2014-04-01T00:25:04Z ccl-logbot joined #sbcl 2014-04-01T00:25:04Z 2014-04-01T00:25:04Z names: ccl-logbot psilord drmeister eudoxia ehaliewicz oleo harovali redline6561 Bicyclidine ASau DGASAU hlavaty nyef asedeno jdz luis nicdev irsol prxq christoph_debian tmh_ kludge` xymox jaimef fikusz Quadrescence bege hzp joshe loke ferada pkhuong |3b| ams @Krystof leoc fe[nl]ix Munksgaard pjb flip214 Bike ivan`` faheem yauz Guest50092 Posterdati milosn foom minion specbot pchrist daimrod Blkt White_Flame scymtym_ antoszka jsnell_ brucem mood 2014-04-01T00:26:22Z slyrus joined #sbcl 2014-04-01T00:58:59Z ltbarcly joined #sbcl 2014-04-01T01:09:57Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-01T01:15:39Z ccl-logbot joined #sbcl 2014-04-01T01:15:39Z 2014-04-01T01:15:39Z names: ccl-logbot slyrus psilord drmeister eudoxia ehaliewicz oleo harovali redline6561 Bicyclidine ASau DGASAU hlavaty nyef asedeno jdz luis nicdev irsol prxq christoph_debian tmh_ kludge` xymox jaimef fikusz Quadrescence bege hzp joshe loke ferada pkhuong |3b| ams @Krystof leoc fe[nl]ix Munksgaard pjb flip214 Bike ivan`` faheem yauz Guest50092 Posterdati milosn foom minion specbot pchrist daimrod Blkt White_Flame mood brucem jsnell_ antoszka scymtym_ 2014-04-01T01:39:55Z pjb` joined #sbcl 2014-04-01T01:41:56Z pjb quit (Ping timeout: 252 seconds) 2014-04-01T01:50:03Z eudoxia quit (Quit: leaving) 2014-04-01T01:53:59Z Bicyclidine quit (Ping timeout: 252 seconds) 2014-04-01T02:18:07Z krzysz00 joined #sbcl 2014-04-01T02:26:45Z loke_ joined #sbcl 2014-04-01T02:35:19Z ehaliewicz quit (Ping timeout: 264 seconds) 2014-04-01T02:37:05Z christoph_debian quit (Read error: Operation timed out) 2014-04-01T02:39:24Z prxq_ joined #sbcl 2014-04-01T02:42:48Z prxq quit (Ping timeout: 255 seconds) 2014-04-01T02:52:51Z christoph_debian joined #sbcl 2014-04-01T03:07:55Z nyef quit (Quit: G'night all.) 2014-04-01T03:11:16Z echo-area joined #sbcl 2014-04-01T03:25:39Z pjb` is now known as pjb 2014-04-01T03:26:08Z pjb is now known as Guest62906 2014-04-01T03:26:09Z Guest62906 is now known as pjb` 2014-04-01T03:26:25Z pjb` is now known as pjb 2014-04-01T04:01:44Z pranavrc joined #sbcl 2014-04-01T04:26:06Z yacks joined #sbcl 2014-04-01T04:36:27Z attila_lendvai joined #sbcl 2014-04-01T04:42:09Z kanru joined #sbcl 2014-04-01T05:02:12Z oleo quit (Quit: Leaving) 2014-04-01T05:08:32Z drmeister quit (Remote host closed the connection) 2014-04-01T05:25:35Z michael_lee joined #sbcl 2014-04-01T05:27:12Z michael_lee quit (Max SendQ exceeded) 2014-04-01T05:35:44Z sdemarre joined #sbcl 2014-04-01T05:44:30Z michael_lee joined #sbcl 2014-04-01T05:52:57Z michael_lee quit (Max SendQ exceeded) 2014-04-01T05:54:08Z michael_lee joined #sbcl 2014-04-01T06:03:31Z krzysz00 quit (Ping timeout: 245 seconds) 2014-04-01T06:04:34Z kanru quit (Remote host closed the connection) 2014-04-01T06:11:40Z prxq_ is now known as prxq 2014-04-01T06:22:01Z angavrilov joined #sbcl 2014-04-01T06:30:31Z echo-area quit (Ping timeout: 255 seconds) 2014-04-01T06:45:18Z sdemarre left #sbcl 2014-04-01T06:53:13Z ASau quit (Remote host closed the connection) 2014-04-01T06:54:39Z ASau joined #sbcl 2014-04-01T06:57:22Z drmeister joined #sbcl 2014-04-01T07:01:44Z drmeister quit (Ping timeout: 246 seconds) 2014-04-01T07:03:31Z ASau quit (Ping timeout: 245 seconds) 2014-04-01T07:28:04Z yacks quit (Quit: Leaving) 2014-04-01T07:47:16Z yacks joined #sbcl 2014-04-01T08:10:14Z antoszka quit (Ping timeout: 252 seconds) 2014-04-01T08:14:46Z harovali quit (Ping timeout: 245 seconds) 2014-04-01T08:15:53Z harovali joined #sbcl 2014-04-01T08:28:48Z antoszka joined #sbcl 2014-04-01T08:46:09Z drmeister joined #sbcl 2014-04-01T08:50:36Z drmeister quit (Ping timeout: 245 seconds) 2014-04-01T09:55:26Z kludge` quit (Ping timeout: 252 seconds) 2014-04-01T09:57:55Z kludge` joined #sbcl 2014-04-01T10:02:27Z edgar-rft joined #sbcl 2014-04-01T10:35:01Z drmeister joined #sbcl 2014-04-01T10:39:51Z drmeister quit (Ping timeout: 268 seconds) 2014-04-01T11:00:27Z krzysz00 joined #sbcl 2014-04-01T11:01:12Z edgar-rft quit (Quit: happening vanished by nuclear disaster) 2014-04-01T11:15:25Z yacks quit (Read error: Connection reset by peer) 2014-04-01T11:21:12Z pranavrc_ joined #sbcl 2014-04-01T11:23:11Z pranavrc quit (Ping timeout: 246 seconds) 2014-04-01T12:11:01Z attila_lendvai quit (Quit: Leaving.) 2014-04-01T12:28:01Z drmeister joined #sbcl 2014-04-01T12:29:45Z pranavrc_ quit (Remote host closed the connection) 2014-04-01T12:30:04Z pranavrc joined #sbcl 2014-04-01T12:36:11Z drmeister quit (Remote host closed the connection) 2014-04-01T12:36:23Z pranavrc quit 2014-04-01T12:36:39Z attila_lendvai joined #sbcl 2014-04-01T12:45:08Z krzysz00 quit (Ping timeout: 240 seconds) 2014-04-01T12:59:07Z psilord quit (Quit: Leaving.) 2014-04-01T13:19:39Z eudoxia joined #sbcl 2014-04-01T13:31:55Z drmeister joined #sbcl 2014-04-01T13:35:12Z oleo joined #sbcl 2014-04-01T13:36:31Z drmeister quit (Ping timeout: 264 seconds) 2014-04-01T13:46:50Z edgar-rft joined #sbcl 2014-04-01T14:17:40Z yacks joined #sbcl 2014-04-01T14:23:57Z psilord joined #sbcl 2014-04-01T15:28:06Z slyrus quit (Ping timeout: 245 seconds) 2014-04-01T15:44:53Z jdz quit (Quit: Leaving) 2014-04-01T16:05:18Z hlavaty: hi, what does "failed AVER: (FUNCTIONP INITFUN)" mean? 2014-04-01T16:06:13Z drmeister joined #sbcl 2014-04-01T16:14:15Z Krystof: "you shouldn't be able to make this happen" 2014-04-01T16:14:38Z jdz joined #sbcl 2014-04-01T16:14:40Z Krystof: in this instance, it could mean "you have confused CLOS by messing around with the MOP" 2014-04-01T16:14:49Z Krystof: or it could be something like that but different 2014-04-01T16:16:02Z hlavaty: ok, unfortunatelly it happens 2014-04-01T16:16:56Z Krystof: yes 2014-04-01T16:17:00Z Krystof: how? :) 2014-04-01T16:17:13Z hlavaty: is that clos internal stuff or lisp compiler? 2014-04-01T16:17:17Z hlavaty: i dont know yet 2014-04-01T16:19:22Z slyrus joined #sbcl 2014-04-01T16:34:09Z hlavaty: Krystof: HEAD: (ERROR 'SB-PCL::SLOTD-INITIALIZATION-ERROR :INITARG :INITFORM :KIND :MISSING) 2014-04-01T16:36:06Z hlavaty: whatever is going on, it should throw error, not that AVER, right? 2014-04-01T16:42:05Z |3b|: AVER usually means compiler bug or you did something very non-conformant... normal code shouldn't cause it 2014-04-01T16:42:28Z hlavaty: that simple form causes it, shall i file a bug? 2014-04-01T16:43:31Z hlavaty: both x86 and x86_64 2014-04-01T16:43:58Z |3b|: is that the actual code, or part of an expansion of something? 2014-04-01T16:44:18Z hlavaty: from a stacktrace 2014-04-01T16:44:54Z |3b|: is that frame in sbcl source? 2014-04-01T16:44:57Z hlavaty: it seems to me it is trying to report an error and fails with AVER failure 2014-04-01T16:45:39Z drmeister quit (Ping timeout: 255 seconds) 2014-04-01T16:45:55Z |3b|: yeah, does sound like something is wrong there 2014-04-01T16:47:27Z |3b|: works in 1.0.48.25, aver in 1.1.13.24 2014-04-01T16:48:34Z hlavaty: the original error is not, not sure what is actually going on as it is some hu.dwim stuff, but that particula form is in pcl/init.lisp:141 2014-04-01T16:48:46Z |3b|: actually, maybe doesn't work in 1.0.48... it is behaving oddly 2014-04-01T16:49:19Z |3b|: ok, just making sure it isn't user-code calling sb-pcl:: stuff 2014-04-01T16:50:48Z |3b|: hmm, this 1.0.48 doesn't seem to error on any error form, so it might not be a useful reference 2014-04-01T16:54:09Z hlavaty: in 1.0.58 i get "Lock on package SB-PCL violated when interning SLOTD-INITIALIZATION-ERROR while in package COMMON-LISP-USER." so it's probably something new 2014-04-01T16:54:18Z |3b|: yeah, was just noticing that 2014-04-01T16:54:21Z attila_lendvai: this is enough for me in itself: (ERROR 'SB-PCL::SLOTD-INITIALIZATION-ERROR) 2014-04-01T16:55:16Z hlavaty: attila_lendvai: hi, yeah there are two issues thogh, the bug is that reporting the error fails badly 2014-04-01T16:55:20Z |3b| would say file a bug, with more of the backtrace if possible 2014-04-01T16:57:41Z attila_lendvai: hi! I see you're still pushing things forward... 2014-04-01T16:58:43Z yacks quit (Ping timeout: 264 seconds) 2014-04-01T17:01:28Z attila_lendvai: if I change the 'value slot''s initform from nil to :initform 42, then the error goes away 2014-04-01T17:01:49Z attila_lendvai: some code somewhere does a nil-check on the initform and ignores it wrongly 2014-04-01T17:02:19Z attila_lendvai: thus no initform and initfunction is recorded for the 'value slot of slotd-initialization-error 2014-04-01T17:20:22Z hlavaty: it fails in find-slot-default, code/condition.lisp:192 2014-04-01T17:20:29Z hlavaty: i suspect 885a956ae2044a0c5c4a2b55db8e32b7f6b48d05 broke it 2014-04-01T17:21:46Z hlavaty: it for some reason expects that condition-slot-initfunction is not nil and is a function 2014-04-01T17:28:43Z slyrus quit (Ping timeout: 268 seconds) 2014-04-01T17:35:46Z attila_lendvai: hlavaty: read what I wrote above. something somewhere at definition time ignores the :initform nil specification on the slot called 'value, that should be turned into a (constantly nil) 2014-04-01T17:37:22Z hlavaty: attila_lendvai: yeah, but i dont understand why 885a956ae2044a0c5c4a2b55db8e32b7f6b48d05 changed it that way :-) 2014-04-01T17:37:44Z attila_lendvai: ah, ok 2014-04-01T17:38:52Z milosn quit (Read error: No route to host) 2014-04-01T17:39:03Z milosn joined #sbcl 2014-04-01T17:40:36Z michael_lee quit (Ping timeout: 245 seconds) 2014-04-01T17:42:08Z attila_lendvai: if I guard this not to push the slot in question into the "hairy" slots (whatever they are), then it works: (push slot (condition-classoid-hairy-slots class)) 2014-04-01T17:42:24Z attila_lendvai: after C-c C-c ing the defcondition 2014-04-01T17:43:05Z attila_lendvai: the comment states: ;; a list of all the effective instance allocation slots of this class that have a non-constant initform or default-initarg. 2014-04-01T17:43:47Z |3b| thought C-c C-c the defcondition makes it work without changes 2014-04-01T17:46:29Z michael_lee joined #sbcl 2014-04-01T17:46:45Z attila_lendvai: not for me, but I'm distracted, so I may have made a mistake 2014-04-01T17:47:03Z |3b|: on slotd-initialization-error? 2014-04-01T17:47:24Z attila_lendvai: it looks like define-condition makes a decision based on (functionp (condition-slot-initfunction effective-slot)), but that's true even for :initform nil 2014-04-01T17:47:34Z attila_lendvai: yes 2014-04-01T17:47:38Z attila_lendvai double checks 2014-04-01T17:48:11Z |3b|: yeah, c-c c-c complains about readers being normal functions, but after overriding them it errors instead of aver 2014-04-01T17:48:16Z attila_lendvai: err, I was wrong, sorry. a simple redefinition is enough. 2014-04-01T17:48:31Z attila_lendvai: btw, my test form is a simple (make-condition 'SB-PCL::SLOTD-INITIALIZATION-ERROR) 2014-04-01T17:48:41Z |3b|: yeah, same 2014-04-01T17:50:36Z attila_lendvai: the (aver (functionp initfun)) and the fact that there's a conditional on (functionp (condition-slot-initfunction slot)) in define-condition smells contradictory to me 2014-04-01T17:51:33Z attila_lendvai: as if someone made an optimization not to generate an initfuncion for constants, but forgot to update the place where the aver is currently 2014-04-01T17:54:17Z attila_lendvai needs to give up looking for it 2014-04-01T17:55:23Z attila_lendvai writes a comment on the bug 2014-04-01T18:16:58Z eudoxia quit (Quit: leaving) 2014-04-01T18:20:52Z attila_lendvai quit (Quit: Leaving.) 2014-04-01T18:21:35Z tmh_ quit (Ping timeout: 255 seconds) 2014-04-01T18:38:00Z ltbarcly joined #sbcl 2014-04-01T18:42:54Z slyrus joined #sbcl 2014-04-01T18:45:31Z Krystof: hlavaty: I think I see it. Typo. 2014-04-01T18:47:17Z Krystof: oh, stassats has already seen it 2014-04-01T19:19:43Z tmh_ joined #sbcl 2014-04-01T19:21:05Z krzysz00 joined #sbcl 2014-04-01T19:28:38Z ASau joined #sbcl 2014-04-01T19:34:05Z fiveop joined #sbcl 2014-04-01T19:38:30Z davazp joined #sbcl 2014-04-01T19:59:38Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-01T20:04:57Z angavrilov quit (Remote host closed the connection) 2014-04-01T20:06:29Z slyrus quit (Read error: Connection reset by peer) 2014-04-01T20:15:28Z harovali1 joined #sbcl 2014-04-01T20:17:00Z drmeister joined #sbcl 2014-04-01T20:17:15Z slyrus joined #sbcl 2014-04-01T20:17:16Z harovali quit (Ping timeout: 245 seconds) 2014-04-01T20:34:50Z scymtym joined #sbcl 2014-04-01T20:58:39Z drmeister quit (Remote host closed the connection) 2014-04-01T21:05:02Z drmeister joined #sbcl 2014-04-01T21:06:46Z ams quit (Ping timeout: 265 seconds) 2014-04-01T21:11:05Z Munksgaard quit (Ping timeout: 246 seconds) 2014-04-01T21:11:24Z Munksgaard joined #sbcl 2014-04-01T21:26:10Z drmeister quit (Remote host closed the connection) 2014-04-01T21:31:24Z ltbarcly joined #sbcl 2014-04-01T21:42:37Z irsol quit (Ping timeout: 265 seconds) 2014-04-01T21:54:43Z jaimef quit (Ping timeout: 255 seconds) 2014-04-01T21:57:38Z drmeister joined #sbcl 2014-04-01T21:59:00Z jaimef joined #sbcl 2014-04-01T22:10:43Z drmeister quit (Read error: Connection reset by peer) 2014-04-01T22:11:22Z drmeister joined #sbcl 2014-04-01T22:17:28Z irsol joined #sbcl 2014-04-01T22:23:37Z ltbarcly quit (Ping timeout: 265 seconds) 2014-04-01T22:25:22Z ltbarcly joined #sbcl 2014-04-01T22:32:27Z drmeister quit (Read error: Connection reset by peer) 2014-04-01T22:33:58Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-01T22:46:43Z oleo quit (Ping timeout: 264 seconds) 2014-04-01T22:47:24Z drmeister joined #sbcl 2014-04-01T22:47:30Z oleo joined #sbcl 2014-04-01T22:53:00Z ltbarcly joined #sbcl 2014-04-01T22:59:29Z eudoxia joined #sbcl 2014-04-01T23:17:08Z ccl-logbot joined #sbcl 2014-04-01T23:17:08Z 2014-04-01T23:17:08Z names: ccl-logbot eudoxia ltbarcly oleo drmeister irsol jaimef Munksgaard scymtym slyrus harovali1 fiveop ASau krzysz00 tmh_ michael_lee milosn jdz edgar-rft kludge` antoszka christoph_debian prxq loke_ pjb redline6561 DGASAU hlavaty asedeno luis nicdev xymox fikusz Quadrescence bege hzp joshe loke ferada pkhuong |3b| @Krystof leoc fe[nl]ix flip214 Bike ivan`` faheem yauz Guest50092 Posterdati foom minion specbot mood brucem jsnell_ scymtym_ White_Flame Blkt 2014-04-01T23:17:08Z names: daimrod pchrist 2014-04-01T23:18:38Z ltbarcly quit (Ping timeout: 240 seconds) 2014-04-01T23:22:37Z ltbarcly joined #sbcl 2014-04-01T23:23:16Z ltbarcly quit (Client Quit) 2014-04-01T23:23:35Z ltbarcly joined #sbcl 2014-04-01T23:28:03Z ltbarcly quit (Ping timeout: 255 seconds) 2014-04-01T23:31:50Z ltbarcly joined #sbcl 2014-04-01T23:39:45Z slyrus quit (Ping timeout: 255 seconds) 2014-04-01T23:55:57Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-02T00:01:52Z psilord joined #sbcl 2014-04-02T00:32:24Z ltbarcly joined #sbcl 2014-04-02T00:37:38Z ltbarcly quit (Ping timeout: 240 seconds) 2014-04-02T00:40:41Z ltbarcly joined #sbcl 2014-04-02T00:45:14Z fiveop quit (Quit: leaving) 2014-04-02T00:52:26Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-02T00:56:00Z ltbarcly joined #sbcl 2014-04-02T00:58:50Z edgar-rft quit (Quit: existence destroyed by permanent disaster) 2014-04-02T01:05:11Z eudoxia quit (Quit: leaving) 2014-04-02T01:32:20Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-02T01:33:27Z White_Flame quit (Read error: Connection reset by peer) 2014-04-02T01:33:57Z prxq_ joined #sbcl 2014-04-02T01:34:29Z prxq quit (Read error: Operation timed out) 2014-04-02T01:34:45Z White_Flame joined #sbcl 2014-04-02T02:01:41Z loke_ quit (Remote host closed the connection) 2014-04-02T02:02:23Z krzysz00 quit (Ping timeout: 268 seconds) 2014-04-02T02:19:18Z kanru joined #sbcl 2014-04-02T02:21:11Z jaimef quit (Excess Flood) 2014-04-02T02:27:59Z jaimef joined #sbcl 2014-04-02T02:33:02Z loke_ joined #sbcl 2014-04-02T02:39:23Z christoph_debian quit (Ping timeout: 268 seconds) 2014-04-02T02:43:00Z loke_ quit (Remote host closed the connection) 2014-04-02T02:52:26Z christoph_debian joined #sbcl 2014-04-02T02:52:43Z krzysz00 joined #sbcl 2014-04-02T02:53:31Z Quadrescence quit (Quit: This computer has gone to sleep) 2014-04-02T02:55:25Z slyrus joined #sbcl 2014-04-02T03:30:03Z slyrus quit (Ping timeout: 265 seconds) 2014-04-02T03:39:09Z yacks joined #sbcl 2014-04-02T04:09:03Z slyrus joined #sbcl 2014-04-02T04:31:40Z loke_ joined #sbcl 2014-04-02T04:51:08Z slyrus quit (Ping timeout: 240 seconds) 2014-04-02T05:03:54Z oleo quit (Quit: Leaving) 2014-04-02T05:13:55Z scymtym quit (Ping timeout: 240 seconds) 2014-04-02T05:23:59Z attila_lendvai joined #sbcl 2014-04-02T05:41:55Z slyrus joined #sbcl 2014-04-02T05:48:52Z pranavrc joined #sbcl 2014-04-02T05:54:32Z angavrilov joined #sbcl 2014-04-02T06:04:43Z krzysz00 quit (Ping timeout: 264 seconds) 2014-04-02T06:58:02Z ASau quit (Ping timeout: 255 seconds) 2014-04-02T08:06:25Z c4h joined #sbcl 2014-04-02T08:15:11Z ams joined #sbcl 2014-04-02T08:16:04Z ams quit (Changing host) 2014-04-02T08:16:04Z ams joined #sbcl 2014-04-02T08:18:17Z harovali1 quit (Ping timeout: 246 seconds) 2014-04-02T08:20:50Z harovali joined #sbcl 2014-04-02T08:21:04Z yacks quit (Quit: Leaving) 2014-04-02T08:51:50Z yacks joined #sbcl 2014-04-02T08:53:21Z pranavrc quit (Remote host closed the connection) 2014-04-02T09:09:44Z pranavrc joined #sbcl 2014-04-02T09:09:47Z pranavrc quit (Changing host) 2014-04-02T09:09:47Z pranavrc joined #sbcl 2014-04-02T09:12:58Z DGASAU quit (Remote host closed the connection) 2014-04-02T09:17:18Z DGASAU joined #sbcl 2014-04-02T09:33:45Z kludge` quit (Ping timeout: 255 seconds) 2014-04-02T09:36:28Z kludge` joined #sbcl 2014-04-02T09:48:55Z hlavaty: Krystof & stassats: thank you! it works now 2014-04-02T09:59:05Z Krystof: sorry for the error 2014-04-02T10:09:35Z hlavaty: no problem 2014-04-02T10:14:33Z hlavaty: Krystof: in pcl/init.lisp:153 does (and initfunp (not initformp)) need to be an error? i get this failure only on sbcl, acl and ccl dont throw the error in this case 2014-04-02T10:15:47Z Krystof: The :initform argument is a form. The :initform argument defaults to nil. An error is signaled if the :initform argument is supplied, but the :initfunction argument is not supplied. 2014-04-02T10:15:48Z Krystof: The :initfunction argument is a function of zero arguments which, when called, evaluates the :initform in the appropriate lexical environment. The :initfunction argument defaults to false. An error is signaled if the :initfunction argument is supplied, but the :initform argument is not supplied. 2014-04-02T10:15:52Z Krystof: (says the MOP) 2014-04-02T10:16:37Z hlavaty: ok thanks! 2014-04-02T10:18:40Z hlavaty: i can see it in the spec now :-) 2014-04-02T10:55:47Z prxq_ is now known as prxq 2014-04-02T11:01:07Z krzysz00 joined #sbcl 2014-04-02T11:40:51Z angavrilov quit (Remote host closed the connection) 2014-04-02T11:44:13Z angavrilov joined #sbcl 2014-04-02T11:55:58Z antoszka quit (Read error: Operation timed out) 2014-04-02T12:00:51Z antoszka joined #sbcl 2014-04-02T12:02:18Z drmeister quit (Remote host closed the connection) 2014-04-02T12:05:31Z drmeister joined #sbcl 2014-04-02T12:14:06Z drmeister quit (Remote host closed the connection) 2014-04-02T12:21:41Z drmeister joined #sbcl 2014-04-02T12:36:01Z drmeister quit (Remote host closed the connection) 2014-04-02T12:41:08Z pranavrc quit 2014-04-02T12:46:58Z edgar-rft joined #sbcl 2014-04-02T13:00:59Z eudoxia joined #sbcl 2014-04-02T13:02:54Z yacks quit (Ping timeout: 255 seconds) 2014-04-02T13:24:47Z jdz quit (Quit: Leaving) 2014-04-02T13:25:58Z yacks joined #sbcl 2014-04-02T13:35:47Z eudoxia quit (Quit: leaving) 2014-04-02T13:40:31Z drmeister joined #sbcl 2014-04-02T13:41:12Z drmeister quit (Remote host closed the connection) 2014-04-02T14:10:55Z michael_lee quit (Ping timeout: 240 seconds) 2014-04-02T14:22:51Z pranavrc joined #sbcl 2014-04-02T14:22:51Z pranavrc quit (Changing host) 2014-04-02T14:22:51Z pranavrc joined #sbcl 2014-04-02T14:25:56Z michael_lee joined #sbcl 2014-04-02T14:28:54Z drmeister joined #sbcl 2014-04-02T14:37:33Z oleo joined #sbcl 2014-04-02T14:38:53Z drmeister quit (Read error: Connection reset by peer) 2014-04-02T14:39:33Z drmeister joined #sbcl 2014-04-02T14:41:13Z eudoxia joined #sbcl 2014-04-02T15:19:03Z tmh_ quit (Changing host) 2014-04-02T15:19:03Z tmh_ joined #sbcl 2014-04-02T15:36:19Z eudoxia quit (Quit: leaving) 2014-04-02T15:41:19Z krzysz00 quit (Ping timeout: 268 seconds) 2014-04-02T15:42:49Z ASau joined #sbcl 2014-04-02T15:49:08Z drmeister quit (Ping timeout: 265 seconds) 2014-04-02T16:37:15Z c4h quit (Quit: quit) 2014-04-02T16:37:26Z yacks quit (Ping timeout: 268 seconds) 2014-04-02T16:52:13Z yacks joined #sbcl 2014-04-02T17:27:35Z pranavrc quit (Quit: Ping timeout: ∞) 2014-04-02T17:38:18Z yacks quit (*.net *.split) 2014-04-02T17:38:18Z oleo quit (*.net *.split) 2014-04-02T17:45:25Z yacks joined #sbcl 2014-04-02T17:46:17Z oleo joined #sbcl 2014-04-02T17:52:42Z slyrus quit (Ping timeout: 255 seconds) 2014-04-02T17:58:20Z ltbarcly joined #sbcl 2014-04-02T19:22:30Z fiveop joined #sbcl 2014-04-02T19:32:05Z slyrus joined #sbcl 2014-04-02T19:53:39Z slyrus quit (Ping timeout: 252 seconds) 2014-04-02T20:00:23Z angavrilov quit (Remote host closed the connection) 2014-04-02T20:12:16Z slyrus joined #sbcl 2014-04-02T20:19:23Z harovali quit (Read error: Connection reset by peer) 2014-04-02T20:27:27Z slyrus quit (Ping timeout: 268 seconds) 2014-04-02T20:37:39Z harovali joined #sbcl 2014-04-02T20:54:41Z fiveop quit 2014-04-02T21:17:36Z michael_lee quit (Remote host closed the connection) 2014-04-02T21:21:27Z fiveop joined #sbcl 2014-04-02T21:32:45Z fiveop quit 2014-04-02T21:35:10Z attila_lendvai quit (Quit: Leaving.) 2014-04-02T21:39:40Z yacks quit (Read error: Operation timed out) 2014-04-02T21:39:58Z scymtym joined #sbcl 2014-04-02T22:00:46Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-02T22:06:07Z afleck joined #sbcl 2014-04-02T22:06:32Z afleck: Hi, is it possible to build sbcl on NetBSD amd64? 2014-04-02T22:15:23Z eudoxia joined #sbcl 2014-04-02T22:48:05Z Guest50092: http://sbcl.org/platform-table.html 2014-04-02T22:48:12Z Guest50092 is now known as reb` 2014-04-02T23:09:01Z oleo quit (Ping timeout: 268 seconds) 2014-04-02T23:21:51Z oleo joined #sbcl 2014-04-02T23:46:24Z eudoxia quit (Quit: leaving) 2014-04-03T01:06:18Z michael_lee joined #sbcl 2014-04-03T01:34:27Z prxq_ joined #sbcl 2014-04-03T01:38:00Z prxq quit (Ping timeout: 255 seconds) 2014-04-03T02:18:38Z christoph_debian quit (Ping timeout: 240 seconds) 2014-04-03T02:20:14Z christoph_debian joined #sbcl 2014-04-03T02:34:03Z ltbarcly joined #sbcl 2014-04-03T02:38:45Z christoph_debian quit (Ping timeout: 255 seconds) 2014-04-03T02:38:59Z krzysz00 joined #sbcl 2014-04-03T02:42:09Z ltbarcly quit (Quit: Textual IRC Client: www.textualapp.com) 2014-04-03T02:43:10Z ltbarcly joined #sbcl 2014-04-03T02:51:58Z christoph_debian joined #sbcl 2014-04-03T03:29:32Z edgar-rft quit (Quit: lifeform experiment stopped by sudden death) 2014-04-03T03:47:32Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-03T04:29:43Z ASau quit (Remote host closed the connection) 2014-04-03T04:37:24Z yacks joined #sbcl 2014-04-03T05:02:48Z oleo quit (Quit: Leaving) 2014-04-03T05:14:23Z pranavrc joined #sbcl 2014-04-03T05:14:23Z pranavrc quit (Changing host) 2014-04-03T05:14:23Z pranavrc joined #sbcl 2014-04-03T05:44:01Z slyrus joined #sbcl 2014-04-03T05:46:14Z scymtym quit (Ping timeout: 240 seconds) 2014-04-03T05:51:46Z attila_lendvai joined #sbcl 2014-04-03T06:04:24Z krzysz00 quit (Ping timeout: 255 seconds) 2014-04-03T06:15:03Z prxq_ is now known as prxq 2014-04-03T06:17:40Z angavrilov joined #sbcl 2014-04-03T06:33:38Z tmh_ quit (Ping timeout: 268 seconds) 2014-04-03T07:05:09Z jaimef quit (Ping timeout: 255 seconds) 2014-04-03T07:12:49Z ASau joined #sbcl 2014-04-03T07:19:23Z jaimef joined #sbcl 2014-04-03T07:23:20Z tmh_ joined #sbcl 2014-04-03T07:27:39Z yacks quit (Ping timeout: 255 seconds) 2014-04-03T07:54:58Z ASau quit (Ping timeout: 240 seconds) 2014-04-03T07:57:15Z antoszka quit (Quit: +++ killed by SIGSEGV +++) 2014-04-03T08:14:01Z jdz joined #sbcl 2014-04-03T08:23:54Z harovali quit (Ping timeout: 240 seconds) 2014-04-03T08:39:52Z harovali joined #sbcl 2014-04-03T08:58:14Z yacks joined #sbcl 2014-04-03T09:08:04Z antoszka joined #sbcl 2014-04-03T09:21:14Z jdz quit (Quit: Leaving) 2014-04-03T09:33:32Z kludge` quit (Ping timeout: 252 seconds) 2014-04-03T09:36:25Z kludge` joined #sbcl 2014-04-03T09:45:27Z davazp joined #sbcl 2014-04-03T11:01:10Z krzysz00 joined #sbcl 2014-04-03T11:14:07Z jdz joined #sbcl 2014-04-03T11:28:05Z eudoxia joined #sbcl 2014-04-03T11:30:12Z eudoxia quit (Client Quit) 2014-04-03T11:31:14Z davazp quit (Ping timeout: 240 seconds) 2014-04-03T11:33:06Z ltbarcly joined #sbcl 2014-04-03T12:01:24Z davazp joined #sbcl 2014-04-03T12:18:13Z LiamH joined #sbcl 2014-04-03T12:46:01Z krzysz00 quit (Ping timeout: 240 seconds) 2014-04-03T13:05:00Z LiamH quit (Quit: Leaving.) 2014-04-03T13:09:59Z pranavrc quit 2014-04-03T13:16:50Z sdemarre joined #sbcl 2014-04-03T13:21:49Z segv- joined #sbcl 2014-04-03T13:29:54Z davazp quit (Ping timeout: 240 seconds) 2014-04-03T13:30:08Z ltbarcly_ joined #sbcl 2014-04-03T13:31:51Z ltbarcly_ quit (Client Quit) 2014-04-03T13:32:58Z ltbarcly quit (Ping timeout: 240 seconds) 2014-04-03T13:40:36Z nyef joined #sbcl 2014-04-03T13:40:49Z nyef: G'morning all. 2014-04-03T13:49:32Z oleo joined #sbcl 2014-04-03T13:51:58Z ltbarcly joined #sbcl 2014-04-03T13:54:00Z hlavaty quit (Remote host closed the connection) 2014-04-03T14:23:13Z slyrus quit (Ping timeout: 240 seconds) 2014-04-03T14:33:04Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-03T14:41:24Z jdz quit (Quit: Leaving) 2014-04-03T14:56:08Z sdemarre quit (Ping timeout: 240 seconds) 2014-04-03T15:10:04Z slyrus joined #sbcl 2014-04-03T15:19:20Z attila_lendvai quit (Quit: Leaving.) 2014-04-03T15:19:28Z attila_lendvai joined #sbcl 2014-04-03T15:19:28Z attila_lendvai quit (Changing host) 2014-04-03T15:19:28Z attila_lendvai joined #sbcl 2014-04-03T16:02:52Z sdemarre joined #sbcl 2014-04-03T16:22:37Z ltbarcly joined #sbcl 2014-04-03T16:25:13Z sdemarre quit (Ping timeout: 240 seconds) 2014-04-03T16:27:50Z ltbarcly quit (Ping timeout: 252 seconds) 2014-04-03T16:31:27Z ltbarcly joined #sbcl 2014-04-03T17:06:58Z slyrus quit (Ping timeout: 240 seconds) 2014-04-03T17:07:58Z sdemarre joined #sbcl 2014-04-03T17:26:05Z fiveop joined #sbcl 2014-04-03T17:26:11Z ltbarcly quit (Read error: Operation timed out) 2014-04-03T17:26:46Z ltbarcly joined #sbcl 2014-04-03T17:39:34Z ltbarcly quit (Ping timeout: 240 seconds) 2014-04-03T17:41:41Z ltbarcly joined #sbcl 2014-04-03T17:46:31Z ltbarcly_ joined #sbcl 2014-04-03T17:47:24Z ltbarcly quit (Ping timeout: 252 seconds) 2014-04-03T18:01:42Z ltbarcly_ quit (Ping timeout: 255 seconds) 2014-04-03T18:04:35Z ltbarcly joined #sbcl 2014-04-03T18:25:25Z |3b| quit (Read error: Connection reset by peer) 2014-04-03T18:26:15Z |3b| joined #sbcl 2014-04-03T18:31:00Z angavrilov quit (Remote host closed the connection) 2014-04-03T18:37:45Z attila_lendvai quit (Quit: Leaving.) 2014-04-03T18:37:57Z attila_lendvai joined #sbcl 2014-04-03T18:37:57Z attila_lendvai quit (Changing host) 2014-04-03T18:37:57Z attila_lendvai joined #sbcl 2014-04-03T18:51:32Z krzysz00 joined #sbcl 2014-04-03T18:52:48Z slyrus joined #sbcl 2014-04-03T19:09:12Z ltbarcly quit (Ping timeout: 255 seconds) 2014-04-03T19:09:48Z michael_lee quit (Remote host closed the connection) 2014-04-03T19:11:02Z ltbarcly joined #sbcl 2014-04-03T19:21:36Z ASau joined #sbcl 2014-04-03T19:25:24Z andreh joined #sbcl 2014-04-03T19:35:18Z sdemarre quit (Ping timeout: 240 seconds) 2014-04-03T19:46:41Z edgar-rft joined #sbcl 2014-04-03T20:16:53Z andreh quit (Ping timeout: 268 seconds) 2014-04-03T20:26:36Z harovali quit (Ping timeout: 255 seconds) 2014-04-03T20:26:55Z harovali joined #sbcl 2014-04-03T20:50:59Z tmh_ quit (Changing host) 2014-04-03T20:50:59Z tmh_ joined #sbcl 2014-04-03T20:58:23Z andreh joined #sbcl 2014-04-03T20:58:53Z andreh quit (Client Quit) 2014-04-03T21:19:37Z attila_lendvai quit (Ping timeout: 240 seconds) 2014-04-03T21:20:28Z attila_lendvai joined #sbcl 2014-04-03T21:34:46Z fiveop quit 2014-04-03T21:36:12Z |3b| quit (Remote host closed the connection) 2014-04-03T21:36:20Z |3b| joined #sbcl 2014-04-03T21:41:58Z cmack joined #sbcl 2014-04-03T21:54:51Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-03T21:58:01Z attila_lendvai quit (Ping timeout: 240 seconds) 2014-04-03T22:02:46Z Bicyclidine joined #sbcl 2014-04-03T22:11:41Z nyef quit (Quit: G'night all.) 2014-04-03T22:27:14Z krzysz00 quit (Ping timeout: 240 seconds) 2014-04-03T23:22:33Z segv- quit (Ping timeout: 255 seconds) 2014-04-04T00:34:49Z eudoxia joined #sbcl 2014-04-04T00:36:38Z Bicyclidine quit (Ping timeout: 240 seconds) 2014-04-04T00:39:32Z brucem: foom: you around? I got the CVS history for Goo and have a couple of questions for you. (easy ones) 2014-04-04T00:42:03Z oleo quit (Ping timeout: 268 seconds) 2014-04-04T00:48:50Z slyrus quit (Ping timeout: 268 seconds) 2014-04-04T00:55:00Z oleo joined #sbcl 2014-04-04T01:20:24Z eudoxia quit (Quit: leaving) 2014-04-04T01:35:02Z prxq_ joined #sbcl 2014-04-04T01:38:08Z prxq quit (Ping timeout: 240 seconds) 2014-04-04T01:40:40Z edgar-rft quit (Quit: conversation disappeared because no future found) 2014-04-04T02:38:34Z christoph_debian quit (Read error: Operation timed out) 2014-04-04T02:52:14Z drmeister joined #sbcl 2014-04-04T02:52:45Z christoph_debian joined #sbcl 2014-04-04T03:12:12Z dimkras joined #sbcl 2014-04-04T03:44:14Z ehaliewicz joined #sbcl 2014-04-04T04:02:33Z dimkras quit (Remote host closed the connection) 2014-04-04T04:21:58Z difortran joined #sbcl 2014-04-04T04:40:18Z drmeister quit (Ping timeout: 240 seconds) 2014-04-04T05:05:29Z oleo quit (Quit: Leaving) 2014-04-04T05:13:03Z attila_lendvai joined #sbcl 2014-04-04T05:34:02Z sdemarre joined #sbcl 2014-04-04T05:35:49Z pranavrc joined #sbcl 2014-04-04T05:39:44Z angavrilov joined #sbcl 2014-04-04T06:01:08Z michael_lee joined #sbcl 2014-04-04T06:05:34Z michael_lee quit (Max SendQ exceeded) 2014-04-04T06:23:12Z michael_lee joined #sbcl 2014-04-04T06:33:10Z michael_lee quit (Max SendQ exceeded) 2014-04-04T06:50:29Z michael_lee joined #sbcl 2014-04-04T06:57:00Z difortran quit (Remote host closed the connection) 2014-04-04T07:24:10Z prxq_ is now known as prxq 2014-04-04T07:27:54Z ASau quit (Ping timeout: 240 seconds) 2014-04-04T07:31:25Z sdemarre quit (Read error: Connection reset by peer) 2014-04-04T07:31:30Z sdemarre1 joined #sbcl 2014-04-04T07:37:08Z yacks quit (Ping timeout: 240 seconds) 2014-04-04T07:43:08Z sdemarre1 quit (Ping timeout: 240 seconds) 2014-04-04T08:03:09Z yacks joined #sbcl 2014-04-04T08:28:52Z harovali quit (Ping timeout: 268 seconds) 2014-04-04T08:44:40Z harovali joined #sbcl 2014-04-04T08:51:33Z pranavrc quit (Remote host closed the connection) 2014-04-04T09:23:54Z pranavrc joined #sbcl 2014-04-04T09:23:55Z pranavrc quit (Changing host) 2014-04-04T09:23:55Z pranavrc joined #sbcl 2014-04-04T09:24:08Z pranavrc quit (Remote host closed the connection) 2014-04-04T09:24:21Z pranavrc joined #sbcl 2014-04-04T09:24:35Z sdemarre joined #sbcl 2014-04-04T09:42:10Z pranavrc_ joined #sbcl 2014-04-04T09:43:08Z pranavrc quit (Ping timeout: 240 seconds) 2014-04-04T09:56:54Z michael_lee quit (Ping timeout: 240 seconds) 2014-04-04T10:10:17Z michael_lee joined #sbcl 2014-04-04T10:21:19Z michael_lee quit (Max SendQ exceeded) 2014-04-04T10:40:30Z michael_lee joined #sbcl 2014-04-04T10:43:20Z michael_lee quit (Excess Flood) 2014-04-04T10:46:58Z ehaliewicz quit (Ping timeout: 240 seconds) 2014-04-04T10:50:12Z segv- joined #sbcl 2014-04-04T10:52:56Z DGASAU quit (Ping timeout: 255 seconds) 2014-04-04T11:04:01Z loke quit (Ping timeout: 240 seconds) 2014-04-04T11:14:00Z eudoxia joined #sbcl 2014-04-04T11:15:56Z loke joined #sbcl 2014-04-04T11:16:32Z DGASAU joined #sbcl 2014-04-04T11:31:01Z DGASAU quit (Remote host closed the connection) 2014-04-04T11:35:23Z DGASAU joined #sbcl 2014-04-04T11:44:08Z DGASAU quit (Ping timeout: 240 seconds) 2014-04-04T11:55:46Z DGASAU joined #sbcl 2014-04-04T12:09:48Z edgar-rft joined #sbcl 2014-04-04T12:18:01Z scymtym joined #sbcl 2014-04-04T12:25:26Z jdz joined #sbcl 2014-04-04T12:30:21Z oleo joined #sbcl 2014-04-04T12:32:34Z yacks quit (Quit: Leaving) 2014-04-04T12:36:02Z pranavrc_ quit (Remote host closed the connection) 2014-04-04T12:39:20Z pranavrc_ joined #sbcl 2014-04-04T12:40:27Z sdemarre quit (Ping timeout: 268 seconds) 2014-04-04T12:51:24Z attila_lendvai quit (Quit: Leaving.) 2014-04-04T12:53:52Z jdz quit (Quit: Leaving) 2014-04-04T13:02:34Z pranavrc_ quit 2014-04-04T13:17:29Z yacks joined #sbcl 2014-04-04T13:28:14Z oleo quit (Ping timeout: 240 seconds) 2014-04-04T13:36:24Z eudoxia quit (Quit: leaving) 2014-04-04T13:41:32Z oleo joined #sbcl 2014-04-04T14:48:08Z scymtym quit (Ping timeout: 240 seconds) 2014-04-04T15:01:05Z nyef joined #sbcl 2014-04-04T15:01:14Z nyef: G'morning all. 2014-04-04T15:07:42Z oleo: morning 2014-04-04T15:08:25Z sdemarre joined #sbcl 2014-04-04T15:16:28Z sdemarre quit (Ping timeout: 268 seconds) 2014-04-04T15:16:58Z Krystof quit (Read error: Operation timed out) 2014-04-04T15:37:46Z sdemarre joined #sbcl 2014-04-04T15:38:58Z Krystof joined #sbcl 2014-04-04T15:38:58Z ChanServ has set mode +o Krystof 2014-04-04T16:15:15Z milanj joined #sbcl 2014-04-04T16:43:51Z sdemarre quit (Ping timeout: 255 seconds) 2014-04-04T17:17:31Z slyrus joined #sbcl 2014-04-04T17:23:58Z milanj quit (Quit: Leaving) 2014-04-04T17:42:02Z sdemarre joined #sbcl 2014-04-04T17:49:04Z ltbarcly joined #sbcl 2014-04-04T18:03:23Z ASau joined #sbcl 2014-04-04T18:24:14Z sdemarre quit (Ping timeout: 240 seconds) 2014-04-04T18:42:37Z milosn quit (Read error: No route to host) 2014-04-04T18:42:54Z milosn joined #sbcl 2014-04-04T19:11:00Z irsol quit (Ping timeout: 252 seconds) 2014-04-04T19:18:35Z irsol joined #sbcl 2014-04-04T19:21:07Z slyrus quit (Ping timeout: 252 seconds) 2014-04-04T19:24:55Z slyrus joined #sbcl 2014-04-04T19:45:47Z davazp joined #sbcl 2014-04-04T20:27:38Z yacks quit (Ping timeout: 240 seconds) 2014-04-04T20:32:49Z Bicyclidine joined #sbcl 2014-04-04T20:33:14Z harovali quit (Ping timeout: 240 seconds) 2014-04-04T20:47:41Z harovali joined #sbcl 2014-04-04T21:01:15Z nyef quit (Quit: G'night all.) 2014-04-04T21:08:17Z ehaliewicz joined #sbcl 2014-04-04T21:18:46Z davazp quit (Remote host closed the connection) 2014-04-04T22:36:23Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-04T22:43:13Z slyrus quit (Ping timeout: 240 seconds) 2014-04-04T22:44:30Z ltbarcly joined #sbcl 2014-04-04T22:45:12Z slyrus joined #sbcl 2014-04-04T23:16:38Z Bicyclidine quit (Ping timeout: 240 seconds) 2014-04-04T23:35:55Z Bike quit (Quit: leaving) 2014-04-04T23:37:53Z Bike joined #sbcl 2014-04-05T01:11:27Z eudoxia joined #sbcl 2014-04-05T01:31:40Z davazp joined #sbcl 2014-04-05T01:37:43Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-05T01:38:54Z prxq quit (Ping timeout: 240 seconds) 2014-04-05T01:51:33Z davazp quit (Remote host closed the connection) 2014-04-05T01:51:39Z prxq joined #sbcl 2014-04-05T02:05:36Z ltbarcly joined #sbcl 2014-04-05T02:20:49Z oleo quit (Ping timeout: 240 seconds) 2014-04-05T02:21:42Z eudoxia quit (Quit: leaving) 2014-04-05T02:21:46Z oleo joined #sbcl 2014-04-05T02:38:54Z christoph_debian quit (Ping timeout: 240 seconds) 2014-04-05T02:52:17Z christoph_debian joined #sbcl 2014-04-05T03:08:15Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-05T03:30:43Z ltbarcly joined #sbcl 2014-04-05T03:54:36Z edgar-rft quit (Quit: mental process disappeared because everything died) 2014-04-05T04:00:17Z yacks joined #sbcl 2014-04-05T04:18:51Z drmeister joined #sbcl 2014-04-05T05:46:52Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-05T06:02:18Z ltbarcly joined #sbcl 2014-04-05T06:05:04Z attila_lendvai joined #sbcl 2014-04-05T06:52:45Z attila_lendvai quit (Quit: Leaving.) 2014-04-05T06:56:28Z ltbarcly quit (Read error: Connection reset by peer) 2014-04-05T06:57:05Z ltbarcly joined #sbcl 2014-04-05T07:55:19Z bishuk joined #sbcl 2014-04-05T07:56:55Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-05T08:02:36Z bishuk quit (Remote host closed the connection) 2014-04-05T08:04:38Z sdemarre joined #sbcl 2014-04-05T08:07:42Z fiveop joined #sbcl 2014-04-05T08:14:47Z pochuta joined #sbcl 2014-04-05T08:24:58Z pochuta quit (Ping timeout: 240 seconds) 2014-04-05T08:35:33Z harovali quit (Ping timeout: 268 seconds) 2014-04-05T08:44:38Z ehaliewicz quit (Ping timeout: 240 seconds) 2014-04-05T08:50:46Z harovali joined #sbcl 2014-04-05T09:26:56Z attila_lendvai joined #sbcl 2014-04-05T09:32:54Z kludge` quit (Ping timeout: 268 seconds) 2014-04-05T09:37:13Z kludge` joined #sbcl 2014-04-05T09:49:42Z michael_lee joined #sbcl 2014-04-05T09:51:19Z michael_lee quit (Max SendQ exceeded) 2014-04-05T09:52:32Z michael_lee joined #sbcl 2014-04-05T10:00:50Z michael_lee quit (Max SendQ exceeded) 2014-04-05T10:03:36Z fiveop quit 2014-04-05T10:06:34Z begemoth joined #sbcl 2014-04-05T10:09:54Z sdemarre quit (Ping timeout: 240 seconds) 2014-04-05T10:19:18Z michael_lee joined #sbcl 2014-04-05T12:14:57Z edgar-rft joined #sbcl 2014-04-05T12:46:30Z sdemarre joined #sbcl 2014-04-05T14:03:18Z LiamH joined #sbcl 2014-04-05T14:05:41Z root_empire joined #sbcl 2014-04-05T14:08:25Z michael_lee quit (Ping timeout: 240 seconds) 2014-04-05T14:10:59Z root_empire quit (Max SendQ exceeded) 2014-04-05T14:14:19Z begemoth quit (Remote host closed the connection) 2014-04-05T14:43:08Z oleo quit (Ping timeout: 240 seconds) 2014-04-05T14:43:30Z oleo joined #sbcl 2014-04-05T14:54:03Z sdemarre quit (Ping timeout: 255 seconds) 2014-04-05T15:19:42Z harovali quit (Ping timeout: 255 seconds) 2014-04-05T15:46:12Z drmeister quit (Remote host closed the connection) 2014-04-05T16:04:01Z Hydan joined #sbcl 2014-04-05T16:07:57Z leuler joined #sbcl 2014-04-05T16:12:19Z scymtym joined #sbcl 2014-04-05T16:15:44Z ltbarcly joined #sbcl 2014-04-05T17:05:00Z harovali joined #sbcl 2014-04-05T17:44:46Z pnpuff joined #sbcl 2014-04-05T17:57:55Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-05T18:19:38Z segv- quit (Ping timeout: 240 seconds) 2014-04-05T18:30:21Z attila_lendvai quit (Quit: Leaving.) 2014-04-05T19:24:53Z pnpuff quit (Remote host closed the connection) 2014-04-05T20:08:19Z Hydan quit (Remote host closed the connection) 2014-04-05T20:16:46Z fikusz quit (Quit: Leaving) 2014-04-05T20:20:26Z fikusz joined #sbcl 2014-04-05T20:25:53Z michael_lee joined #sbcl 2014-04-05T20:30:38Z sdemarre joined #sbcl 2014-04-05T20:37:38Z harovali quit (Ping timeout: 240 seconds) 2014-04-05T20:38:43Z harovali joined #sbcl 2014-04-05T20:46:06Z sdemarre quit (Read error: Operation timed out) 2014-04-05T20:55:06Z drmeister joined #sbcl 2014-04-05T21:04:25Z michael_lee quit (Quit: Ex-Chat) 2014-04-05T21:35:54Z jaimef quit (Excess Flood) 2014-04-05T21:48:57Z jaimef joined #sbcl 2014-04-05T22:03:35Z leuler quit (Quit: ERC Version 5.1.2 $Revision: 1.796.2.6 $ (IRC client for Emacs)) 2014-04-05T23:05:34Z ehaliewicz joined #sbcl 2014-04-06T00:26:08Z yacks quit (Ping timeout: 240 seconds) 2014-04-06T00:27:27Z yacks joined #sbcl 2014-04-06T00:56:47Z slyrus_ joined #sbcl 2014-04-06T00:56:51Z slyrus quit (Ping timeout: 252 seconds) 2014-04-06T00:57:00Z slyrus_ is now known as slyrus 2014-04-06T01:36:06Z prxq_ joined #sbcl 2014-04-06T01:38:58Z prxq quit (Ping timeout: 240 seconds) 2014-04-06T01:42:23Z reb` quit (Remote host closed the connection) 2014-04-06T02:05:30Z brown joined #sbcl 2014-04-06T02:05:56Z brown is now known as Guest86564 2014-04-06T02:06:21Z Guest86564 is now known as reb 2014-04-06T02:38:05Z christoph_debian quit (Read error: Operation timed out) 2014-04-06T02:42:03Z yacks quit (Read error: Operation timed out) 2014-04-06T02:48:54Z slyrus quit (Ping timeout: 268 seconds) 2014-04-06T02:49:28Z eudoxia joined #sbcl 2014-04-06T02:52:10Z christoph_debian joined #sbcl 2014-04-06T03:07:38Z oleo quit (Ping timeout: 240 seconds) 2014-04-06T03:08:25Z oleo joined #sbcl 2014-04-06T03:11:21Z drmeiste_ joined #sbcl 2014-04-06T03:13:51Z drmeister quit (Ping timeout: 255 seconds) 2014-04-06T03:19:58Z eudoxia quit (Quit: leaving) 2014-04-06T04:47:19Z LiamH quit (Quit: Leaving.) 2014-04-06T04:50:08Z ehaliewicz quit (Ping timeout: 240 seconds) 2014-04-06T05:21:38Z scymtym quit (Ping timeout: 240 seconds) 2014-04-06T05:24:45Z drmeiste_ quit (Remote host closed the connection) 2014-04-06T07:06:12Z sdemarre joined #sbcl 2014-04-06T07:17:09Z harovali quit (Ping timeout: 268 seconds) 2014-04-06T07:17:42Z attila_lendvai joined #sbcl 2014-04-06T07:29:08Z sdemarre quit (Ping timeout: 240 seconds) 2014-04-06T07:42:04Z sdemarre joined #sbcl 2014-04-06T08:47:08Z sdemarre quit (Ping timeout: 240 seconds) 2014-04-06T09:32:49Z kludge` quit (Ping timeout: 268 seconds) 2014-04-06T09:37:12Z kludge` joined #sbcl 2014-04-06T10:57:53Z fiveop joined #sbcl 2014-04-06T11:31:18Z nyef joined #sbcl 2014-04-06T11:31:33Z nyef: G'morning all. 2014-04-06T11:32:26Z sdemarre joined #sbcl 2014-04-06T12:02:12Z attila_lendvai quit (Quit: Leaving.) 2014-04-06T12:08:27Z ASau` joined #sbcl 2014-04-06T12:12:32Z ASau quit (Ping timeout: 268 seconds) 2014-04-06T12:14:10Z ASau` is now known as ASau 2014-04-06T12:35:38Z harovali joined #sbcl 2014-04-06T12:51:38Z cmack quit (Ping timeout: 240 seconds) 2014-04-06T13:00:18Z scymtym joined #sbcl 2014-04-06T13:07:21Z LiamH joined #sbcl 2014-04-06T13:42:08Z sdemarre quit (Ping timeout: 240 seconds) 2014-04-06T14:54:42Z yacks joined #sbcl 2014-04-06T14:54:56Z drmeister joined #sbcl 2014-04-06T14:55:01Z yacks quit (Read error: Connection reset by peer) 2014-04-06T15:12:20Z fridim_ joined #sbcl 2014-04-06T15:17:05Z nyef: Hunh. Threads aren't enabled by default on PPC? 2014-04-06T15:23:29Z drmeister quit (Remote host closed the connection) 2014-04-06T15:25:25Z attila_lendvai joined #sbcl 2014-04-06T15:25:25Z attila_lendvai quit (Changing host) 2014-04-06T15:25:25Z attila_lendvai joined #sbcl 2014-04-06T15:36:44Z oleo quit (Read error: Operation timed out) 2014-04-06T15:37:16Z oleo joined #sbcl 2014-04-06T15:41:36Z LiamH quit (Quit: Leaving.) 2014-04-06T15:48:59Z fridim_ quit (Ping timeout: 268 seconds) 2014-04-06T16:00:40Z fridim_ joined #sbcl 2014-04-06T16:19:16Z pkhuong: nyef: I'm not even sure we enable them by default on darwin/x86 2014-04-06T16:19:53Z nyef: I'm on Linux, so that's not quite at issue. But it's probably a lot easier to break PPC threading by accident than it is to break x86oid threading. 2014-04-06T16:24:04Z nyef: Okay, threaded PPC passes the test suite reasonably, time to do a cheneygc build... 2014-04-06T16:24:13Z jaimef quit (Excess Flood) 2014-04-06T16:29:08Z jaimef joined #sbcl 2014-04-06T16:30:54Z pkhuong: Oh right. I have a bunch of barrier and CAS stuff to look at. 2014-04-06T16:32:56Z nyef: ... I need a better SPARC. This one is maxed out doing something not-SBCL-related, so I can't even run a build right now... And a build takes north of an hour, and the other task is probably going to take another three and a half. /-: 2014-04-06T16:33:17Z pkhuong: The GHC people used to have a couple niagara. 2014-04-06T16:34:11Z nyef: I'm probably just going to order something that's at least two CPUs and at least 800MHz off of eBay one of these days. 2014-04-06T16:35:18Z pkhuong: anyone willing to write the intro to my thesis? I'm told that half a page is too short ;) 2014-04-06T16:43:32Z jaimef quit (Excess Flood) 2014-04-06T16:46:38Z jaimef joined #sbcl 2014-04-06T16:52:24Z Krystof: pkhuong: sure. Can I relate your work to early-universe cosmology? 2014-04-06T16:53:04Z Krystof: there are a lot of integer parameters in the Universe 2014-04-06T16:54:00Z nyef: Hunh. I might have a scheme to allow threaded cheneygc systems. 2014-04-06T16:55:04Z jaimef quit (Excess Flood) 2014-04-06T16:56:38Z nyef: Might be amusing, but I'm not at all convinced that it's worth doing. 2014-04-06T17:01:08Z jaimef joined #sbcl 2014-04-06T17:03:57Z pkhuong: Krystof: ah, so you actually understand why we should be excited by the BICEP2 findings. 2014-04-06T17:04:05Z fridim_ quit (Read error: Operation timed out) 2014-04-06T17:04:39Z Krystof: well, we sort-of shouldn't 2014-04-06T17:04:46Z Krystof: science-by-press-release tends not to work too well 2014-04-06T17:05:07Z Krystof: also, you should remember that my perception of cosmology is basically that every theory has a free function 2014-04-06T17:05:27Z Krystof: so you should take claims that the data fits the theory as pretty vacuous :) 2014-04-06T17:05:42Z Krystof: but, I confess to not having actually read the paper 2014-04-06T17:09:40Z nyef: Oh, good. sb-rotate-byte failing to build on PPC/cheneygc is repeatable. 2014-04-06T17:12:41Z nyef: Now, if the failure would just make sense, or at least provide more information than a corruption warning immediately after building module-setup.fasl... 2014-04-06T17:13:05Z pkhuong: huh. can't even come up with an hypothesis for that one. 2014-04-06T17:14:29Z nyef: Yeah, exactly. 2014-04-06T17:24:11Z nyef: It's a memory fault with a page-aligned address within dynamic-space-0. 2014-04-06T17:26:03Z nyef: Program counter is not in range for any space, and it is lower than the addresses in sbcl.nm. 2014-04-06T17:35:07Z flip214: nyef: is valgrind an option? perhaps with --db-attach=yes? 2014-04-06T17:35:38Z nyef: I don't know that I even HAVE valgrind. 2014-04-06T17:36:06Z nyef: Oh, and if I build just sb-rotate-byte it works fine. 2014-04-06T17:40:42Z nyef: Here's another fun bit: If I build just sb-md5 and sb-rotate-byte, I only get a build failure on sb-rotate-byte if sb-md5 is earlier on the command line. 2014-04-06T17:41:43Z flip214: nyef: try to use valgrind. 2014-04-06T17:42:01Z flip214: to me that sounds like some uninitialized variable ... but what do I know? 2014-04-06T17:42:51Z flip214: using sb-md5 earlier might set _exactly_ that stack location to some "bad" value (like != 0 ;) 2014-04-06T17:43:04Z nyef: Separate process instances. 2014-04-06T17:43:13Z nyef: Each contrib is built in its own process. 2014-04-06T17:43:19Z nyef: (In theory, at least.) 2014-04-06T17:43:27Z flip214: so, what do they share, then? 2014-04-06T17:43:38Z nyef: The difference is that with sb-md5 being built first, the fasls for sb-rotate-byte have already been built. 2014-04-06T17:43:41Z flip214: what's the OS? with debian you'd have lots of debugging toys. 2014-04-06T17:43:52Z pkhuong: flip214: valgrind will be useless with sbcl unless we add a lot more of code. 2014-04-06T17:44:07Z pkhuong: valgrind on nyef's ppc also sounds like torture. 2014-04-06T17:44:28Z nyef: At least this one is a G5, not one of my G4s. 2014-04-06T17:47:03Z joshe: nyef: I have a niagra sitting idle, it may be possible to set up a linux or solaris vm for sbcl development 2014-04-06T17:47:57Z joshe: preferably linux, I hear installing solaris into one of those VMs is a real ordeal 2014-04-06T17:48:49Z flip214: nyef: what exactly is causing the problem, and what does work? do you have commandlines for me to test? 2014-04-06T17:49:04Z nyef: joshe: My local box should become usable in about two more hours, and I've got about a week during which I can reasonably bang on SBCL while doing other things. So, thank you, but I'm good for now. 2014-04-06T17:49:41Z joshe: ok 2014-04-06T17:50:00Z pkhuong: joshe: you just have a niagara sitting there? 2014-04-06T17:50:01Z joshe: I suspect it would take rather more than two hours to set something up for you ;) 2014-04-06T17:50:36Z nyef: flip214: If I knew what was causing the problem I'd be a lot closer to fixing it. I did a build --without-gencgc --with-cheneygc, at which point I spotted the failure, and for testing building individual contribs I use sh ./make-target-contrib.sh {contrib names}. 2014-04-06T17:50:53Z joshe: pkhuong: I ebayed it a while ago and haven't some up with anything interesting to do with it yet 2014-04-06T17:51:25Z joshe: *come 2014-04-06T17:52:29Z joshe: the first generation ones are becoming quite cheap on ebay nowadays 2014-04-06T17:52:53Z nyef: Yeah, less than $400 it looks like. 2014-04-06T17:54:01Z nyef: I'm seriously considering buying one myself at this rate. 2014-04-06T17:54:21Z joshe: of course the trick is then getting rails and finding some place to rack it where you don't have to hear it 2014-04-06T17:54:49Z nyef: Yeah, I was also looking at colo prices earlier this week. 2014-04-06T17:55:59Z joshe: it's a good fit for colo as it doesn't use a lot of power 2014-04-06T17:56:34Z nyef: I might-or-might-not be getting a bit of a windfall soon, and if I do and it's large enough then I'm thinking to spring for a decent alpha system. And I might grab an XServe by the end of the month... 2014-04-06T17:56:41Z joshe: and if you have another machine to get into the alom from, remote console is trivial 2014-04-06T17:57:45Z joshe: I scored some sort of ppc xserv for $100 at an equipment recycler which I haven't even tried to use yet 2014-04-06T17:58:10Z pkhuong: I recently looked into colo prices for my 4U (4x E5-4617), and it wasn't pretty. 2014-04-06T17:58:11Z joshe: ppc mac minis make much better ppc machines to tuck away somewhere 2014-04-06T17:59:12Z nyef: PPC mac mini is insufficient to requirements: Too slow, not enough CPUs. 2014-04-06T17:59:29Z joshe: I would like to have an alpha someday, maybe take a shot at finishing up the 64-bit alpha work and port to openbsd 2014-04-06T17:59:32Z nyef: Okay, some of them aren't too slow, but still not enough CPUs. 2014-04-06T17:59:56Z joshe: I have a pa-risc machine in the closet but it no longer powers on 2014-04-06T18:00:14Z nyef: Mmm. And PA-RISC costs about as much as Alpha does on eBay. 2014-04-06T18:01:51Z nyef: The real trick would be to find a decently strong rack-mount MIPS box. (minimum requirements: 2 CPUs, 800 MHz) 2014-04-06T18:03:04Z flip214: nyef: I'm using foreign arch debootstraps with qemu-*-static for foreign package building .... 2014-04-06T18:03:15Z flip214: don't know whether that would be fast enough for you, though. 2014-04-06T18:04:11Z nyef: I don't trust simulators. 2014-04-06T18:04:29Z nyef: The true test of software is real hardware. 2014-04-06T18:04:46Z joshe: indeed 2014-04-06T18:06:55Z nyef: Similar to how a good burn-in test for a new or newly-repaired box is a Linux kernel build. 2014-04-06T18:07:19Z nyef: Or a fresh stage-1 gentoo install. 2014-04-06T18:08:19Z joshe: I've used parallel sbcl builds in a loop for that before 2014-04-06T18:13:08Z nyef: Now, if we could just parallelize SBCL builds. (-: 2014-04-06T18:14:22Z joshe: how hard can it be? ;) 2014-04-06T18:15:03Z nyef: Exactly! 2014-04-06T18:16:07Z nyef: Doesn't even have to be in the same process, for that matter, so we wouldn't have to worry about any large locks or anything. 2014-04-06T18:17:51Z fiveop: Do we load what we compile during the build process (make-host-1 and -2)? 2014-04-06T18:18:05Z fiveop: (within each stage not in between) 2014-04-06T18:19:01Z nyef: host-1, yes. host-2, no. 2014-04-06T18:19:16Z nyef: But there are still compile-time effects. 2014-04-06T18:19:47Z fiveop: you have to load all macros even in host-2, don't you? 2014-04-06T18:20:47Z nyef: It's not just that, there are data structures built up in memory, there's at least one place that uses EVAL to define inlined functions (or maybe macros) at macroexpand time (unless that's been fixed recently?) and so on. 2014-04-06T18:27:04Z drmeister joined #sbcl 2014-04-06T18:58:14Z nyef: I'm still drawing a blank on the sb-rotate-byte thing for PPC/cheneygc, but I've thought of another test that I could run: Does it also fail on SPARC/cheneygc? 2014-04-06T19:28:44Z nyef: And, once again, LDB is completely bloody useless in the face of a memory fault. 2014-04-06T19:29:04Z nyef: (Now that I've thought to hack the contrib build script to not --disable-debugger.) 2014-04-06T19:33:55Z nyef: Okay, http://paste.lisp.org/display/141935 has a backtrace, for whatever it may be worth. 2014-04-06T19:34:59Z nyef: (Removed both lose-on-corruption and disable-debugger to get this.) 2014-04-06T19:48:11Z nyef: Lovely. Something in libc, but not something that shows up in the backtrace. 2014-04-06T19:49:28Z nyef: (gdb backtrace and proc/pid/maps attached.) 2014-04-06T20:02:41Z nyef: ... "no function contains specified address." 2014-04-06T20:03:00Z nyef: Ah, and my SPARC is no longer busy. 2014-04-06T20:20:41Z fiveop quit 2014-04-06T20:40:08Z DGASAU` joined #sbcl 2014-04-06T20:42:09Z pjb` joined #sbcl 2014-04-06T20:43:45Z antoszka_ joined #sbcl 2014-04-06T20:44:14Z antoszka quit (Disconnected by services) 2014-04-06T20:44:21Z antoszka_ is now known as antoszka 2014-04-06T20:47:08Z redline6561_ joined #sbcl 2014-04-06T20:47:28Z redline6561_ is now known as redline6561` 2014-04-06T20:48:10Z jaimef_ joined #sbcl 2014-04-06T20:48:19Z psilord quit (*.net *.split) 2014-04-06T20:48:19Z pjb quit (*.net *.split) 2014-04-06T20:48:19Z redline6561 quit (*.net *.split) 2014-04-06T20:48:20Z jaimef quit (*.net *.split) 2014-04-06T20:48:20Z oleo quit (*.net *.split) 2014-04-06T20:48:20Z attila_lendvai quit (*.net *.split) 2014-04-06T20:48:21Z harovali quit (*.net *.split) 2014-04-06T20:48:21Z nyef quit (*.net *.split) 2014-04-06T20:48:22Z fikusz quit (*.net *.split) 2014-04-06T20:48:22Z DGASAU quit (*.net *.split) 2014-04-06T20:48:27Z jaimef_ is now known as jaimef 2014-04-06T20:54:51Z fikusz joined #sbcl 2014-04-06T20:55:20Z oleo joined #sbcl 2014-04-06T20:55:47Z nyef joined #sbcl 2014-04-06T20:56:06Z psilord joined #sbcl 2014-04-06T20:56:34Z harovali joined #sbcl 2014-04-06T21:29:53Z ccl-logbot joined #sbcl 2014-04-06T21:29:53Z 2014-04-06T21:29:53Z names: ccl-logbot harovali psilord nyef oleo fikusz jaimef redline6561` antoszka pjb` DGASAU` drmeister scymtym ASau kludge` christoph_debian reb prxq_ edgar-rft Bike irsol milosn @Krystof loke angavrilov |3b| tmh_ afleck ams loke_ kanru White_Flame Munksgaard asedeno luis nicdev xymox bege hzp joshe ferada pkhuong leoc fe[nl]ix flip214 ivan`` faheem yauz Posterdati foom minion specbot mood brucem jsnell_ scymtym_ Blkt daimrod pchrist 2014-04-06T21:34:19Z prxq_ is now known as prxq 2014-04-06T21:51:33Z ASau is now known as people 2014-04-06T21:51:42Z people is now known as ASau 2014-04-06T22:26:17Z nyef: Oh, hell. fd-stream-read-n-bytes is trying to read into a pinned vector (hence without-gcing) with a boxed address of #x4fe4f357 and a length of 8192... And it's faulting while accessing address 0x4fe50000. 2014-04-06T22:28:50Z nyef: At least I have some idea of where to look now. /-: 2014-04-06T22:38:41Z nyef: Might even be able to generate a useful test case for this, too. 2014-04-06T23:42:26Z yurisy joined #sbcl 2014-04-06T23:48:50Z nyef: Oh, hell. cheneygc_handle_wp_violation(), within the gc trigger range, from a foreign function call. 2014-04-06T23:49:44Z nyef: Yeah, that's the damage. And there's even a fun comment in src/assembly/ppc/array.lisp, allocate-vector-on-heap. 2014-04-06T23:58:08Z pkhuong: great. mprotect strikes even on cheneygc 2014-04-06T23:58:32Z pkhuong: Is there a nice reference for the meaning of (let ((*x* 42)) (makunbound '*x*) (setf *x* 5)) ? 2014-04-07T00:28:15Z nyef: Presuming X is special? 2014-04-07T00:28:47Z nyef: Hrm. 2014-04-07T00:36:37Z harovali quit (Read error: Connection reset by peer) 2014-04-07T00:39:17Z drmeister quit (Ping timeout: 246 seconds) 2014-04-07T00:40:01Z harovali joined #sbcl 2014-04-07T00:40:13Z nyef: pkhuong: As I just mentioned in #lisp, you're looking for CLHS 3.1.2.1.1.2. 2014-04-07T00:55:18Z nyef: Okay, for a test case, I'm thinking to set up an impure test or a .sh test that sets up a WITHOUT-GCING, then calls bzero() on something two pages beyond the GC trigger. 2014-04-07T00:56:57Z nyef: Alternately, GC, allocate an array large enough to blow past the GC trigger, pin it, then bzero() it. 2014-04-07T00:57:30Z pkhuong: nyef: so we would trigger a gc only when the page is accessed?! 2014-04-07T00:58:39Z nyef: On cheneygc, we trigger a GC whenever there's a write to a page within the current dynamic space beyond the GC trigger point from lisp code. 2014-04-07T00:59:07Z nyef: So I could just bzero the end of the current dynamic space and that'd do the trick, but I'd rather cover an actual, plausible usage scenario. 2014-04-07T00:59:48Z pkhuong: allocate a huge vector? 2014-04-07T01:00:03Z nyef: That'd be the second of my two proposed test cases, yes. 2014-04-07T01:00:42Z pkhuong: I get why read(2) would fail, but not why it'd crash our runtime though. 2014-04-07T01:01:07Z nyef: Have a look at the last function in cheneygc.c: It only kicks in if it's invoked from lisp, not from alien code. 2014-04-07T01:01:19Z pkhuong: :\ 2014-04-07T01:01:28Z nyef: /That's/ the bug right there. 2014-04-07T01:01:42Z pkhuong: arguably, the bug lies in the allocation path ;) 2014-04-07T01:02:02Z nyef: Hah! Have a look at the allocation path. src/assembly/ppc/array.lisp. 2014-04-07T01:03:10Z pkhuong: I'm pretty sure the problem is touching the last page but not any of the intervening ones 2014-04-07T01:03:39Z pkhuong: relatively clearn though, and I like the fact that it's an out of line assembly routine 2014-04-07T01:04:04Z nyef: On gencgc.c, maybe. On cheneygc, IIUC, the heap is read-write up to the gc trigger page. 2014-04-07T01:04:21Z pkhuong: and past it? 2014-04-07T01:04:24Z nyef: And once the trigger has been hit, the entire dynamic space is read-write. 2014-04-07T01:04:34Z nyef: Past it is read-only until it gets hit. 2014-04-07T01:04:42Z nyef: That's my current understanding, at least. 2014-04-07T01:05:33Z nyef: Ooh. Look at the same routine for SPARC. 2014-04-07T01:05:38Z pkhuong: I don't see how cheney avoids allocating a lot more than heap size 2014-04-07T01:07:08Z nyef: Heh. And ALPHA. 2014-04-07T01:07:20Z nyef: Okay, we definitely need a test case for this, one way or another. 2014-04-07T01:07:37Z pkhuong: we need a test case for allocating zero-elements array too 2014-04-07T01:08:03Z pkhuong: the -1 workaround doesn't just fail on nil arrays, it fails on anything with 0 data word. 2014-04-07T01:09:10Z pkhuong: So yeah. if I have a 256 MB heap, what's keeping a cheneygc build from allocation 2GB and overwriting its stack or something silly? 2014-04-07T01:09:15Z pkhuong: *allocating 2014-04-07T01:09:33Z nyef: Okay, on gencgc we don't care: either it's a small object in which case the region is read-write anyway or it's a large object in which case it will be allocated in an unboxed (read-write) region if it's going to be passed to alien code. 2014-04-07T01:10:06Z pkhuong: we might care just after coreload 2014-04-07T01:10:30Z pkhuong: but, iirc, the result was EINTR or something, not a runtime crash. 2014-04-07T01:10:48Z nyef: Yeah, gencgc is a lot more graceful with sig_memory_fault. 2014-04-07T01:11:00Z nyef: Well, within heap space, at least. 2014-04-07T01:11:20Z nyef: I have a fix in mind for cheneygc, the thing that I'm concerned about is a test case. 2014-04-07T01:12:06Z pkhuong: can you try and (make-array 0) on a cheney sparc, btw? I'm curious. 2014-04-07T01:12:28Z pkhuong: oh wait no, it doesn't write to alloc - 1 anymore 2014-04-07T01:13:11Z nyef: Right, the fact that it doesn't do the write on gencgc means that it can just stomp the word after the allocation, which will Do The Right Thing on cheneygc. 2014-04-07T01:13:27Z pkhuong: well, more or less. 2014-04-07T01:13:39Z nyef: Sure, it might trigger the GC a hair earlier than necessary. 2014-04-07T01:13:54Z nyef: To which I say "Binary File Descriptor". 2014-04-07T01:14:17Z pkhuong: It'll probably trigger the guard page (although, again, what's preventing me from allocating into a much farther area that's actually mapped in) 2014-04-07T01:14:49Z pkhuong: and I guess that's enough to unprotect earlier pages, I see. 2014-04-07T01:15:13Z nyef: To my mind, the "right thing" is to allow ANY write to the post-trigger part of the current dynamic space to trigger a GC, lisp code or alien. 2014-04-07T01:15:15Z pkhuong: Still worried about overflowing our magic page, like we used to do with stack alloc on x86oids. 2014-04-07T01:16:13Z nyef: set_auto_gc_trigger() protects the ENTIRE dynamic space beyond the trigger point, so blowing past it would have to be by quite a bit. 2014-04-07T01:16:55Z pkhuong: that, and inserting a check for allocating past the end of dynamic space for allocations larger than our safety buffer. 2014-04-07T01:18:41Z nyef: Yeah, the blowing past the end of dynamic space thing doesn't have an obvious check that I can see. 2014-04-07T01:19:06Z pkhuong: the good thing is it's probably only needed for vectors. 2014-04-07T01:19:34Z pkhuong: just like my hack to trigger GC before running out of heap on gencgc. 2014-04-07T01:20:22Z nyef: So, speaking of... there was a question recently in #lisp about returning to lisp after entering LDB. 2014-04-07T01:20:42Z pkhuong: don't think that's gonna happen. 2014-04-07T01:21:33Z nyef: And it occurred to me that in order for it to happen the runtime would have to be able to define its own unwind-protect mechanism, plus circumstances under which it makes sense to attempt to return to lisp. 2014-04-07T01:21:56Z pkhuong: but if we could make that happen, why drop to ldb at all? 2014-04-07T01:23:01Z nyef: Yeah, that was the other side of it. With the exception of some signal-handling games, almost all of the invocations of lose() in the runtime are things that it just doesn't make sense to recover from at all. 2014-04-07T01:23:50Z pkhuong: have you seen the stack overflow non-recovery bug on windows, btw? 2014-04-07T01:24:22Z nyef: Haven't bothered, TBH. 2014-04-07T01:25:12Z nyef: The only reason I got involved with SBCL on windows was that I was bored, I had a windows box at the time, and I was convinced that the explanation for why a windows port would be stupidly hard was almost completely wrong. 2014-04-07T01:25:55Z pkhuong: all right (: SBCL/win32: built on throw-away hacks. 2014-04-07T01:26:13Z nyef: Yup! 2014-04-07T01:26:28Z LiamH joined #sbcl 2014-04-07T01:27:19Z nyef: On the other hand, I'm actually kindof proud of how closely I managed to tie into the SEH framework, even if it did get screwed up by the PACK semantics and the ordering requirement on the stack. 2014-04-07T01:30:58Z nyef: Anyway, PPC: affected by this issue. SPARC: not bloody likely. MIPS and HPPA: I have no idea. 2014-04-07T01:32:55Z nyef: Actually, I should have more that one test case: One for a large allocation without tripping the GC trigger and one for simply accessing the current dynamic space past the end of the GC trigger and allocation pointer. 2014-04-07T01:36:38Z prxq_ joined #sbcl 2014-04-07T01:37:06Z prxq quit (Read error: Operation timed out) 2014-04-07T02:06:56Z nyef quit (Quit: G'night all.) 2014-04-07T02:26:03Z ltbarcly joined #sbcl 2014-04-07T02:29:13Z slyrus joined #sbcl 2014-04-07T02:31:36Z pjb` is now known as pjb 2014-04-07T02:38:38Z christoph_debian quit (Ping timeout: 240 seconds) 2014-04-07T02:52:08Z christoph_debian joined #sbcl 2014-04-07T03:09:02Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-07T03:30:45Z LiamH quit (Quit: Leaving.) 2014-04-07T04:03:08Z hzp quit (Ping timeout: 240 seconds) 2014-04-07T04:07:08Z scymtym quit (Ping timeout: 240 seconds) 2014-04-07T04:19:45Z michael_lee joined #sbcl 2014-04-07T04:23:38Z oleo quit (Ping timeout: 240 seconds) 2014-04-07T04:24:27Z oleo joined #sbcl 2014-04-07T04:54:41Z slyrus quit (Ping timeout: 252 seconds) 2014-04-07T05:05:51Z oleo quit (Quit: Leaving) 2014-04-07T05:05:59Z slyrus joined #sbcl 2014-04-07T05:07:17Z attila_lendvai joined #sbcl 2014-04-07T05:07:17Z attila_lendvai quit (Changing host) 2014-04-07T05:07:17Z attila_lendvai joined #sbcl 2014-04-07T05:18:16Z kanru quit (Ping timeout: 265 seconds) 2014-04-07T05:19:24Z yacks joined #sbcl 2014-04-07T05:37:56Z pranavrc joined #sbcl 2014-04-07T05:46:38Z ASau quit (Ping timeout: 240 seconds) 2014-04-07T06:28:22Z DGASAU` is now known as DGASAU 2014-04-07T06:30:27Z zenyfish joined #sbcl 2014-04-07T06:47:01Z kanru joined #sbcl 2014-04-07T06:59:38Z edgar-rft quit (Quit: memory access corrupted because lifeform experiment discontinued) 2014-04-07T07:35:14Z yacks quit (Read error: Operation timed out) 2014-04-07T07:47:18Z Bike quit (Ping timeout: 240 seconds) 2014-04-07T07:50:24Z Bike joined #sbcl 2014-04-07T07:53:30Z yacks joined #sbcl 2014-04-07T08:17:28Z jdz joined #sbcl 2014-04-07T08:51:57Z pranavrc quit (Remote host closed the connection) 2014-04-07T09:05:42Z pranavrc joined #sbcl 2014-04-07T09:05:42Z pranavrc quit (Changing host) 2014-04-07T09:05:42Z pranavrc joined #sbcl 2014-04-07T09:31:08Z kludge` quit (Ping timeout: 240 seconds) 2014-04-07T09:36:38Z kludge` joined #sbcl 2014-04-07T10:03:53Z yurisy quit (Remote host closed the connection) 2014-04-07T11:02:38Z momo-reina joined #sbcl 2014-04-07T11:05:47Z momo-reina quit (Remote host closed the connection) 2014-04-07T12:02:38Z nyef joined #sbcl 2014-04-07T12:02:45Z nyef: G'morning all. 2014-04-07T12:07:06Z nyef: ... Every so often I run into evidence that our debugger was either planned to have a number of features that were never implemented or was at some point gutted. 2014-04-07T12:08:50Z flip214: "it may be both" 2014-04-07T12:09:24Z nyef: That too. 2014-04-07T12:09:36Z nyef: Half-implemented and then torn out is a distinct possibility. 2014-04-07T12:19:24Z nyef: ... And I see some of it being torn out in commit 6d9ecc45cb21a1208deb8c4d128adc04aa289c9d 2014-04-07T12:20:33Z nyef: Although by that point it was already fairly obviously dead. 2014-04-07T12:21:37Z eudoxia joined #sbcl 2014-04-07T12:34:42Z nyef: Oh. I can almost-sortof see the semantics involved in this mess that might make it workable even now. 2014-04-07T12:38:37Z harovali quit (Read error: Connection reset by peer) 2014-04-07T12:40:05Z harovali joined #sbcl 2014-04-07T12:56:45Z pranavrc quit 2014-04-07T13:40:51Z oleo joined #sbcl 2014-04-07T13:47:30Z prxq_ quit (Remote host closed the connection) 2014-04-07T13:58:44Z prxq joined #sbcl 2014-04-07T14:02:12Z segv- joined #sbcl 2014-04-07T14:40:18Z irsol quit (Ping timeout: 240 seconds) 2014-04-07T14:41:39Z edgar-rft joined #sbcl 2014-04-07T14:46:36Z hlavaty joined #sbcl 2014-04-07T14:57:23Z ltbarcly joined #sbcl 2014-04-07T14:57:54Z ltbarcly quit (Client Quit) 2014-04-07T14:59:12Z irsol joined #sbcl 2014-04-07T15:42:27Z ltbarcly joined #sbcl 2014-04-07T15:43:50Z ltbarcly quit (Client Quit) 2014-04-07T15:56:05Z nyef sighs. 2014-04-07T15:56:25Z nyef: Time to break out my poor, anemic MIPS system and try to get it running again. 2014-04-07T16:01:20Z sdemarre joined #sbcl 2014-04-07T16:13:39Z segv- quit (Remote host closed the connection) 2014-04-07T16:14:40Z jdz quit (Quit: Leaving) 2014-04-07T16:14:51Z segv- joined #sbcl 2014-04-07T16:28:51Z sdemarre quit (Read error: Operation timed out) 2014-04-07T16:32:25Z ltbarcly joined #sbcl 2014-04-07T16:34:51Z oleo quit (Ping timeout: 255 seconds) 2014-04-07T16:40:08Z ltbarcly quit (Ping timeout: 252 seconds) 2014-04-07T16:43:29Z ltbarcly joined #sbcl 2014-04-07T16:48:56Z nyef: ... lovely. My PPC just shut down partway through an SBCL build. /-: 2014-04-07T16:50:15Z oleo joined #sbcl 2014-04-07T16:56:08Z nicdev quit (Ping timeout: 240 seconds) 2014-04-07T16:57:02Z attila_lendvai quit (Disconnected by services) 2014-04-07T16:57:02Z attila_lendvai1 joined #sbcl 2014-04-07T16:57:02Z attila_lendvai1 quit (Changing host) 2014-04-07T16:57:02Z attila_lendvai1 joined #sbcl 2014-04-07T16:58:40Z nicdev joined #sbcl 2014-04-07T17:05:19Z nyef: ... And again. Clearly today is not my day for PPC stuff. 2014-04-07T17:10:32Z ASau joined #sbcl 2014-04-07T17:44:36Z ltbarcly quit (Quit: Textual IRC Client: www.textualapp.com) 2014-04-07T17:45:01Z ltbarcly joined #sbcl 2014-04-07T17:53:15Z kanru quit (Ping timeout: 255 seconds) 2014-04-07T18:09:38Z eudoxia quit (Quit: leaving) 2014-04-07T18:09:48Z eudoxia joined #sbcl 2014-04-07T18:11:22Z ltbarcly quit (Read error: Operation timed out) 2014-04-07T18:11:44Z sdemarre joined #sbcl 2014-04-07T18:12:02Z ltbarcly joined #sbcl 2014-04-07T18:19:00Z eudoxia quit (Quit: Lost terminal) 2014-04-07T18:36:27Z ltbarcly quit (Ping timeout: 255 seconds) 2014-04-07T18:38:02Z ltbarcly joined #sbcl 2014-04-07T18:41:33Z hzp joined #sbcl 2014-04-07T19:19:19Z zenyfish left #sbcl 2014-04-07T19:24:38Z ltbarcly quit (Ping timeout: 240 seconds) 2014-04-07T19:26:18Z ltbarcly joined #sbcl 2014-04-07T19:30:48Z sdemarre quit (Ping timeout: 255 seconds) 2014-04-07T19:52:58Z nyef: Heh! In compiler/sparc/subprim, can the NOT-LIST error trap EVER be triggered? 2014-04-07T19:57:59Z nyef: I guess under low-SAFETY it might. But at that point we probably don't care. 2014-04-07T20:03:26Z scymtym__ joined #sbcl 2014-04-07T20:04:01Z White__Flame joined #sbcl 2014-04-07T20:04:50Z jeti` joined #sbcl 2014-04-07T20:09:16Z ivan``_ joined #sbcl 2014-04-07T20:10:15Z White_Flame quit (*.net *.split) 2014-04-07T20:10:17Z ivan`` quit (*.net *.split) 2014-04-07T20:10:19Z scymtym_ quit (*.net *.split) 2014-04-07T20:15:38Z jeti` quit (Ping timeout: 252 seconds) 2014-04-07T20:19:40Z prxq quit (Quit: Leaving) 2014-04-07T20:24:29Z michael_lee quit (Quit: Ex-Chat) 2014-04-07T20:55:42Z ASau quit (Remote host closed the connection) 2014-04-07T20:59:01Z ASau joined #sbcl 2014-04-07T21:01:48Z jeti` joined #sbcl 2014-04-07T22:05:30Z bege left #sbcl 2014-04-07T22:22:15Z milosn quit (Ping timeout: 255 seconds) 2014-04-07T22:23:00Z milosn joined #sbcl 2014-04-07T22:25:35Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-07T22:41:50Z drmeister joined #sbcl 2014-04-07T22:42:03Z ivan``_ quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2014-04-07T22:43:42Z segv- quit (Remote host closed the connection) 2014-04-07T22:47:24Z scymtym joined #sbcl 2014-04-07T22:48:13Z ivan`` joined #sbcl 2014-04-07T22:48:13Z ivan`` quit (Changing host) 2014-04-07T22:48:13Z ivan`` joined #sbcl 2014-04-07T22:57:54Z attila_lendvai1 quit (Ping timeout: 255 seconds) 2014-04-07T23:05:14Z drmeister quit (Read error: Connection reset by peer) 2014-04-07T23:13:19Z drmeister joined #sbcl 2014-04-07T23:22:12Z drmeister quit (Remote host closed the connection) 2014-04-08T00:39:41Z jeti` quit (Read error: Connection reset by peer) 2014-04-08T00:42:37Z harovali1 joined #sbcl 2014-04-08T00:44:26Z harovali quit (Ping timeout: 246 seconds) 2014-04-08T01:09:10Z eudoxia joined #sbcl 2014-04-08T01:30:37Z drmeister joined #sbcl 2014-04-08T01:39:53Z pjb` joined #sbcl 2014-04-08T01:42:08Z pjb quit (Ping timeout: 240 seconds) 2014-04-08T01:42:22Z eudoxia quit (Quit: leaving) 2014-04-08T01:51:35Z zRecursive joined #sbcl 2014-04-08T02:15:25Z ccl-logbot joined #sbcl 2014-04-08T02:15:26Z 2014-04-08T02:15:26Z names: ccl-logbot zRecursive pjb` drmeister harovali1 ivan`` scymtym milosn ASau White__Flame scymtym__ hzp nicdev oleo irsol hlavaty edgar-rft nyef kludge` yacks Bike slyrus christoph_debian psilord fikusz jaimef redline6561` antoszka DGASAU reb @Krystof loke angavrilov |3b| tmh_ afleck ams loke_ Munksgaard asedeno luis xymox joshe ferada pkhuong leoc fe[nl]ix flip214 faheem yauz Posterdati foom minion specbot pchrist daimrod Blkt jsnell_ brucem mood 2014-04-08T02:15:26Z kanru joined #sbcl 2014-04-08T02:23:29Z kanru quit (Ping timeout: 252 seconds) 2014-04-08T02:25:03Z nyef quit (Quit: G'night all.) 2014-04-08T02:38:08Z christoph_debian quit (Ping timeout: 240 seconds) 2014-04-08T02:51:46Z christoph_debian joined #sbcl 2014-04-08T03:01:18Z slyrus quit (Ping timeout: 252 seconds) 2014-04-08T03:18:48Z slyrus joined #sbcl 2014-04-08T03:38:08Z scymtym quit (Ping timeout: 240 seconds) 2014-04-08T04:01:35Z pranavrc joined #sbcl 2014-04-08T04:13:02Z pjb` is now known as pjb 2014-04-08T04:22:32Z drmeister quit (Remote host closed the connection) 2014-04-08T04:45:06Z kanru joined #sbcl 2014-04-08T05:05:13Z oleo quit (Ping timeout: 240 seconds) 2014-04-08T05:26:48Z Bike quit (Quit: leaving) 2014-04-08T05:30:04Z attila_lendvai joined #sbcl 2014-04-08T05:30:04Z attila_lendvai quit (Changing host) 2014-04-08T05:30:04Z attila_lendvai joined #sbcl 2014-04-08T05:36:46Z yacks quit (Read error: Operation timed out) 2014-04-08T05:37:42Z yacks joined #sbcl 2014-04-08T05:46:13Z sdemarre joined #sbcl 2014-04-08T06:00:33Z sdemarre quit (Ping timeout: 252 seconds) 2014-04-08T06:08:03Z zRecursive left #sbcl 2014-04-08T06:09:00Z DGASAU quit (Ping timeout: 255 seconds) 2014-04-08T06:11:23Z drmeister joined #sbcl 2014-04-08T06:15:53Z drmeister quit (Ping timeout: 240 seconds) 2014-04-08T06:39:21Z momo-reina joined #sbcl 2014-04-08T06:43:06Z ASau quit (Ping timeout: 255 seconds) 2014-04-08T07:02:52Z momo-reina quit (Remote host closed the connection) 2014-04-08T07:59:00Z DGASAU joined #sbcl 2014-04-08T08:00:15Z drmeister joined #sbcl 2014-04-08T08:00:46Z loke_ quit (Remote host closed the connection) 2014-04-08T08:04:50Z drmeister quit (Ping timeout: 252 seconds) 2014-04-08T08:14:41Z michael_lee joined #sbcl 2014-04-08T08:15:58Z michael_lee quit (Max SendQ exceeded) 2014-04-08T08:24:35Z jdz joined #sbcl 2014-04-08T08:50:44Z loke_ joined #sbcl 2014-04-08T08:57:01Z ltbarcly joined #sbcl 2014-04-08T09:31:38Z kludge` quit (Ping timeout: 240 seconds) 2014-04-08T09:36:47Z kludge` joined #sbcl 2014-04-08T09:39:01Z DGASAU quit (*.net *.split) 2014-04-08T09:39:02Z pranavrc quit (*.net *.split) 2014-04-08T09:39:06Z hlavaty quit (*.net *.split) 2014-04-08T09:39:08Z antoszka quit (*.net *.split) 2014-04-08T09:39:10Z xymox quit (*.net *.split) 2014-04-08T09:39:12Z Blkt quit (*.net *.split) 2014-04-08T09:39:18Z Blkt joined #sbcl 2014-04-08T09:39:19Z hlavaty joined #sbcl 2014-04-08T09:39:36Z pranavrc joined #sbcl 2014-04-08T09:39:36Z pranavrc quit (Changing host) 2014-04-08T09:39:36Z pranavrc joined #sbcl 2014-04-08T09:39:46Z christoph_debian quit (*.net *.split) 2014-04-08T09:39:47Z harovali1 quit (*.net *.split) 2014-04-08T09:39:54Z antoszka joined #sbcl 2014-04-08T09:40:01Z loke quit (*.net *.split) 2014-04-08T09:40:01Z afleck quit (*.net *.split) 2014-04-08T09:40:01Z joshe quit (*.net *.split) 2014-04-08T09:40:04Z foom quit (*.net *.split) 2014-04-08T09:40:07Z afleck joined #sbcl 2014-04-08T09:40:15Z loke joined #sbcl 2014-04-08T09:40:17Z joshe joined #sbcl 2014-04-08T09:40:21Z foom joined #sbcl 2014-04-08T09:40:24Z christoph_debian joined #sbcl 2014-04-08T09:40:38Z DGASAU joined #sbcl 2014-04-08T09:41:21Z scymtym__ quit (*.net *.split) 2014-04-08T09:41:27Z flip214 quit (*.net *.split) 2014-04-08T09:41:29Z pchrist quit (*.net *.split) 2014-04-08T09:41:29Z daimrod quit (*.net *.split) 2014-04-08T09:42:07Z scymtym__ joined #sbcl 2014-04-08T09:42:08Z pchrist joined #sbcl 2014-04-08T09:42:12Z ltbarcly quit (*.net *.split) 2014-04-08T09:42:13Z jdz quit (*.net *.split) 2014-04-08T09:42:15Z pjb quit (*.net *.split) 2014-04-08T09:42:15Z ivan`` quit (*.net *.split) 2014-04-08T09:42:15Z milosn quit (*.net *.split) 2014-04-08T09:42:17Z irsol quit (*.net *.split) 2014-04-08T09:42:17Z edgar-rft quit (*.net *.split) 2014-04-08T09:42:23Z angavrilov quit (*.net *.split) 2014-04-08T09:42:27Z ferada quit (*.net *.split) 2014-04-08T09:42:27Z pkhuong quit (*.net *.split) 2014-04-08T09:42:39Z milosn joined #sbcl 2014-04-08T09:42:44Z angavrilov joined #sbcl 2014-04-08T09:43:01Z ltbarcly joined #sbcl 2014-04-08T09:44:20Z edgar-rft joined #sbcl 2014-04-08T09:44:26Z pjb joined #sbcl 2014-04-08T09:46:07Z harovali joined #sbcl 2014-04-08T09:46:40Z jdz joined #sbcl 2014-04-08T09:47:50Z ivan`` joined #sbcl 2014-04-08T09:47:50Z irsol joined #sbcl 2014-04-08T09:51:49Z ferada joined #sbcl 2014-04-08T09:52:30Z flip214 joined #sbcl 2014-04-08T09:52:30Z flip214 quit (Changing host) 2014-04-08T09:52:30Z flip214 joined #sbcl 2014-04-08T09:54:32Z pkhuong joined #sbcl 2014-04-08T09:55:15Z xymox joined #sbcl 2014-04-08T10:03:29Z scymtym__ quit (*.net *.split) 2014-04-08T10:03:31Z foom quit (*.net *.split) 2014-04-08T10:03:32Z loke quit (*.net *.split) 2014-04-08T10:03:37Z reb quit (*.net *.split) 2014-04-08T10:03:38Z |3b| quit (*.net *.split) 2014-04-08T10:03:51Z reb joined #sbcl 2014-04-08T10:03:52Z scymtym__ joined #sbcl 2014-04-08T10:03:54Z loke joined #sbcl 2014-04-08T10:03:57Z foom joined #sbcl 2014-04-08T10:04:19Z |3b|` joined #sbcl 2014-04-08T10:08:52Z |3b|` is now known as |3b| 2014-04-08T10:13:56Z |3b|` joined #sbcl 2014-04-08T10:13:58Z reb` joined #sbcl 2014-04-08T10:15:18Z reb quit (Ping timeout: 264 seconds) 2014-04-08T10:17:41Z |3b| quit (Ping timeout: 264 seconds) 2014-04-08T10:26:49Z pchrist quit (*.net *.split) 2014-04-08T10:26:54Z joshe quit (*.net *.split) 2014-04-08T10:26:55Z DGASAU quit (*.net *.split) 2014-04-08T10:27:02Z yacks quit (*.net *.split) 2014-04-08T10:27:10Z ams quit (*.net *.split) 2014-04-08T10:27:17Z faheem quit (*.net *.split) 2014-04-08T10:27:19Z Posterdati quit (*.net *.split) 2014-04-08T10:27:28Z yacks joined #sbcl 2014-04-08T10:27:36Z Posterdati joined #sbcl 2014-04-08T10:27:50Z faheem joined #sbcl 2014-04-08T10:28:00Z DGASAU joined #sbcl 2014-04-08T10:28:19Z ams joined #sbcl 2014-04-08T10:30:58Z joshe joined #sbcl 2014-04-08T10:39:48Z xymox quit (*.net *.split) 2014-04-08T10:39:52Z ltbarcly quit (*.net *.split) 2014-04-08T10:39:56Z pranavrc quit (*.net *.split) 2014-04-08T10:39:57Z kludge` quit (*.net *.split) 2014-04-08T10:40:03Z nicdev quit (*.net *.split) 2014-04-08T10:40:29Z nicdev` joined #sbcl 2014-04-08T10:40:29Z kludge` joined #sbcl 2014-04-08T10:40:33Z pranavrc joined #sbcl 2014-04-08T10:40:33Z pranavrc quit (Changing host) 2014-04-08T10:40:33Z pranavrc joined #sbcl 2014-04-08T10:40:34Z xymox joined #sbcl 2014-04-08T10:40:36Z ltbarcly joined #sbcl 2014-04-08T10:45:56Z pchrist joined #sbcl 2014-04-08T10:51:40Z yacks quit (Quit: Leaving) 2014-04-08T10:52:00Z yacks joined #sbcl 2014-04-08T11:37:44Z drmeister joined #sbcl 2014-04-08T11:38:14Z yacks quit (Ping timeout: 255 seconds) 2014-04-08T11:42:17Z drmeister quit (Ping timeout: 264 seconds) 2014-04-08T11:56:18Z yacks joined #sbcl 2014-04-08T11:59:24Z ccl-logbot joined #sbcl 2014-04-08T11:59:24Z 2014-04-08T11:59:24Z names: ccl-logbot yacks pchrist ltbarcly xymox pranavrc kludge` nicdev` joshe ams DGASAU faheem Posterdati reb` |3b|` foom loke scymtym__ pkhuong flip214 ferada irsol ivan`` jdz harovali pjb edgar-rft angavrilov milosn christoph_debian afleck antoszka hlavaty Blkt loke_ attila_lendvai kanru slyrus White__Flame hzp psilord fikusz jaimef redline6561` @Krystof tmh_ mood brucem jsnell_ specbot minion yauz fe[nl]ix leoc luis asedeno Munksgaard 2014-04-08T12:07:04Z nyef joined #sbcl 2014-04-08T12:07:14Z nyef: G'morning all. 2014-04-08T12:17:30Z reb`` joined #sbcl 2014-04-08T12:18:33Z reb` quit (Ping timeout: 252 seconds) 2014-04-08T12:18:33Z loke quit (Ping timeout: 252 seconds) 2014-04-08T12:18:55Z foom quit (Ping timeout: 252 seconds) 2014-04-08T12:20:22Z |3b|` quit (Ping timeout: 252 seconds) 2014-04-08T12:20:44Z scymtym__ quit (Ping timeout: 252 seconds) 2014-04-08T12:20:55Z scymtym joined #sbcl 2014-04-08T12:21:03Z loke joined #sbcl 2014-04-08T12:22:04Z foom joined #sbcl 2014-04-08T12:24:41Z DGASAU quit (Remote host closed the connection) 2014-04-08T12:27:57Z drmeister joined #sbcl 2014-04-08T12:29:10Z DGASAU joined #sbcl 2014-04-08T12:31:46Z eudoxia joined #sbcl 2014-04-08T12:32:36Z drmeister quit (Remote host closed the connection) 2014-04-08T12:45:14Z harovali quit (Ping timeout: 240 seconds) 2014-04-08T12:45:56Z harovali joined #sbcl 2014-04-08T12:50:16Z milanj joined #sbcl 2014-04-08T12:52:29Z pranavrc quit 2014-04-08T12:55:14Z drmeister joined #sbcl 2014-04-08T12:55:55Z psilord quit (Quit: Leaving.) 2014-04-08T13:38:00Z ccl-logbot joined #sbcl 2014-04-08T13:38:00Z 2014-04-08T13:38:00Z names: ccl-logbot drmeister oleo ltbarcly milanj harovali eudoxia DGASAU foom loke scymtym reb`` nyef yacks pchrist xymox kludge` nicdev` joshe ams faheem Posterdati pkhuong flip214 ferada irsol ivan`` jdz mood brucem jsnell_ specbot minion yauz fe[nl]ix leoc luis asedeno Munksgaard tmh_ @Krystof redline6561` jaimef fikusz hzp White__Flame slyrus kanru attila_lendvai loke_ Blkt hlavaty antoszka afleck christoph_debian milosn angavrilov edgar-rft pjb 2014-04-08T13:48:29Z nicdev` is now known as nicdev 2014-04-08T13:49:23Z |3b|`` joined #sbcl 2014-04-08T14:09:17Z psilord joined #sbcl 2014-04-08T14:12:41Z ltbarcly quit (Ping timeout: 246 seconds) 2014-04-08T14:13:54Z cmack joined #sbcl 2014-04-08T14:14:30Z ltbarcly joined #sbcl 2014-04-08T14:17:12Z segv- joined #sbcl 2014-04-08T14:45:04Z flip214 quit (*.net *.split) 2014-04-08T14:45:06Z edgar-rft quit (*.net *.split) 2014-04-08T14:45:06Z pjb quit (*.net *.split) 2014-04-08T14:45:08Z angavrilov quit (*.net *.split) 2014-04-08T14:45:09Z christoph_debian quit (*.net *.split) 2014-04-08T14:45:11Z antoszka quit (*.net *.split) 2014-04-08T14:45:13Z kanru quit (*.net *.split) 2014-04-08T14:45:14Z White__Flame quit (*.net *.split) 2014-04-08T14:45:16Z fikusz quit (*.net *.split) 2014-04-08T14:45:16Z jaimef quit (*.net *.split) 2014-04-08T14:45:16Z redline6561` quit (*.net *.split) 2014-04-08T14:45:17Z asedeno quit (*.net *.split) 2014-04-08T14:45:17Z luis quit (*.net *.split) 2014-04-08T14:45:20Z minion quit (*.net *.split) 2014-04-08T14:45:20Z specbot quit (*.net *.split) 2014-04-08T14:45:50Z luis joined #sbcl 2014-04-08T14:46:09Z specbot joined #sbcl 2014-04-08T14:46:18Z edgar-rft joined #sbcl 2014-04-08T14:46:23Z minion joined #sbcl 2014-04-08T14:46:54Z pjb joined #sbcl 2014-04-08T14:47:11Z redline6561 joined #sbcl 2014-04-08T14:48:23Z fikusz joined #sbcl 2014-04-08T14:48:56Z White_Flame joined #sbcl 2014-04-08T14:50:42Z flip214 joined #sbcl 2014-04-08T14:50:42Z flip214 quit (Changing host) 2014-04-08T14:50:42Z flip214 joined #sbcl 2014-04-08T14:51:23Z antoszka joined #sbcl 2014-04-08T14:53:38Z |3b|`` is now known as |3b| 2014-04-08T14:55:22Z angavrilov joined #sbcl 2014-04-08T14:55:37Z segv- quit (Remote host closed the connection) 2014-04-08T14:55:51Z asedeno joined #sbcl 2014-04-08T14:58:38Z segv- joined #sbcl 2014-04-08T14:58:56Z jaimef joined #sbcl 2014-04-08T15:17:54Z christoph_debian joined #sbcl 2014-04-08T15:19:47Z eudoxia quit (Quit: leaving) 2014-04-08T15:27:17Z jdz quit (Ping timeout: 255 seconds) 2014-04-08T15:29:26Z sdemarre joined #sbcl 2014-04-08T15:30:14Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-08T15:42:54Z chichimaru joined #sbcl 2014-04-08T15:49:28Z slyrus quit (Remote host closed the connection) 2014-04-08T15:51:23Z slyrus joined #sbcl 2014-04-08T15:53:30Z nyef: Would someone be so kind as to confirm the analysis in my last commit message? (d7b75c3bfa5316a78b986f049b70d54eb9b10746, compiler/sparc/subprim: Clean up a FIXME...) 2014-04-08T15:58:23Z ASau joined #sbcl 2014-04-08T15:58:43Z ltbarcly joined #sbcl 2014-04-08T16:23:51Z ltbarcly quit (Read error: Connection reset by peer) 2014-04-08T16:24:25Z ltbarcly joined #sbcl 2014-04-08T16:26:41Z chichimaru left #sbcl 2014-04-08T16:31:19Z sdemarre quit (Ping timeout: 240 seconds) 2014-04-08T16:40:56Z Bike joined #sbcl 2014-04-08T16:43:48Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-08T16:45:20Z fe[nl]ix quit (Remote host closed the connection) 2014-04-08T16:45:21Z Blkt quit (Remote host closed the connection) 2014-04-08T16:45:29Z Blkt joined #sbcl 2014-04-08T16:45:32Z fe[nl]ix joined #sbcl 2014-04-08T17:02:15Z fiveop joined #sbcl 2014-04-08T17:02:47Z fiveop: Is douglas sometimes here? 2014-04-08T17:03:27Z nyef: douglas who? 2014-04-08T17:09:38Z fe[nl]ix: catman 2014-04-08T17:10:43Z Krystof: not as far as I know 2014-04-08T17:12:33Z fiveop: I like that he's cleaning things up. 2014-04-08T17:14:38Z ltbarcly joined #sbcl 2014-04-08T17:16:09Z nyef: Krystof: If you have a couple of minutes, would you mind having a look at commit d7b75c3bfa5316a78b986f049b70d54eb9b10746 and letting me know if I'm right about the circumstances under which the first type test can trigger? 2014-04-08T17:32:01Z fcsa joined #sbcl 2014-04-08T17:32:32Z ltbarcly quit (Quit: Textual IRC Client: www.textualapp.com) 2014-04-08T17:32:41Z fcsa: Hello, #sbcl. I would like to present a free as in freedom Skype replacement: Tox. Tox is a P2P, totally distributed, trustless network. The core is written in ANSI C and there are many clients/frontends for it. 2014-04-08T17:32:46Z fcsa: I would like to invite you to make a SBCL wrapper and Client for it, as there is no good Multi-platform (Windows, Linux, OS X, BSD, etc.) for it. 2014-04-08T17:32:53Z fcsa: Website: http://tox.im/ Github: https://github.com/irungentoo/ProjectTox-Core Development Introduction: http://api.libtoxcore.so/ 2014-04-08T17:32:55Z milanj quit (Ping timeout: 240 seconds) 2014-04-08T17:33:06Z nyef: Okay, now you're just spamming. 2014-04-08T17:33:33Z fcsa: Sorry, nyef, I thought there could be someone only at #sbcl and not at #lisp 2014-04-08T17:33:43Z ltbarcly joined #sbcl 2014-04-08T17:33:54Z fcsa: So besides #lisp I tried #guile and #sbcl as well, my bad. I apologize. 2014-04-08T17:34:17Z nyef: #guile might be a good choice, I'd expect far less overlap there. 2014-04-08T17:35:46Z angavrilov quit (Remote host closed the connection) 2014-04-08T17:39:43Z oleo quit (Ping timeout: 240 seconds) 2014-04-08T17:50:33Z ltbarcly_ joined #sbcl 2014-04-08T17:52:42Z ltbarcly quit (Ping timeout: 240 seconds) 2014-04-08T18:23:09Z oleo joined #sbcl 2014-04-08T18:48:23Z attila_lendvai quit (Quit: Leaving.) 2014-04-08T18:48:32Z attila_lendvai joined #sbcl 2014-04-08T18:48:34Z attila_lendvai quit (Changing host) 2014-04-08T18:48:34Z attila_lendvai joined #sbcl 2014-04-08T18:50:03Z attila_lendvai1 joined #sbcl 2014-04-08T18:50:03Z attila_lendvai quit (Disconnected by services) 2014-04-08T18:50:03Z attila_lendvai1 quit (Changing host) 2014-04-08T18:50:03Z attila_lendvai1 joined #sbcl 2014-04-08T18:58:32Z Krystof: nyef: you shouldn't take my word for it, but I think your reasoning is sound 2014-04-08T18:59:23Z nyef: Okay, thanks. 2014-04-08T18:59:39Z nyef: I've also got a patch for doing the same on PPC. 2014-04-08T19:27:56Z ltbarcly_ quit (Quit: Textual IRC Client: www.textualapp.com) 2014-04-08T19:29:07Z fcsa quit (Quit: Page closed) 2014-04-08T19:30:13Z ltbarcly joined #sbcl 2014-04-08T19:36:17Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-08T19:40:47Z fiveop quit 2014-04-08T19:59:18Z scymtym_ joined #sbcl 2014-04-08T20:18:40Z Krystof quit (Read error: Operation timed out) 2014-04-08T20:25:49Z ltbarcly joined #sbcl 2014-04-08T20:32:38Z attila_lendvai1 quit (Ping timeout: 240 seconds) 2014-04-08T20:36:33Z milanj joined #sbcl 2014-04-08T20:51:59Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-08T21:02:28Z Krystof joined #sbcl 2014-04-08T21:02:28Z ChanServ has set mode +o Krystof 2014-04-08T21:04:53Z slyrus_ joined #sbcl 2014-04-08T21:06:18Z slyrus quit (Ping timeout: 240 seconds) 2014-04-08T21:06:33Z slyrus_ is now known as slyrus 2014-04-08T21:13:18Z ltbarcly joined #sbcl 2014-04-08T21:13:34Z ltbarcly quit (Client Quit) 2014-04-08T21:16:08Z ltbarcly joined #sbcl 2014-04-08T21:42:18Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-08T21:48:49Z ltbarcly joined #sbcl 2014-04-08T21:48:49Z ltbarcly quit (Client Quit) 2014-04-08T21:51:27Z davazp joined #sbcl 2014-04-08T21:56:13Z drmeister quit (Remote host closed the connection) 2014-04-08T22:02:57Z slyrus quit (Ping timeout: 250 seconds) 2014-04-08T22:09:37Z drmeister joined #sbcl 2014-04-08T22:13:38Z slyrus joined #sbcl 2014-04-08T22:22:12Z ltbarcly joined #sbcl 2014-04-08T22:26:24Z psilord quit (Quit: Leaving.) 2014-04-08T22:32:56Z drmeister quit (Read error: Connection reset by peer) 2014-04-08T22:43:06Z slyrus quit (Ping timeout: 240 seconds) 2014-04-08T22:43:07Z drmeister joined #sbcl 2014-04-08T22:44:06Z slyrus joined #sbcl 2014-04-08T22:48:25Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-08T22:51:42Z drmeister quit (Remote host closed the connection) 2014-04-08T23:07:53Z slyrus quit (Remote host closed the connection) 2014-04-08T23:15:12Z ltbarcly joined #sbcl 2014-04-08T23:20:47Z segv- quit (Ping timeout: 246 seconds) 2014-04-08T23:30:15Z cmack` joined #sbcl 2014-04-08T23:31:38Z cmack quit (Ping timeout: 246 seconds) 2014-04-08T23:44:19Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-09T00:03:55Z ltbarcly joined #sbcl 2014-04-09T00:19:58Z drmeister joined #sbcl 2014-04-09T00:29:25Z antoszka quit (Ping timeout: 265 seconds) 2014-04-09T00:29:40Z antoszka joined #sbcl 2014-04-09T00:31:18Z milanj quit (Ping timeout: 240 seconds) 2014-04-09T00:48:14Z harovali quit (Ping timeout: 240 seconds) 2014-04-09T00:48:37Z psilord joined #sbcl 2014-04-09T00:49:00Z harovali joined #sbcl 2014-04-09T01:29:29Z segv- joined #sbcl 2014-04-09T01:33:04Z nyef: So, anyone up for some sort of compile-time type-system or type-propagation bug causing spurious deletion of typecase clauses? 2014-04-09T01:34:18Z nyef: (Was just reported by jasom in #lisp.) 2014-04-09T01:35:30Z pkhuong: sure 2014-04-09T01:37:06Z nyef: http://paste.lisp.org/display/141960#1 looks quite interesting in compiler trace output. 2014-04-09T01:37:15Z nyef: No sign of the (vector * 8) clause at all. 2014-04-09T01:38:05Z nyef: And I have no clue as to what's going on there, as I have a hard enough time with the normal type system, let alone the compile-time magic. 2014-04-09T01:41:32Z pkhuong: looks like a constraint bug 2014-04-09T01:41:48Z pkhuong: (vector * 4) expands into (and (vectorp ...) [length check]) 2014-04-09T01:43:00Z pkhuong: the if-false successor of vectorp is the (vector * 8) type test and it gets marked as dead immediately. 2014-04-09T01:43:45Z nyef: Ugh. The if-false successor of VECTORP? 2014-04-09T01:44:02Z pkhuong: yeah 2014-04-09T01:44:24Z nyef: How does THAT happen? 2014-04-09T01:45:45Z pkhuong: this is way worse. 2014-04-09T01:46:03Z pkhuong: add a (not vector) clause at the head of that typecase 2014-04-09T01:46:14Z pkhuong: and gets converted into if 2014-04-09T01:48:50Z nyef: ... Wha...? That's vicious. 2014-04-09T01:50:47Z pkhuong: vicious? what part? 2014-04-09T01:50:57Z pkhuong: converting AND into IF? 2014-04-09T01:50:59Z nyef: Dumping almost the entire typecase? 2014-04-09T01:51:04Z pkhuong: oh yeah. 2014-04-09T01:51:18Z pkhuong: clearly it's got to do with constraint. 2014-04-09T01:52:21Z nyef: It doesn't happen in 1.0.23, so it's a regression at some point. 2014-04-09T01:53:21Z pkhuong: i'd blame my if-if rewriting or constraint 2014-04-09T01:54:38Z nyef: Mmm. Okay, so there's a starting point. 2014-04-09T01:59:54Z scymtym_ quit (Ping timeout: 240 seconds) 2014-04-09T02:07:49Z nyef: I'll start bisecting this mess... probably tomorrow. 2014-04-09T02:10:57Z pkhuong: we got some serious nconcing going on somewhere 2014-04-09T02:18:59Z pkhuong: bah, no. foiled by overly smart lvar printing 2014-04-09T02:19:34Z jaimef quit (Excess Flood) 2014-04-09T02:19:52Z nyef: Ugh. I'm bisecting from a revision made while ths was still with us. 2014-04-09T02:24:35Z jaimef joined #sbcl 2014-04-09T02:28:42Z segv- quit (Ping timeout: 240 seconds) 2014-04-09T02:30:18Z ltbarcly quit (Ping timeout: 240 seconds) 2014-04-09T02:35:25Z kanru joined #sbcl 2014-04-09T02:36:41Z joshe quit 2014-04-09T02:38:30Z joshe joined #sbcl 2014-04-09T02:38:41Z joshe quit (Remote host closed the connection) 2014-04-09T02:38:54Z christoph_debian quit (Ping timeout: 240 seconds) 2014-04-09T02:40:25Z pkhuong: got it down to somehow propagating that a variable is both VECTORP and not VECTORP 2014-04-09T02:40:40Z joshe joined #sbcl 2014-04-09T02:40:42Z pkhuong: sprinkling avers everywhere to figure out where that happens 2014-04-09T02:41:06Z pkhuong: but so far, it looks like it's downstream of any IF node 2014-04-09T02:42:18Z joshe quit (Remote host closed the connection) 2014-04-09T02:43:02Z joshe joined #sbcl 2014-04-09T02:43:27Z nyef: I'm on the second bisection step of about ten. 2014-04-09T02:44:13Z nyef: At least this isn't my PPC or my SPARC. (-: 2014-04-09T02:47:56Z jaimef: nyef what OS? 2014-04-09T02:48:05Z nyef: Linux, in all cases. 2014-04-09T02:48:19Z jaimef: oic 2014-04-09T02:48:29Z ltbarcly joined #sbcl 2014-04-09T02:48:43Z nyef: The issue being that a full build takes 40 minutes on my PPC, and far longer on my SPARC. 2014-04-09T02:49:05Z jaimef: what model ppc? 2014-04-09T02:49:19Z jaimef: nyef: what does "openssl speed aes-256-cbc" report on the PPC and the SPARC? 2014-04-09T02:49:50Z jaimef: I need to see if I can get sbcl working on half of my platforms. 2014-04-09T02:50:17Z nyef: The PPC is an iMac G5 ALS, and the SPARC is a 600MHz SunBlade 150. Both are currently switched off. 2014-04-09T02:50:33Z jaimef: oh clearly that's an easy guess :P 2014-04-09T02:50:42Z jaimef: sounds like the ppc is not optimized very well 2014-04-09T02:51:26Z jaimef never found a linux distro that worked well on the g5. what are you using 2014-04-09T02:51:30Z nyef: A MIPS build on what I have available would take even longer, as it's a DD-WRT box. 2014-04-09T02:51:36Z nyef: I'm using an oldish Debian. 2014-04-09T02:52:07Z joshe: that's a nice ppc, I think mine takes about an hour 2014-04-09T02:52:08Z jaimef: fastest mips I have is a r12k 2014-04-09T02:52:19Z nyef: Modern Debian on a desktop machine is obnoxious. 2014-04-09T02:52:22Z jaimef: yeah G5 was pretty fast. compare to power 6 even 2014-04-09T02:52:30Z joshe: I have faster sparc though 2014-04-09T02:52:35Z christoph_debian joined #sbcl 2014-04-09T02:52:39Z jaimef gets too many handme downs 2014-04-09T02:52:47Z nyef: Fastest MIPS I have is either a PS2 or a Linksys E3000. 2014-04-09T02:52:58Z joshe: no mips though, uninteresting as I couldn't run sbcl there on openbsd 2014-04-09T02:53:02Z jaimef: http://linbsd.org/rack.png no x86 there :P 2014-04-09T02:53:24Z jaimef: openbsd have good sbcl support? I use netbsd exclusivel 2014-04-09T02:53:27Z joshe: http://elsasser.org/misc/tower.jpg 2014-04-09T02:53:37Z joshe: quite out of date, but you get the idea 2014-04-09T02:54:07Z joshe: that's a lot of sun gear 2014-04-09T02:54:22Z joshe: and twice as many nextstations as mine :) 2014-04-09T02:54:27Z nyef: You both have '040 NeXT hardware? 2014-04-09T02:54:47Z jaimef: yes, still up and running netbsd 2014-04-09T02:54:50Z joshe: mine didn't boot the last time I tried :/ 2014-04-09T02:54:51Z jaimef: boots SOO fast 2014-04-09T02:55:07Z jaimef: makes the bebox seem like poop. bug bebox has to boot floppy 2014-04-09T02:55:22Z jaimef: keeping working floppies around is harder than keeping disks working on this ancient crap 2014-04-09T02:55:49Z joshe: I had to buy an x86 floppy drive the last time I installed openbsd on 68k hardware 2014-04-09T02:56:04Z jaimef: :D 2014-04-09T02:56:13Z jaimef: the next has a great network boot rom. but only 10mb 2014-04-09T02:56:25Z joshe: but I admit I've lost my taste for the really old crap 2014-04-09T02:56:27Z jaimef: so I have to keep a 10/100 router around to netboot it. 2014-04-09T02:56:35Z jaimef: yeah. mostly here as well 2014-04-09T02:56:37Z joshe: (although I still want a microvax) 2014-04-09T02:56:44Z jaimef: got bebox working, and it was total poo 2014-04-09T02:56:56Z nyef: Either of you have a decent Alpha? 2014-04-09T02:57:04Z joshe: I have no alpha 2014-04-09T02:58:44Z jaimef: mine is a little optia 2014-04-09T02:58:46Z jaimef: 166mhz 2014-04-09T02:58:59Z jaimef: the alpha's about evaporated 2014-04-09T02:59:05Z jaimef: Multia that's it 2014-04-09T03:01:55Z pkhuong: # U # -> # 2014-04-09T03:02:02Z pkhuong: awesome. 2014-04-09T03:02:22Z pkhuong: Krystof: your turn 2014-04-09T03:05:03Z pkhuong: (sb-kernel:type-union (sb-kernel:specifier-type 'string) (sb-kernel:specifier-type '(vector * 4))) => # 2014-04-09T03:05:07Z pkhuong: really doesn't look right (: 2014-04-09T03:07:29Z pkhuong: nyef: looks like an old bug that got tickled by improvements to our typep-propagation 2014-04-09T03:07:52Z nyef: Possibly. Still bisecting here, and will probably have to finish the bisection tomorrow. 2014-04-09T03:08:09Z pkhuong: got it down to a type union bug 2014-04-09T03:08:15Z nyef: Also trying to remember my way around the whole type-class mess. 2014-04-09T03:08:16Z pkhuong: that should be easier to look at. 2014-04-09T03:09:35Z pkhuong: I'd expect 8be78b41ac1695e1c9b5d7f6c8bd41e3c7e6e6fc 2014-04-09T03:09:46Z nyef: Yeah, that still happens on 1.0.23, so... 2014-04-09T03:11:26Z pkhuong: what still happens? the bad union? 2014-04-09T03:11:31Z nyef: Yup. 2014-04-09T03:11:42Z pkhuong: array types are subtly broken in many ways. 2014-04-09T03:12:44Z pkhuong: Starting at 8be78, we got a lot cleverer about typep (and thus typecase) 2014-04-09T03:13:27Z pkhuong: a bit too clever, in fact, with array types (we should only propagate upgraded types) 2014-04-09T03:21:02Z nyef: Already down to the 1.1.12 to 1.1.16 range. 2014-04-09T03:26:09Z pkhuong: in the worst case constrain-ref-type could start using quick and lossy union/intersection 2014-04-09T03:28:51Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-09T03:28:54Z drmeister quit (Remote host closed the connection) 2014-04-09T03:29:22Z pkhuong: it'd probably be a compile time win with little effect on code quality... but the union bug could still bite us anywhere else 2014-04-09T03:32:47Z nyef: Hang on, string-type on the one side is a union of three specicalized vector types without length restrictions, and (vector * 4) on the other is an unspecialized vector type WITH a length restriction... There's a certain twisted sense to the union being VECTOR. 2014-04-09T03:33:57Z ltbarcly joined #sbcl 2014-04-09T03:34:54Z pkhuong: right. I can see the sort of logic that would get there. 2014-04-09T03:35:02Z pkhuong: wrong but plausible 2014-04-09T03:38:37Z nyef: TYPE-UNION2 on base-string and (vector * 4) returns vector. 2014-04-09T03:38:56Z nyef: This is called from simplify-unions. 2014-04-09T03:38:56Z pkhuong: awesome. no quick fix for me. 2014-04-09T03:39:29Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-09T03:40:23Z nyef: How so? 2014-04-09T03:41:40Z pkhuong: I was going to try and call type-union2 directly and hope that avoiding the simplification logic would be enough to hide the issue 2014-04-09T03:41:46Z nyef: Ah. 2014-04-09T03:42:04Z nyef: Unfortunately, type-union2 appears to BE the issue. 2014-04-09T03:42:14Z yacks quit (Ping timeout: 240 seconds) 2014-04-09T03:42:46Z nyef: late-type.lisp, around line 2590. 2014-04-09T03:43:09Z yacks joined #sbcl 2014-04-09T03:44:14Z pkhuong: this csubtypep eltype logic doesn't look right 2014-04-09T03:44:25Z pkhuong: but anyway 2014-04-09T03:45:17Z pkhuong: I see what it does 2014-04-09T03:45:23Z pkhuong: this is about the declared eltype 2014-04-09T03:46:43Z pkhuong: its treatment of dimension is completely wrong 2014-04-09T03:47:09Z pkhuong: it should just return nil when there isn't a clear subsumption 2014-04-09T03:49:23Z pkhuong: yeah. This requires some more thinking. good night 2014-04-09T03:49:33Z nyef: Heh. G'night. 2014-04-09T03:56:38Z pkhuong: well, it's an easy solution, because there is no non-trivial solution. 2014-04-09T03:56:52Z pkhuong: array types can only express cartesian products 2014-04-09T03:57:30Z pkhuong: if neither array type is a subtype of the other, there is no array type that covers exactly their union and nothing more 2014-04-09T03:58:04Z pkhuong: the specified eltype is the only exception: we can take the union there. 2014-04-09T03:58:14Z pkhuong: (if everything else is equal) 2014-04-09T03:59:12Z pkhuong: Y axis has 2 rows: T or BIT. X axis has 2 columns: 1, 2 dimensions. 2014-04-09T04:01:02Z pkhuong: array types can only express rectangles that choose a single row/column, or all rows/all columns 2014-04-09T04:01:12Z pkhuong: so if we have two incomparable types, we're screwed. 2014-04-09T04:02:45Z nyef: I think I understand what you're getting at, but not sufficiently to be able to try and write a patch. 2014-04-09T04:03:10Z pkhuong: and I should really get some sleep. 2014-04-09T04:03:51Z nyef: So should I, really. 2014-04-09T04:04:53Z nyef: I'll just update #lisp / jasom to the effect that we think we've found the underlying bug, but not yet a fix. 2014-04-09T04:05:30Z nyef: ... And I may-or-may-not finish the bisect tonight. Probably not. 2014-04-09T04:24:50Z segv- joined #sbcl 2014-04-09T04:36:14Z attila_lendvai joined #sbcl 2014-04-09T04:36:14Z attila_lendvai quit (Changing host) 2014-04-09T04:36:14Z attila_lendvai joined #sbcl 2014-04-09T04:45:49Z michael_lee joined #sbcl 2014-04-09T04:48:17Z michael_lee quit (Max SendQ exceeded) 2014-04-09T04:49:37Z michael_lee joined #sbcl 2014-04-09T04:54:55Z ASau quit (Ping timeout: 240 seconds) 2014-04-09T04:57:51Z michael_lee quit (Max SendQ exceeded) 2014-04-09T05:04:12Z oleo quit (Quit: Leaving) 2014-04-09T05:14:31Z michael_lee joined #sbcl 2014-04-09T05:17:48Z drmeister joined #sbcl 2014-04-09T05:19:10Z ltbarcly joined #sbcl 2014-04-09T05:19:10Z ltbarcly quit (Client Quit) 2014-04-09T05:21:59Z drmeister quit (Ping timeout: 246 seconds) 2014-04-09T05:22:19Z nyef quit (Quit: Why am I still awake this late?) 2014-04-09T05:34:44Z pranavrc joined #sbcl 2014-04-09T05:34:44Z pranavrc quit (Changing host) 2014-04-09T05:34:44Z pranavrc joined #sbcl 2014-04-09T05:36:56Z sdemarre joined #sbcl 2014-04-09T05:46:16Z jaimef quit (Excess Flood) 2014-04-09T05:51:37Z jaimef joined #sbcl 2014-04-09T06:09:43Z sdemarre left #sbcl 2014-04-09T06:32:29Z Krystof: gah 2014-04-09T06:32:42Z Krystof: pkhuong: I'm going on holiday! 2014-04-09T06:59:56Z angavrilov joined #sbcl 2014-04-09T07:06:34Z drmeister joined #sbcl 2014-04-09T07:11:11Z drmeister quit (Ping timeout: 246 seconds) 2014-04-09T07:40:03Z davazp quit (Remote host closed the connection) 2014-04-09T08:25:30Z milanj joined #sbcl 2014-04-09T08:32:21Z pranavrc_ joined #sbcl 2014-04-09T08:33:54Z pranavrc quit (Ping timeout: 240 seconds) 2014-04-09T08:55:24Z drmeister joined #sbcl 2014-04-09T08:59:03Z pranavrc_ quit (Read error: Connection reset by peer) 2014-04-09T08:59:54Z drmeister quit (Ping timeout: 240 seconds) 2014-04-09T09:06:35Z pranavrc joined #sbcl 2014-04-09T09:31:30Z kludge` quit (Ping timeout: 264 seconds) 2014-04-09T09:35:54Z milanj quit (Ping timeout: 240 seconds) 2014-04-09T09:36:15Z kludge` joined #sbcl 2014-04-09T09:41:40Z pranavrc quit (Read error: Connection reset by peer) 2014-04-09T09:48:21Z pranavrc joined #sbcl 2014-04-09T09:48:33Z pranavrc quit (Changing host) 2014-04-09T09:48:33Z pranavrc joined #sbcl 2014-04-09T09:52:18Z pranavrc quit (Read error: Connection reset by peer) 2014-04-09T09:55:58Z pranavrc joined #sbcl 2014-04-09T09:55:58Z pranavrc quit (Changing host) 2014-04-09T09:55:58Z pranavrc joined #sbcl 2014-04-09T09:56:13Z pranavrc quit (Read error: Connection reset by peer) 2014-04-09T09:57:28Z pranavrc joined #sbcl 2014-04-09T09:57:28Z pranavrc quit (Changing host) 2014-04-09T09:57:28Z pranavrc joined #sbcl 2014-04-09T09:57:30Z pranavrc quit (Read error: Connection reset by peer) 2014-04-09T10:05:35Z pranavrc joined #sbcl 2014-04-09T10:30:27Z pranavrc_ joined #sbcl 2014-04-09T10:32:55Z pranavrc quit (Ping timeout: 240 seconds) 2014-04-09T10:38:08Z pranavrc_ quit (Read error: Connection reset by peer) 2014-04-09T10:41:43Z pranavrc joined #sbcl 2014-04-09T10:44:07Z drmeister joined #sbcl 2014-04-09T10:46:07Z pranavrc quit (Read error: Connection reset by peer) 2014-04-09T10:48:32Z drmeister quit (Ping timeout: 246 seconds) 2014-04-09T10:49:11Z pranavrc joined #sbcl 2014-04-09T10:49:11Z pranavrc quit (Changing host) 2014-04-09T10:49:11Z pranavrc joined #sbcl 2014-04-09T10:54:16Z pranavrc_ joined #sbcl 2014-04-09T10:55:19Z pranavrc quit (Ping timeout: 240 seconds) 2014-04-09T10:56:55Z yacks quit (Ping timeout: 240 seconds) 2014-04-09T11:18:29Z yacks joined #sbcl 2014-04-09T11:42:31Z pranavrc_ quit (Ping timeout: 240 seconds) 2014-04-09T12:16:49Z pranavrc joined #sbcl 2014-04-09T12:25:47Z segv- quit (Remote host closed the connection) 2014-04-09T12:26:34Z nyef joined #sbcl 2014-04-09T12:26:42Z nyef: G'morning all. 2014-04-09T12:27:02Z segv- joined #sbcl 2014-04-09T12:28:40Z drmeister joined #sbcl 2014-04-09T12:29:19Z drmeister quit (Remote host closed the connection) 2014-04-09T12:31:08Z drmeister joined #sbcl 2014-04-09T12:32:44Z Krystof: morning 2014-04-09T12:32:56Z drmeister quit (Remote host closed the connection) 2014-04-09T12:51:16Z harovali1 joined #sbcl 2014-04-09T12:51:17Z harovali quit (Ping timeout: 250 seconds) 2014-04-09T13:01:13Z attila_lendvai quit (Quit: Leaving.) 2014-04-09T13:03:22Z drmeister joined #sbcl 2014-04-09T13:05:24Z pranavrc quit 2014-04-09T13:09:56Z pkhuong: actually, even declared eltypes can't be recursively type-unioned 2014-04-09T13:10:13Z nyef: Oh, good. Hello pkhuong. 2014-04-09T13:10:38Z pkhuong: another form of the bug where you declare a deftransform on (OR A B) instead of a transform on A and another on B 2014-04-09T13:11:50Z nyef: I'm on the last bisect step, having just marked the opaque-typep commit as bad. 2014-04-09T13:12:25Z pkhuong: otoh, it's probably *safe* to type union declared eltypes 2014-04-09T13:12:42Z pkhuong: because we will (should?) nuke them away on negation/type-difference 2014-04-09T13:14:27Z eudoxia joined #sbcl 2014-04-09T13:15:42Z fe[nl]ix: I'd like to merge the code in static-vectors into sbcl 2014-04-09T13:15:48Z fe[nl]ix: any gotcha that I should be aware of ? 2014-04-09T13:16:01Z fe[nl]ix: it's https://github.com/sionescu/static-vectors/blob/master/src/impl-sbcl.lisp 2014-04-09T13:16:43Z sdemarre joined #sbcl 2014-04-09T13:18:06Z nyef: Well, there's another string widetag, even if it's unlikely to be allocated as a static vector. 2014-04-09T13:18:59Z fe[nl]ix: apart from sb-vm:simple-base-string-widetag and sb-vm:simple-character-string-widetag ? 2014-04-09T13:19:18Z nyef: Yes. Arrays specialized to NIL are strings. 2014-04-09T13:19:35Z nyef: (NIL being a subtype of character) 2014-04-09T13:19:59Z fe[nl]ix: I see 2014-04-09T13:20:31Z drmeister quit (Ping timeout: 240 seconds) 2014-04-09T13:20:43Z Krystof: I don't think we allocate a final #\Null for them though 2014-04-09T13:20:54Z Krystof: I can't honestly remember whether we do for simple-character-strings 2014-04-09T13:21:08Z sdemarre quit (Ping timeout: 246 seconds) 2014-04-09T13:21:29Z Krystof: nyef: I'll fix the logic in array :simple-union2 2014-04-09T13:21:33Z Krystof: sometime soon 2014-04-09T13:22:18Z Krystof: honestly, everyone advocating rewriting the internet in safe languages should take a look at what happens when you let lisp programmes build type systems :-/ 2014-04-09T13:22:24Z nyef: Heh. 2014-04-09T13:23:24Z nyef: "You think that Lisp can't have buffer overruns? Think again." 2014-04-09T13:23:54Z Krystof: buffer overruns are harder, but we can certainly merrily delete all your code 2014-04-09T13:25:15Z drmeister joined #sbcl 2014-04-09T13:25:41Z pkhuong: so will gcc, excpet, in their case, the standard usually lets them ;) 2014-04-09T13:27:36Z nyef: Bisected to commit 8be78b41ac1695e1c9b5d7f6c8bd41e3c7e6e6fc, as pkhuong predicted. 2014-04-09T13:27:58Z oleo joined #sbcl 2014-04-09T13:47:05Z pkhuong: I should get myself some profiles of SBCL compiling large applications 2014-04-09T13:48:04Z pkhuong: I'd like to know if constraint propagation is slow because of constrain-ref-type. If so, there's an interesting localised project with easily quantifiable impact. 2014-04-09T13:48:25Z drmeister quit (Read error: Connection reset by peer) 2014-04-09T13:48:41Z drmeiste_ joined #sbcl 2014-04-09T13:49:28Z Kevin_Xi joined #sbcl 2014-04-09T13:50:00Z Kevin_Xi left #sbcl 2014-04-09T13:50:24Z pkhuong: also, quick way to probably get the propagation IR1 depends on but not much more: stop propagating along back edges. 2014-04-09T13:52:17Z pkhuong: logic being that we need it propagation for expressions like (if (foo-p x) [x is a foo] [better know that x is not a foo or the compiler will aver out]) 2014-04-09T13:52:54Z pkhuong: but afaik, these transforms only emit dispatch trees (or even ladders), never loops. 2014-04-09T13:54:08Z pkhuong: you'll often get horribad code from assuming that everything is of type T at loop entry, but that's life. 2014-04-09T13:54:38Z drmeiste_ quit (Ping timeout: 240 seconds) 2014-04-09T14:00:55Z drmeister joined #sbcl 2014-04-09T14:02:50Z milanj joined #sbcl 2014-04-09T14:17:50Z ltbarcly_ joined #sbcl 2014-04-09T14:20:10Z tmh_ quit (Ping timeout: 252 seconds) 2014-04-09T14:23:18Z ltbarcly_ quit (Ping timeout: 240 seconds) 2014-04-09T14:27:05Z tmh_ joined #sbcl 2014-04-09T14:36:58Z redline6561 quit (Quit: ZNC - http://znc.in) 2014-04-09T14:43:11Z drmeiste_ joined #sbcl 2014-04-09T14:45:19Z drmeister quit (Ping timeout: 240 seconds) 2014-04-09T14:48:16Z rpg joined #sbcl 2014-04-09T14:48:18Z ltbarcly joined #sbcl 2014-04-09T14:54:12Z ASau joined #sbcl 2014-04-09T14:57:30Z redline6561 joined #sbcl 2014-04-09T14:57:57Z rpg_ joined #sbcl 2014-04-09T15:01:54Z rpg quit (Ping timeout: 240 seconds) 2014-04-09T15:03:43Z ltbarcly quit (Ping timeout: 240 seconds) 2014-04-09T15:15:31Z yacks quit (Ping timeout: 240 seconds) 2014-04-09T15:16:39Z ltbarcly joined #sbcl 2014-04-09T15:21:53Z attila_lendvai joined #sbcl 2014-04-09T15:24:18Z ltbarcly quit (Ping timeout: 240 seconds) 2014-04-09T15:31:53Z Blkt quit (Quit: No Ping reply in 180 seconds.) 2014-04-09T15:32:29Z Blkt joined #sbcl 2014-04-09T15:51:27Z michael_lee quit (Quit: Ex-Chat) 2014-04-09T15:51:28Z foom quit (Read error: Connection reset by peer) 2014-04-09T16:17:15Z rpg joined #sbcl 2014-04-09T16:17:46Z sdemarre joined #sbcl 2014-04-09T16:19:30Z ltbarcly joined #sbcl 2014-04-09T16:20:55Z rpg_ quit (Ping timeout: 240 seconds) 2014-04-09T16:21:54Z rpg quit (Ping timeout: 240 seconds) 2014-04-09T16:23:50Z ltbarcly quit (Ping timeout: 246 seconds) 2014-04-09T16:31:47Z kludge` quit (Remote host closed the connection) 2014-04-09T16:42:10Z fiveop joined #sbcl 2014-04-09T16:45:29Z kludge` joined #sbcl 2014-04-09T16:47:39Z rpg joined #sbcl 2014-04-09T16:50:21Z fiveop quit 2014-04-09T16:52:18Z rpg quit (Ping timeout: 240 seconds) 2014-04-09T16:55:06Z drmeiste_ quit (Remote host closed the connection) 2014-04-09T16:55:45Z drmeister joined #sbcl 2014-04-09T17:12:11Z ltbarcly joined #sbcl 2014-04-09T17:15:28Z rpg joined #sbcl 2014-04-09T17:16:42Z rpg_ joined #sbcl 2014-04-09T17:16:42Z sdemarre quit (Ping timeout: 246 seconds) 2014-04-09T17:20:18Z rpg quit (Ping timeout: 240 seconds) 2014-04-09T17:25:05Z yacks joined #sbcl 2014-04-09T17:42:29Z attila_lendvai quit (Quit: Leaving.) 2014-04-09T18:16:17Z milanj quit (Ping timeout: 250 seconds) 2014-04-09T18:34:55Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-09T18:37:47Z eudoxia quit (Quit: Lost terminal) 2014-04-09T18:41:36Z ltbarcly joined #sbcl 2014-04-09T18:44:15Z fiveop joined #sbcl 2014-04-09T18:46:54Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-09T18:50:51Z oleo quit (Ping timeout: 246 seconds) 2014-04-09T18:52:29Z oleo joined #sbcl 2014-04-09T19:26:25Z ltbarcly joined #sbcl 2014-04-09T19:32:05Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-09T19:43:19Z milanj joined #sbcl 2014-04-09T20:16:55Z angavrilov quit (Remote host closed the connection) 2014-04-09T20:38:54Z yacks quit (Ping timeout: 240 seconds) 2014-04-09T20:40:37Z Hydan joined #sbcl 2014-04-09T20:41:03Z scymtym_ joined #sbcl 2014-04-09T20:45:01Z ltbarcly joined #sbcl 2014-04-09T21:01:38Z fiveop quit 2014-04-09T21:56:02Z drmeister quit (Remote host closed the connection) 2014-04-09T21:59:36Z tmh_ quit (Changing host) 2014-04-09T21:59:36Z tmh_ joined #sbcl 2014-04-09T22:05:44Z rpg joined #sbcl 2014-04-09T22:07:51Z drmeister joined #sbcl 2014-04-09T22:09:05Z nyef: Okay, still trying to understand the semantics for array type union. Would I be correct in saying that (array * (3)) and (array * (*)) should unify, but that (array * (3 *)) and (array * (* 3)) should not? 2014-04-09T22:09:25Z rpg_ quit (Ping timeout: 250 seconds) 2014-04-09T22:09:45Z Krystof: yes 2014-04-09T22:10:10Z nyef: I think I have a model as to how this needs to work, then. 2014-04-09T22:10:17Z rpg quit (Ping timeout: 250 seconds) 2014-04-09T22:10:38Z Krystof: ok. I was going to try to fix it sometime at the weekend, but if you beat me to it that's so totally fine it is as fine as a sunny day on the equator 2014-04-09T22:11:42Z nyef: Heh. I'm not sure how much time I really should be spending on this right now, but it's an interesting problem... And I'm not going to commit anything until I've seen an improvement in the generated code for the test case we have AND it's been reviewed by both you and pkhuong. 2014-04-09T22:12:38Z Krystof: OK :) basically what I would do is be much more restrictive in the WHEN clause in array :simple-union2 2014-04-09T22:13:01Z Krystof: returning NIL from that should be what is needed when the types don't have a nice union 2014-04-09T22:13:18Z nyef: Yeah, that's what I see, but the actual unified type might also need work. 2014-04-09T22:15:04Z Krystof: a little 2014-04-09T22:15:06Z Krystof: good 2014-04-09T22:15:59Z nyef: I'm also thinking to write some test cases and explanation for what's what. 2014-04-09T22:16:05Z drmeister quit (Ping timeout: 252 seconds) 2014-04-09T22:16:55Z slyrus joined #sbcl 2014-04-09T22:30:07Z eudoxia joined #sbcl 2014-04-09T22:30:58Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-09T22:33:28Z milanj quit (Quit: Leaving) 2014-04-09T22:36:08Z rpg joined #sbcl 2014-04-09T22:38:06Z ltbarcly joined #sbcl 2014-04-09T22:46:22Z cmack` quit (Remote host closed the connection) 2014-04-09T22:46:36Z rpg_ joined #sbcl 2014-04-09T22:49:07Z drmeister joined #sbcl 2014-04-09T22:50:42Z rpg quit (Ping timeout: 240 seconds) 2014-04-09T23:39:23Z rpg joined #sbcl 2014-04-09T23:43:06Z rpg_ quit (Ping timeout: 240 seconds) 2014-04-09T23:44:19Z rpg quit (Ping timeout: 250 seconds) 2014-04-09T23:45:41Z Hydan quit (Remote host closed the connection) 2014-04-09T23:45:56Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-09T23:47:36Z pkhuong: I don't know that I like being acquainted with the type system ;) 2014-04-09T23:50:51Z pkhuong: I don't know if it's an issue that we could sometimes find a nice union by looking at > 2 array types at a time 2014-04-09T23:53:40Z pkhuong: proably not, it just means that we'll unparse to nasty looking array types 2014-04-09T23:54:29Z afleck: Why does make-array throw an error when I try to do 2014-04-09T23:54:46Z afleck: (make-array ((+ 2 2) 3)) 2014-04-09T23:54:50Z nyef: I don't know that we CAN find a nice union of more than two array types unless it's a union of an exhaustive partition of wild-type. 2014-04-09T23:54:52Z pkhuong: afleck: ask #Lisp 2014-04-09T23:54:55Z eudoxia quit (Quit: leaving) 2014-04-09T23:56:00Z pkhuong: nyef: that's the case I was thinking of 2014-04-09T23:56:23Z pkhuong: aka the union of T arrays and specialised array types. 2014-04-09T23:56:48Z nyef: Yes, but it would have to be every specialized array type for the case to trigger. 2014-04-09T23:57:03Z pkhuong: yes... but we do have a type that's the union of every specialised array types 2014-04-09T23:57:49Z nyef: We do? 2014-04-09T23:58:44Z pkhuong: (sb-kernel:specifier-type 'sb-kernel:simple-unboxed-array) 2014-04-09T23:59:36Z nyef: Ahh. 2014-04-10T00:00:14Z nyef: So if we union that with (array t), we should get (array *)... 2014-04-10T00:00:58Z pkhuong: but we don't and type= and subtypep are all confused. 2014-04-10T00:02:29Z nyef: Ugh. More complexity. THIS is why I don't like messing with the type system. 2014-04-10T00:04:01Z pkhuong: I think I'd try something with a set of upgraded eltypes if I were to redo array types 2014-04-10T00:05:07Z nyef: Where does (ARRAY NIL) fit into this? 2014-04-10T00:05:31Z pkhuong: in what? 2014-04-10T00:06:23Z nyef: I'm... not sure anymore. Somewhat fried right now. /-: 2014-04-10T00:09:48Z rpg joined #sbcl 2014-04-10T00:19:45Z rpg_ joined #sbcl 2014-04-10T00:20:10Z rpg quit (Ping timeout: 276 seconds) 2014-04-10T00:21:00Z rpg joined #sbcl 2014-04-10T00:24:42Z rpg_ quit (Ping timeout: 240 seconds) 2014-04-10T00:34:00Z DGASAU` joined #sbcl 2014-04-10T00:34:34Z hlavaty quit (Read error: Operation timed out) 2014-04-10T00:34:47Z hlavaty joined #sbcl 2014-04-10T00:36:02Z DGASAU quit (Ping timeout: 265 seconds) 2014-04-10T00:52:58Z harovali joined #sbcl 2014-04-10T00:53:39Z harovali1 quit (Ping timeout: 250 seconds) 2014-04-10T00:53:40Z rpg quit (Quit: rpg) 2014-04-10T00:55:18Z slyrus quit (Ping timeout: 240 seconds) 2014-04-10T01:18:03Z specbot quit (Disconnected by services) 2014-04-10T01:18:07Z specbot joined #sbcl 2014-04-10T01:18:28Z minion quit (Disconnected by services) 2014-04-10T01:18:32Z minion joined #sbcl 2014-04-10T01:19:33Z edgar-rfx joined #sbcl 2014-04-10T01:20:11Z pjb` joined #sbcl 2014-04-10T01:21:34Z flip214_ joined #sbcl 2014-04-10T01:22:16Z flip214 quit (Ping timeout: 265 seconds) 2014-04-10T01:22:18Z fikusz quit (Ping timeout: 265 seconds) 2014-04-10T01:22:19Z edgar-rft quit (Ping timeout: 265 seconds) 2014-04-10T01:22:19Z pjb quit (Ping timeout: 265 seconds) 2014-04-10T01:22:52Z fikusz joined #sbcl 2014-04-10T01:44:02Z michael_lee joined #sbcl 2014-04-10T01:52:30Z DGASAU` quit (Ping timeout: 240 seconds) 2014-04-10T02:20:18Z eudoxia joined #sbcl 2014-04-10T02:21:34Z eudoxia quit (Client Quit) 2014-04-10T02:22:41Z nyef: Should the union of (array * (*)) and (array * (* *)) ever be (array * *) (that is, array)? 2014-04-10T02:23:10Z pkhuong: no 2014-04-10T02:23:20Z pkhuong: the third includes (array * (* * *)) 2014-04-10T02:24:12Z nyef: Okay, I figured something like that would be the case. 2014-04-10T02:25:34Z nyef: And the union of (array (integer 15 27)) and (array (integer 29 50)) should be (array (or (integer 15 27) (integer 29 50))), because they share a UAET of ub7. 2014-04-10T02:26:14Z pkhuong: actually... strictly speaking no 2014-04-10T02:26:18Z nyef: No? 2014-04-10T02:26:45Z pkhuong: #(20 30) would be a member of the union but of neither original types 2014-04-10T02:26:58Z nyef: Ah, fair enough. 2014-04-10T02:27:02Z pkhuong: but I don't think it matters for what we can use the declared type for. 2014-04-10T02:27:23Z eudoxia joined #sbcl 2014-04-10T02:27:38Z pkhuong: Krystof understands better the subtlety re declared array element type 2014-04-10T02:28:04Z nyef: Can use legitimately, can use in terms of the limits of possible analysis, or can use in terms of currently implemented analysis? 2014-04-10T02:28:21Z pkhuong: legitimately 2014-04-10T02:29:37Z pkhuong: as well as how we currently use the info, illegitimately. 2014-04-10T02:38:42Z christoph_debian quit (Ping timeout: 240 seconds) 2014-04-10T02:45:51Z eudoxia quit (Quit: leaving) 2014-04-10T02:52:55Z christoph_debian joined #sbcl 2014-04-10T03:20:57Z yacks joined #sbcl 2014-04-10T03:36:31Z attila_lendvai joined #sbcl 2014-04-10T03:36:31Z attila_lendvai quit (Changing host) 2014-04-10T03:36:31Z attila_lendvai joined #sbcl 2014-04-10T03:42:53Z nyef quit (Quit: G'night all.) 2014-04-10T04:23:54Z scymtym_ quit (Ping timeout: 240 seconds) 2014-04-10T04:48:41Z pranavrc joined #sbcl 2014-04-10T04:48:47Z pranavrc quit (Changing host) 2014-04-10T04:48:47Z pranavrc joined #sbcl 2014-04-10T05:05:56Z oleo quit (Quit: Leaving) 2014-04-10T05:23:15Z slyrus joined #sbcl 2014-04-10T05:28:23Z yacks quit (Ping timeout: 250 seconds) 2014-04-10T05:52:54Z |3b| quit (Remote host closed the connection) 2014-04-10T05:58:29Z |3b| joined #sbcl 2014-04-10T06:04:35Z yacks joined #sbcl 2014-04-10T06:43:42Z ASau quit (Ping timeout: 240 seconds) 2014-04-10T06:47:13Z ehaliewicz joined #sbcl 2014-04-10T06:48:54Z kanru quit (Remote host closed the connection) 2014-04-10T06:49:37Z kanru joined #sbcl 2014-04-10T06:51:14Z pjb` is now known as pjb 2014-04-10T06:57:57Z pranavrc quit (Ping timeout: 276 seconds) 2014-04-10T07:03:19Z pranavrc joined #sbcl 2014-04-10T07:03:19Z pranavrc quit (Changing host) 2014-04-10T07:03:19Z pranavrc joined #sbcl 2014-04-10T07:12:00Z Bike quit (Quit: gets hit by an ice-cream truck) 2014-04-10T07:18:30Z sdemarre joined #sbcl 2014-04-10T07:23:49Z DGASAU joined #sbcl 2014-04-10T07:23:56Z michael_lee quit (Quit: Ex-Chat) 2014-04-10T07:30:13Z angavrilov joined #sbcl 2014-04-10T07:41:05Z edgar-rfx is now known as edgar-rft 2014-04-10T07:49:53Z flip214_ is now known as flip214 2014-04-10T07:49:59Z flip214 quit (Changing host) 2014-04-10T07:49:59Z flip214 joined #sbcl 2014-04-10T07:55:47Z yacks quit (Ping timeout: 252 seconds) 2014-04-10T08:00:09Z yacks joined #sbcl 2014-04-10T08:05:42Z sdemarre quit (Ping timeout: 252 seconds) 2014-04-10T08:51:23Z pranavrc quit (Remote host closed the connection) 2014-04-10T09:16:41Z pranavrc joined #sbcl 2014-04-10T09:32:29Z kludge` quit (Ping timeout: 246 seconds) 2014-04-10T09:34:21Z attila_lendvai quit (Disconnected by services) 2014-04-10T09:34:21Z attila_lendvai1 joined #sbcl 2014-04-10T09:34:21Z attila_lendvai1 quit (Changing host) 2014-04-10T09:34:21Z attila_lendvai1 joined #sbcl 2014-04-10T09:38:02Z kludge` joined #sbcl 2014-04-10T09:47:10Z fridim_ joined #sbcl 2014-04-10T10:14:04Z White_Flame quit (Quit: http://quassel-irc.org - Chat comfortably. Anywhere.) 2014-04-10T10:31:37Z ehaliewicz quit (Ping timeout: 252 seconds) 2014-04-10T10:54:00Z White_Flame joined #sbcl 2014-04-10T11:19:14Z hlavaty: (time (handler-case (sb-ext:with-timeout 2 (prog2 (print :@@@a) (sleep 5) (print :@@@z))) (sb-ext:timeout (c) :timeout))) doesnt time out:-( 2014-04-10T11:19:14Z hlavaty: 2014-04-10T11:19:39Z hlavaty: shouldnt that return :timeout? 2014-04-10T11:20:40Z fe[nl]ix: I see :timeout here 2014-04-10T11:21:13Z hlavaty: weird i'm on x86_64 linux commit 30f049cd 2014-04-10T11:21:39Z hlavaty: fe[nl]ix: thanks for help 2014-04-10T11:22:52Z fe[nl]ix: I'm on f72dea37 2014-04-10T11:23:13Z hlavaty: it works on 1.0.57.66-5783625 2014-04-10T11:24:27Z pjb quit (Ping timeout: 276 seconds) 2014-04-10T11:30:32Z hlavaty: actually i was on 32 bit sbcl 2014-04-10T11:30:53Z hlavaty: i'll try with the latest 2014-04-10T11:32:56Z pjb joined #sbcl 2014-04-10T11:39:38Z hlavaty: ok it works with the HEAD 2014-04-10T11:56:31Z drmeister quit (Remote host closed the connection) 2014-04-10T11:58:57Z attila_lendvai joined #sbcl 2014-04-10T12:00:06Z attila_lendvai1 quit (Ping timeout: 240 seconds) 2014-04-10T12:08:44Z attila_lendvai quit (Remote host closed the connection) 2014-04-10T12:09:03Z attila_lendvai joined #sbcl 2014-04-10T12:09:03Z attila_lendvai quit (Changing host) 2014-04-10T12:09:03Z attila_lendvai joined #sbcl 2014-04-10T12:09:53Z drmeister joined #sbcl 2014-04-10T12:10:58Z sdemarre joined #sbcl 2014-04-10T12:31:35Z drmeister quit (Read error: Connection reset by peer) 2014-04-10T12:38:26Z nyef joined #sbcl 2014-04-10T12:38:54Z nyef: G'morning all. 2014-04-10T12:39:53Z LiamH joined #sbcl 2014-04-10T12:41:21Z drmeister joined #sbcl 2014-04-10T12:44:40Z nyef: "The variable 68793800927 is unbound." 2014-04-10T12:44:45Z nyef: I think I did something wrong just now. 2014-04-10T12:46:07Z nyef: Okay, transposing the declaration and value forms for m-v-b can have odd effects. 2014-04-10T12:46:32Z eudoxia joined #sbcl 2014-04-10T12:46:54Z sdemarre quit (Ping timeout: 240 seconds) 2014-04-10T12:54:43Z pranavrc quit 2014-04-10T12:56:45Z harovali quit (Ping timeout: 276 seconds) 2014-04-10T13:05:38Z segv- quit (Ping timeout: 246 seconds) 2014-04-10T13:05:48Z psilord quit (Quit: Leaving.) 2014-04-10T13:08:27Z segv- joined #sbcl 2014-04-10T13:11:07Z harovali joined #sbcl 2014-04-10T13:17:50Z LiamH quit (Quit: Leaving.) 2014-04-10T13:24:30Z attila_lendvai quit (Ping timeout: 240 seconds) 2014-04-10T13:27:16Z oleo joined #sbcl 2014-04-10T13:31:00Z cmack joined #sbcl 2014-04-10T13:33:41Z reb`` quit (Remote host closed the connection) 2014-04-10T13:42:25Z hargettp joined #sbcl 2014-04-10T13:43:15Z attila_lendvai joined #sbcl 2014-04-10T13:43:15Z attila_lendvai quit (Changing host) 2014-04-10T13:43:15Z attila_lendvai joined #sbcl 2014-04-10T13:50:31Z attila_lendvai quit (Ping timeout: 245 seconds) 2014-04-10T14:00:58Z psilord joined #sbcl 2014-04-10T14:02:32Z hlavaty quit (Remote host closed the connection) 2014-04-10T14:13:53Z drmeiste_ joined #sbcl 2014-04-10T14:16:46Z drmeister quit (Ping timeout: 245 seconds) 2014-04-10T14:17:44Z attila_lendvai joined #sbcl 2014-04-10T14:29:50Z brown joined #sbcl 2014-04-10T14:30:13Z brown is now known as Guest11863 2014-04-10T14:31:35Z Guest11863 is now known as reb 2014-04-10T14:38:04Z drmeiste_ quit (Read error: Connection reset by peer) 2014-04-10T14:38:33Z drmeister joined #sbcl 2014-04-10T14:40:03Z drmeiste_ joined #sbcl 2014-04-10T14:40:09Z fe[nl]ix quit (Remote host closed the connection) 2014-04-10T14:40:09Z Blkt quit (Remote host closed the connection) 2014-04-10T14:40:17Z Blkt joined #sbcl 2014-04-10T14:40:19Z fe[nl]ix joined #sbcl 2014-04-10T14:41:16Z drmeist__ joined #sbcl 2014-04-10T14:43:14Z drmeis___ joined #sbcl 2014-04-10T14:43:21Z drmeister quit (Ping timeout: 276 seconds) 2014-04-10T14:44:06Z drmeiste_ quit (Ping timeout: 240 seconds) 2014-04-10T14:46:36Z drmeist__ quit (Ping timeout: 276 seconds) 2014-04-10T14:53:43Z nyef: Okay, here's what I have for the array type unity thing: http://paste.lisp.org/display/141977 2014-04-10T14:54:18Z nyef: And it does fix the issue that jasom reported. 2014-04-10T14:54:42Z nyef: Krystof, pkhuong: Feedback appreciated when you have time. 2014-04-10T14:57:09Z Krystof: suddenly I remember what I was thinking 2014-04-10T14:57:26Z Krystof: I was thinking "this is the wrong representation for array types" 2014-04-10T14:57:34Z nyef: Heh. 2014-04-10T14:58:22Z Krystof: I haven't read all the detail yet, but this is a good step in the right direction 2014-04-10T14:58:48Z nyef: Let me guess, next is to review and fix the intersection logic? 2014-04-10T14:59:10Z Krystof: possibly, but maybe the representation change would be helpful 2014-04-10T14:59:31Z Krystof: the particular case I am thinking of is when you have (or (array nil) (array bit) (array base-char) (array ...) ...) where the ... enumerates *all* the specialized array types 2014-04-10T14:59:48Z Krystof: that should be type= to (array *), but we'll never capture that in our current representation 2014-04-10T15:00:02Z nyef: Yeah, that's a bit that I was thinking about, and decided to ignore for the moment. 2014-04-10T15:00:22Z Krystof: the solution is I think to have an array type a bit like our current one but with a bit-mask (and associated expressed array element-type vector) for element-types 2014-04-10T15:01:09Z Krystof: this is obviously a total pain :) 2014-04-10T15:01:17Z Krystof: ignoring for now is fine 2014-04-10T15:02:18Z nyef: There's a way to detect the case in unite-array-types-element-types, but it'd suck in terms of both performance and just plain ugly code. 2014-04-10T15:02:40Z Krystof: yeah, let's not do that 2014-04-10T15:03:02Z nyef: The representation improvement might also simplify the compiler-side logic for array element type derivation. 2014-04-10T15:03:05Z Krystof: well except that the unite-array-types-element-types is pairwise 2014-04-10T15:03:08Z Krystof: yes! 2014-04-10T15:03:53Z nyef: On the other hand, might not. 2014-04-10T15:04:56Z Bike joined #sbcl 2014-04-10T15:09:54Z nyef: Oh. You're right, we can't actually upgrade to wild-type with this representation. 2014-04-10T15:09:55Z pkhuong: nyef: intersection is easier than union in this case. 2014-04-10T15:10:59Z nyef: Okay, so apart from adding a note about upgrading to wild-type, this looks good so far? 2014-04-10T15:13:29Z Krystof: I'm not sure about (:array-type-union :dimensional-unity :moves-towards-wild) 2014-04-10T15:14:17Z Krystof: I would have said that (array * (2 *)) and (array * (* 3)) didn't have a simple union 2014-04-10T15:14:51Z Krystof: it's all things with two dimensions, but either the first one has to be of size 2 or the second has to be of length 3 2014-04-10T15:15:04Z Krystof: so an (array * (4 4)) for example is not part of that union 2014-04-10T15:15:27Z nyef: You're right, that was a thinko on my part, compounded by it also being the behavior of the old code. 2014-04-10T15:16:43Z nyef: And there's a knock-on effect on some of the other test cases in terms of legitimacy. 2014-04-10T15:17:29Z Krystof: yeah. I think you can fix the test cases by having one of them be (2 *) and the other be (* *) 2014-04-10T15:17:40Z Krystof: because the rest of them are disunity tests 2014-04-10T15:18:45Z nyef: But they might test the wrong thing if they aren't also updated. 2014-04-10T15:18:55Z Krystof: I mean updating them in that way 2014-04-10T15:19:02Z nyef: Ah, yes. 2014-04-10T15:19:14Z Krystof: so that the dimensions would union if there weren't another blocker 2014-04-10T15:21:32Z pkhuong: yeah, partially specified dimension length don't work 2014-04-10T15:21:42Z attila_lendvai quit (Ping timeout: 240 seconds) 2014-04-10T15:22:04Z pkhuong: I'm pretty sure the only way to do this is to count the # of discrepancies and only try and be clever when the two types differ only on a single quality 2014-04-10T15:23:05Z nyef: Yeah, the issue is that I screwed up and didn't apply that down to the set of dimensions. 2014-04-10T15:23:48Z nyef: And the two types can differ on multiple qualities if one is fully a subset of the other. 2014-04-10T15:35:05Z hargettp quit (Quit: Page closed) 2014-04-10T15:36:06Z fridim_ quit (Ping timeout: 240 seconds) 2014-04-10T15:36:29Z attila_lendvai joined #sbcl 2014-04-10T15:36:29Z attila_lendvai quit (Changing host) 2014-04-10T15:36:29Z attila_lendvai joined #sbcl 2014-04-10T15:45:40Z eudoxia quit (Quit: Lost terminal) 2014-04-10T15:45:52Z eudoxia joined #sbcl 2014-04-10T15:52:22Z fridim_ joined #sbcl 2014-04-10T16:07:23Z pkhuong: if one type is a strict subset of the other. 2014-04-10T16:07:33Z nyef: Yes. 2014-04-10T16:08:46Z sdemarre joined #sbcl 2014-04-10T16:09:35Z nyef: Okay, http://paste.lisp.org/display/141977#1 should work a bit better. 2014-04-10T16:11:19Z sdemarre left #sbcl 2014-04-10T16:24:20Z Krystof: looks better to me 2014-04-10T16:24:43Z Krystof: for what it's worth, I found the return convention of the unite-array-types-* functions hard to wrap my head around 2014-04-10T16:27:04Z nyef: It's a bit tricky, but I'm not really sure what else is viable. 2014-04-10T16:31:46Z nyef: Okay, if nobody has further suggestions before then, I'll update the commit with the diff in annotation 1 and a NEWS entry and push it tonight. 2014-04-10T16:47:52Z asedeno quit (Quit: leaving) 2014-04-10T16:48:05Z asedeno joined #sbcl 2014-04-10T16:56:42Z ASau joined #sbcl 2014-04-10T17:03:51Z angavrilov quit (Remote host closed the connection) 2014-04-10T17:18:34Z attila_lendvai1 joined #sbcl 2014-04-10T17:18:47Z attila_lendvai quit (Disconnected by services) 2014-04-10T17:18:47Z attila_lendvai1 quit (Changing host) 2014-04-10T17:18:47Z attila_lendvai1 joined #sbcl 2014-04-10T17:30:11Z fridim_ quit (Read error: Connection reset by peer) 2014-04-10T17:31:46Z attila_lendvai1 quit (Ping timeout: 245 seconds) 2014-04-10T17:47:12Z attila_lendvai joined #sbcl 2014-04-10T17:55:26Z fiveop joined #sbcl 2014-04-10T18:00:51Z davazp joined #sbcl 2014-04-10T18:11:47Z attila_lendvai quit (Ping timeout: 252 seconds) 2014-04-10T18:12:14Z ltbarcly joined #sbcl 2014-04-10T18:13:28Z ltbarcly_ joined #sbcl 2014-04-10T18:17:07Z ltbarcly quit (Ping timeout: 250 seconds) 2014-04-10T18:30:53Z pkhuong quit (Ping timeout: 255 seconds) 2014-04-10T18:38:08Z pkhuong joined #sbcl 2014-04-10T18:38:39Z eudoxia quit (Quit: leaving) 2014-04-10T18:39:49Z eudoxia joined #sbcl 2014-04-10T18:43:54Z drmeis___ quit (Read error: Connection reset by peer) 2014-04-10T18:44:03Z drmeister joined #sbcl 2014-04-10T18:44:06Z prxq joined #sbcl 2014-04-10T18:58:12Z eudoxia quit (Quit: leaving) 2014-04-10T18:58:26Z eudoxia joined #sbcl 2014-04-10T19:03:17Z irsol quit (Ping timeout: 255 seconds) 2014-04-10T19:06:46Z eudoxia quit (Quit: leaving) 2014-04-10T19:07:14Z eudoxia joined #sbcl 2014-04-10T19:07:14Z eudoxia quit (Client Quit) 2014-04-10T19:16:39Z irsol joined #sbcl 2014-04-10T19:27:45Z davazp quit (Read error: Connection reset by peer) 2014-04-10T19:52:06Z ltbarcly_ quit (Ping timeout: 240 seconds) 2014-04-10T19:53:54Z ltbarcly joined #sbcl 2014-04-10T20:06:22Z ltbarcly quit (Ping timeout: 258 seconds) 2014-04-10T20:15:06Z oleo quit (Ping timeout: 240 seconds) 2014-04-10T20:15:34Z oleo joined #sbcl 2014-04-10T20:50:37Z fiveop quit 2014-04-10T21:01:41Z ehaliewicz joined #sbcl 2014-04-10T21:10:21Z scymtym_ joined #sbcl 2014-04-10T21:17:42Z ltbarcly joined #sbcl 2014-04-10T21:20:40Z fikusz quit (Quit: Leaving) 2014-04-10T21:22:54Z fikusz joined #sbcl 2014-04-10T21:23:59Z drmeiste_ joined #sbcl 2014-04-10T21:27:18Z drmeister quit (Ping timeout: 240 seconds) 2014-04-10T21:32:42Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-10T21:39:33Z ltbarcly joined #sbcl 2014-04-10T21:41:20Z scymtym_: SB-REGALLOC::OFFSET-CONFLICTS-IN-SB fails to handle the ALWAYS-LIVE slot in annotation 1 of http://paste.lisp.org/display/141980 (i had to leave the machine after making the paste; will try to reproduce later) 2014-04-10T21:41:34Z scymtym_: does the ALWAYS-LIVE vector look ok? 2014-04-10T21:44:40Z scymtym_: the comment in defstruct finite-sb suggests otherwise 2014-04-10T22:00:32Z drmeiste_ quit (Remote host closed the connection) 2014-04-10T22:01:13Z pkhuong: The stack might be special 2014-04-10T22:02:39Z scymtym_: maybe its a miscompilation or debugger problem, but OFFSET-CONFLICTS-IN-SB doesn't seem to handle empty vectors or integer entries in the ALWAYS-LIVE vector 2014-04-10T22:08:01Z psilord quit (Quit: Leaving.) 2014-04-10T22:09:12Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-10T22:15:23Z drmeister joined #sbcl 2014-04-10T22:18:32Z ccl-logbot joined #sbcl 2014-04-10T22:18:32Z 2014-04-10T22:18:32Z names: ccl-logbot eudoxia drmeister fikusz scymtym_ ehaliewicz oleo irsol prxq pkhuong ASau asedeno Bike fe[nl]ix Blkt reb cmack harovali segv- nyef pjb White_Flame kludge` yacks DGASAU kanru |3b| slyrus christoph_debian flip214 edgar-rft minion specbot redline6561 tmh_ jaimef joshe antoszka @Krystof luis loke scymtym pchrist xymox nicdev ams faheem Posterdati ferada ivan`` milosn afleck loke_ hzp Munksgaard leoc yauz jsnell_ brucem mood 2014-04-10T22:40:27Z drmeister quit (Ping timeout: 276 seconds) 2014-04-10T22:46:02Z yacks quit (Ping timeout: 252 seconds) 2014-04-10T22:47:28Z yacks joined #sbcl 2014-04-10T23:43:09Z eudoxia quit (Quit: leaving) 2014-04-10T23:51:53Z psilord joined #sbcl 2014-04-11T00:04:16Z drmeister joined #sbcl 2014-04-11T00:17:09Z ehaliewi` joined #sbcl 2014-04-11T00:17:39Z ehaliewicz quit (Remote host closed the connection) 2014-04-11T00:35:16Z drmeister quit (Remote host closed the connection) 2014-04-11T00:46:29Z nyef: I'm going to commit the unity fix in a few minutes, unless someone can come up with a reason why not. 2014-04-11T00:58:54Z harovali quit (Ping timeout: 240 seconds) 2014-04-11T01:00:38Z ltbarcly joined #sbcl 2014-04-11T01:00:53Z nyef: And done. 2014-04-11T01:03:54Z eudoxia joined #sbcl 2014-04-11T01:05:59Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-11T01:15:21Z harovali joined #sbcl 2014-04-11T01:25:34Z eudoxia quit (Quit: leaving) 2014-04-11T01:34:15Z prxq_ joined #sbcl 2014-04-11T01:37:06Z prxq quit (Ping timeout: 240 seconds) 2014-04-11T01:42:56Z drmeister joined #sbcl 2014-04-11T01:45:06Z scymtym_ quit (Ping timeout: 240 seconds) 2014-04-11T02:07:17Z ltbarcly joined #sbcl 2014-04-11T02:38:54Z christoph_debian quit (Ping timeout: 240 seconds) 2014-04-11T02:52:24Z christoph_debian joined #sbcl 2014-04-11T03:15:51Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-11T03:30:56Z attila_lendvai joined #sbcl 2014-04-11T03:46:51Z nyef quit (Quit: G'night all.) 2014-04-11T04:00:43Z ltbarcly joined #sbcl 2014-04-11T05:02:44Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-11T05:08:15Z oleo quit (Remote host closed the connection) 2014-04-11T05:22:36Z michael_lee joined #sbcl 2014-04-11T05:38:07Z pranavrc joined #sbcl 2014-04-11T05:38:07Z pranavrc quit (Changing host) 2014-04-11T05:38:07Z pranavrc joined #sbcl 2014-04-11T05:42:48Z michael_lee quit (Remote host closed the connection) 2014-04-11T06:01:44Z ASau quit (Remote host closed the connection) 2014-04-11T06:42:57Z angavrilov joined #sbcl 2014-04-11T06:51:55Z loke_ quit (Remote host closed the connection) 2014-04-11T08:01:46Z ehaliewi` quit (Ping timeout: 245 seconds) 2014-04-11T08:25:51Z jdz joined #sbcl 2014-04-11T08:44:33Z echo-area joined #sbcl 2014-04-11T09:00:28Z pranavrc quit (Read error: Connection reset by peer) 2014-04-11T09:01:05Z pranavrc joined #sbcl 2014-04-11T09:01:05Z pranavrc quit (Changing host) 2014-04-11T09:01:05Z pranavrc joined #sbcl 2014-04-11T09:38:20Z segv- quit (Ping timeout: 252 seconds) 2014-04-11T10:11:14Z christoph_debian quit (Ping timeout: 258 seconds) 2014-04-11T10:16:20Z christoph_debian joined #sbcl 2014-04-11T10:22:24Z davazp joined #sbcl 2014-04-11T10:34:18Z christoph_debian quit (Ping timeout: 240 seconds) 2014-04-11T10:35:34Z christoph_debian joined #sbcl 2014-04-11T10:40:29Z christoph_debian quit (Ping timeout: 252 seconds) 2014-04-11T10:47:52Z davazp quit (Remote host closed the connection) 2014-04-11T10:50:20Z christoph_debian joined #sbcl 2014-04-11T10:56:33Z sdemarre joined #sbcl 2014-04-11T11:01:08Z echo-are` joined #sbcl 2014-04-11T11:02:11Z echo-area quit (Ping timeout: 245 seconds) 2014-04-11T11:13:39Z jdz quit (Quit: Leaving) 2014-04-11T11:24:30Z sdemarre quit (Ping timeout: 240 seconds) 2014-04-11T11:35:50Z segv- joined #sbcl 2014-04-11T11:56:40Z fridim_ joined #sbcl 2014-04-11T12:09:42Z fridim_ quit (Read error: Connection reset by peer) 2014-04-11T12:10:16Z eudoxia joined #sbcl 2014-04-11T12:11:01Z cmack quit (Remote host closed the connection) 2014-04-11T12:13:49Z drmeister quit (Remote host closed the connection) 2014-04-11T12:24:16Z brucem_ joined #sbcl 2014-04-11T12:24:20Z jaimef quit (Ping timeout: 240 seconds) 2014-04-11T12:24:21Z brucem quit (Quit: ZNC - http://znc.sourceforge.net) 2014-04-11T12:25:25Z attila_lendvai quit (Disconnected by services) 2014-04-11T12:25:25Z attila_lendvai1 joined #sbcl 2014-04-11T12:25:25Z attila_lendvai1 quit (Changing host) 2014-04-11T12:25:25Z attila_lendvai1 joined #sbcl 2014-04-11T12:28:03Z jaimef joined #sbcl 2014-04-11T12:29:28Z jdz joined #sbcl 2014-04-11T12:38:10Z oleo joined #sbcl 2014-04-11T12:38:45Z jdz_ joined #sbcl 2014-04-11T12:42:56Z drmeister joined #sbcl 2014-04-11T12:45:43Z christoph_debian quit (Ping timeout: 252 seconds) 2014-04-11T12:51:02Z jdz quit (Quit: Leaving) 2014-04-11T12:51:54Z momo-reina joined #sbcl 2014-04-11T13:00:30Z harovali quit (Ping timeout: 240 seconds) 2014-04-11T13:04:33Z psilord quit (Quit: Leaving.) 2014-04-11T13:05:49Z nyef joined #sbcl 2014-04-11T13:06:45Z nyef: G'morning all. 2014-04-11T13:16:25Z harovali joined #sbcl 2014-04-11T13:16:46Z harovali quit (Max SendQ exceeded) 2014-04-11T13:17:31Z drmeister quit (Read error: Connection reset by peer) 2014-04-11T13:17:50Z pranavrc quit 2014-04-11T13:18:29Z harovali joined #sbcl 2014-04-11T13:18:49Z harovali quit (Max SendQ exceeded) 2014-04-11T13:20:38Z harovali joined #sbcl 2014-04-11T13:21:08Z flip214: what's the status of SBCL on ARM? Any progress? 2014-04-11T13:22:53Z jdz_ quit (Remote host closed the connection) 2014-04-11T13:24:43Z nyef: Not on my side, I'm afraid. 2014-04-11T13:25:12Z attila_lendvai1 quit (Quit: Leaving.) 2014-04-11T13:28:12Z jdz joined #sbcl 2014-04-11T13:30:16Z flip214: nyef: thanks for your answer ... although I didn't specifically asked you. I'd hoped that somebody else speaks up and says "everything up and running, including all of quicklisp, in 100MB on my raspberry" ;) 2014-04-11T13:30:34Z nyef: Oh, wouldn't that be nice? 2014-04-11T13:31:11Z nyef: More seriously, it's probably still bogged down with floating-point and interrupt-handling stuff. 2014-04-11T13:36:42Z brucem_ is now known as brucem 2014-04-11T14:03:34Z momo-reina quit (Remote host closed the connection) 2014-04-11T14:05:46Z psilord joined #sbcl 2014-04-11T14:08:21Z jdz quit (Ping timeout: 250 seconds) 2014-04-11T14:15:10Z reb: nyef/flip214: If someone wanted to hack on the ARM port, what repos/branches should he take a look at? 2014-04-11T14:16:51Z flip214: reb: google finds https://github.com/fiveop/sbcl-arm 2014-04-11T14:17:21Z reb: ah, the google ... thanks! 2014-04-11T14:17:40Z nyef: Yeah, that's fiveop's branch, he's been focussing on the floaty bits. 2014-04-11T14:27:28Z sdemarre joined #sbcl 2014-04-11T14:46:49Z drmeister joined #sbcl 2014-04-11T14:48:03Z specbot quit (Disconnected by services) 2014-04-11T14:48:09Z minion quit (Disconnected by services) 2014-04-11T14:48:13Z minion joined #sbcl 2014-04-11T14:48:49Z specbot joined #sbcl 2014-04-11T15:08:18Z sdemarre quit (Ping timeout: 240 seconds) 2014-04-11T16:47:20Z jdz joined #sbcl 2014-04-11T16:56:28Z ltbarcly joined #sbcl 2014-04-11T17:03:52Z christoph_debian joined #sbcl 2014-04-11T17:10:16Z jdz quit (Read error: No route to host) 2014-04-11T17:10:29Z jdz joined #sbcl 2014-04-11T17:11:49Z jdz quit (Remote host closed the connection) 2014-04-11T17:12:11Z christoph_debian quit (Ping timeout: 245 seconds) 2014-04-11T17:22:36Z sdemarre joined #sbcl 2014-04-11T17:34:18Z nyef: Are we expecting some level of print/read consistency for type objects? Specifically, if we have an ARRAY-TYPE object, are we expecting to be able to extract a specifier for it (unparse it?), then invoke SPECIFIER-TYPE and get something back that is TYPE= to the original? 2014-04-11T17:39:16Z nyef: Because if we are, we're not getting it. (-: 2014-04-11T18:00:09Z jdz joined #sbcl 2014-04-11T18:00:52Z ASau joined #sbcl 2014-04-11T18:03:54Z psilord quit (Ping timeout: 240 seconds) 2014-04-11T18:04:38Z psilord joined #sbcl 2014-04-11T18:27:55Z sdemarre quit (Ping timeout: 250 seconds) 2014-04-11T18:28:10Z christoph_debian joined #sbcl 2014-04-11T18:31:12Z jdz quit (Quit: Leaving...) 2014-04-11T18:42:18Z irsol quit (Ping timeout: 240 seconds) 2014-04-11T18:44:19Z christoph_debian quit (Ping timeout: 252 seconds) 2014-04-11T18:44:55Z eudoxia quit (Quit: Lost terminal) 2014-04-11T18:55:16Z irsol joined #sbcl 2014-04-11T19:09:17Z jaimef quit (Excess Flood) 2014-04-11T19:13:36Z christoph_debian joined #sbcl 2014-04-11T19:14:42Z Krystof: nyef: not completely but mostly 2014-04-11T19:14:53Z Krystof: some things (e.g. array complexity) are hard to represent 2014-04-11T19:15:15Z Krystof: somewhere I have a branch that actually represent T complexity correctly [as (and (array x) (not (simple-array x))) ] 2014-04-11T19:15:41Z jaimef joined #sbcl 2014-04-11T19:19:34Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-11T19:24:37Z nyef: It's currently represented as (and (array x) (not simple-array)), which makes sense to me. 2014-04-11T19:26:18Z nyef: I have a case where the UAET of the element type is not the specialized type, which means that when you round-trip it you get a different specialized type. 2014-04-11T19:32:07Z nyef: Really, the only thing that can be done there is to represent it as an intersection type... except that I'm not even sure that that would work for all cases. 2014-04-11T19:33:14Z Krystof: oh, right, yes, well, that's doomed 2014-04-11T19:33:22Z Krystof: but that case doesn't get round-tripped very often 2014-04-11T19:40:42Z nyef: I think I can make it work for simple numeric ranges, MEMBER types, and so on. 2014-04-11T19:41:05Z ltbarcly joined #sbcl 2014-04-11T19:41:31Z nyef: But it basically requires introducing elements to force the desired specialized element type and using intersection to remove them because they aren't actually part of the type. 2014-04-11T19:41:56Z Krystof: I think this is one of the cases that my planned rewrite of array types was going to deal with 2014-04-11T19:43:28Z angavrilov quit (Remote host closed the connection) 2014-04-11T19:46:03Z echo-are` quit (Read error: Connection reset by peer) 2014-04-11T19:48:27Z jaimef quit (Excess Flood) 2014-04-11T19:49:25Z leuler joined #sbcl 2014-04-11T19:50:56Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-11T19:51:41Z jaimef joined #sbcl 2014-04-11T20:14:35Z attila_lendvai joined #sbcl 2014-04-11T20:14:35Z attila_lendvai quit (Changing host) 2014-04-11T20:14:35Z attila_lendvai joined #sbcl 2014-04-11T20:25:41Z ltbarcly joined #sbcl 2014-04-11T20:26:05Z attila_lendvai quit (Quit: Leaving.) 2014-04-11T20:42:00Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-11T20:43:06Z fridim_ joined #sbcl 2014-04-11T20:58:26Z scymtym_ joined #sbcl 2014-04-11T21:07:36Z psilord quit (Quit: Leaving.) 2014-04-11T21:15:42Z christoph_debian quit (Ping timeout: 240 seconds) 2014-04-11T21:26:15Z drmeister quit (Remote host closed the connection) 2014-04-11T21:29:04Z christoph_debian joined #sbcl 2014-04-11T21:37:06Z drmeister joined #sbcl 2014-04-11T21:39:13Z oleo quit (Ping timeout: 252 seconds) 2014-04-11T21:40:06Z fridim_ quit (Ping timeout: 245 seconds) 2014-04-11T21:41:34Z oleo joined #sbcl 2014-04-11T22:06:23Z psilord joined #sbcl 2014-04-11T22:12:21Z leuler quit (Quit: ERC Version 5.1.2 $Revision: 1.796.2.6 $ (IRC client for Emacs)) 2014-04-11T22:13:49Z drmeister quit (Read error: Connection reset by peer) 2014-04-11T22:25:35Z drmeister joined #sbcl 2014-04-11T22:29:27Z eudoxia joined #sbcl 2014-04-11T22:39:07Z drmeister quit (Read error: Connection reset by peer) 2014-04-11T22:52:18Z drmeister joined #sbcl 2014-04-11T23:16:56Z nyef: ... In ppc-bsd-os.c, there is special handling for the case of LISP_FEATURE_C_STACK_IS_CONTROL_STACK. WTF? 2014-04-11T23:20:47Z ltbarcly joined #sbcl 2014-04-11T23:37:17Z davazp joined #sbcl 2014-04-11T23:41:38Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-12T00:15:28Z ltbarcly joined #sbcl 2014-04-12T00:27:46Z pkhuong: heh 2014-04-12T00:41:24Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-12T01:03:42Z harovali quit (Ping timeout: 240 seconds) 2014-04-12T01:09:55Z ltbarcly joined #sbcl 2014-04-12T01:18:52Z harovali joined #sbcl 2014-04-12T01:19:11Z harovali quit (Max SendQ exceeded) 2014-04-12T01:19:35Z nyef sighs. 2014-04-12T01:20:17Z harovali joined #sbcl 2014-04-12T01:20:24Z nyef: So, it looks like the ARM port isn't actually thoroughly nailed by the signal handling bit, it's trying to use #'< without having two-arg-< defined. 2014-04-12T01:20:36Z harovali quit (Max SendQ exceeded) 2014-04-12T01:20:50Z nyef: Or generic-<, for that matter. 2014-04-12T01:22:23Z harovali joined #sbcl 2014-04-12T01:22:42Z harovali quit (Max SendQ exceeded) 2014-04-12T01:23:31Z harovali joined #sbcl 2014-04-12T01:34:49Z prxq__ joined #sbcl 2014-04-12T01:37:55Z prxq_ quit (Ping timeout: 252 seconds) 2014-04-12T01:49:58Z eudoxia quit (Quit: leaving) 2014-04-12T02:23:49Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-12T02:33:22Z davazp quit (Remote host closed the connection) 2014-04-12T02:39:16Z christoph_debian quit (Ping timeout: 245 seconds) 2014-04-12T02:39:47Z ltbarcly joined #sbcl 2014-04-12T02:41:17Z loke quit (Remote host closed the connection) 2014-04-12T02:51:54Z nyef quit (Quit: G'night all.) 2014-04-12T02:52:57Z christoph_debian joined #sbcl 2014-04-12T02:58:14Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-12T03:23:56Z ltbarcly joined #sbcl 2014-04-12T03:37:04Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-12T03:39:14Z yacks quit (Quit: Leaving) 2014-04-12T04:24:34Z yacks joined #sbcl 2014-04-12T04:52:31Z drmeister quit (Remote host closed the connection) 2014-04-12T05:09:18Z scymtym_ quit (Ping timeout: 240 seconds) 2014-04-12T05:13:34Z _8hzp joined #sbcl 2014-04-12T05:15:57Z ASau quit (Remote host closed the connection) 2014-04-12T05:16:42Z ASau joined #sbcl 2014-04-12T05:17:07Z hzp quit (Ping timeout: 252 seconds) 2014-04-12T05:20:24Z attila_lendvai joined #sbcl 2014-04-12T05:24:30Z ASau quit (Ping timeout: 276 seconds) 2014-04-12T06:07:48Z angavrilov joined #sbcl 2014-04-12T06:26:54Z yacks quit (Ping timeout: 276 seconds) 2014-04-12T06:41:24Z drmeister joined #sbcl 2014-04-12T06:45:56Z drmeister quit (Ping timeout: 245 seconds) 2014-04-12T06:46:17Z fridim_ joined #sbcl 2014-04-12T06:58:11Z sdemarre joined #sbcl 2014-04-12T07:01:32Z prxq__ is now known as prxq 2014-04-12T07:48:17Z sdemarre quit (Ping timeout: 250 seconds) 2014-04-12T07:54:02Z yacks joined #sbcl 2014-04-12T08:25:39Z fridim_ quit (Ping timeout: 252 seconds) 2014-04-12T08:30:08Z drmeister joined #sbcl 2014-04-12T08:34:57Z drmeister quit (Ping timeout: 276 seconds) 2014-04-12T08:40:18Z yacks quit (Ping timeout: 240 seconds) 2014-04-12T08:49:50Z yacks joined #sbcl 2014-04-12T09:34:06Z kludge` quit (Ping timeout: 276 seconds) 2014-04-12T09:39:06Z kludge` joined #sbcl 2014-04-12T09:42:54Z yacks quit (Ping timeout: 240 seconds) 2014-04-12T10:12:15Z oleo quit (Read error: Connection reset by peer) 2014-04-12T10:13:25Z oleo joined #sbcl 2014-04-12T10:16:45Z stassats joined #sbcl 2014-04-12T10:19:06Z drmeister joined #sbcl 2014-04-12T10:23:18Z drmeister quit (Ping timeout: 240 seconds) 2014-04-12T10:29:50Z yacks joined #sbcl 2014-04-12T10:52:08Z ASau joined #sbcl 2014-04-12T12:07:43Z drmeister joined #sbcl 2014-04-12T12:12:10Z drmeister quit (Ping timeout: 258 seconds) 2014-04-12T12:12:44Z fiveop joined #sbcl 2014-04-12T12:15:26Z drmeister joined #sbcl 2014-04-12T12:37:54Z attila_lendvai quit (Quit: Leaving.) 2014-04-12T12:58:23Z nyef joined #sbcl 2014-04-12T12:58:32Z nyef: G'morning all. 2014-04-12T12:59:48Z fiveop: hi 2014-04-12T13:04:43Z harovali quit (Read error: Connection reset by peer) 2014-04-12T13:07:00Z harovali joined #sbcl 2014-04-12T13:07:19Z harovali quit (Max SendQ exceeded) 2014-04-12T13:09:02Z harovali joined #sbcl 2014-04-12T13:09:21Z harovali quit (Max SendQ exceeded) 2014-04-12T13:10:15Z harovali joined #sbcl 2014-04-12T13:47:26Z nyef: ... Why is there no consistency with respect to arithmetic static functions across backends? 2014-04-12T13:47:45Z stassats: because backends were written in different eras by different people? 2014-04-12T13:48:30Z nyef: MIPS, PPC, and SPARC have generic-<=. ALPHA, HPPA, MIPS, and PPC have two-arg-<=, and I fail to see anything for x86 or x86-64. 2014-04-12T13:49:39Z stassats: generic-<= is an assembly routine and two-arg-<= is an ordinary function, right? 2014-04-12T13:50:16Z nyef: two-arg-<= is a static-fun. 2014-04-12T13:50:59Z stassats: but the function itself is an ordinary function? 2014-04-12T13:51:11Z stassats: i view static-funs as just a different way of calling ordinary functions 2014-04-12T13:51:57Z nyef: Ah. Okay, so it's not actually defined AFAICT. 2014-04-12T13:52:09Z nyef: generic-< is defined in code/numbers. 2014-04-12T13:52:28Z nyef: Err... Sorry, two-arg-< is defined in code/numbers. 2014-04-12T13:53:03Z stassats: generic-< is an assembly routines that does the job for fixnums, and otherwise calls two-arg-<, IIRC 2014-04-12T13:53:13Z nyef: There's probably a transform somewhere that I have yet to find that converts (two-arg-<= x y) to two-arg-> y x). 2014-04-12T13:53:35Z stassats: how well does that work with floats? 2014-04-12T13:54:00Z nyef: I really don't care about floats right now. 2014-04-12T13:54:16Z stassats: on x86, <= calls GENERIC-< and GENERIC-= 2014-04-12T13:54:33Z nyef: That'd do it as well. 2014-04-12T13:54:39Z nyef: Which means that it could easily be a macro instead. 2014-04-12T13:55:03Z nyef: Okay, I have a rough plan of attack now. Thank you. 2014-04-12T13:55:37Z stassats: i can see generic-<= being more efficient 2014-04-12T13:56:18Z stassats: but i'm not really convinced that generic-anything-arithmetic is a good idea, at least on modern x86-64 2014-04-12T13:56:39Z nyef: Which is why we have DECLARE TYPE. 2014-04-12T14:00:25Z stassats: i can see assembly routines for things which do not clobber registers on the fast path, like bignum allocation 2014-04-12T14:00:41Z stassats: but generic-< are the same as a full-call 2014-04-12T14:00:56Z stassats: generic* 2014-04-12T14:03:55Z nyef: Mostly. You still manage to shave a few instructions off the path if you're using fixnums after all. 2014-04-12T14:05:52Z stassats: i haven't measured the effect 2014-04-12T14:06:56Z nyef: So, +, -, *, /, <, >, =, and, ior, xor, eqv, lcm, and gcd. 2014-04-12T14:07:29Z stassats: maybe it would be better to automate it, since there already all the fixnum vops for those operations 2014-04-12T14:07:57Z nyef: ... Except that x86 only defines /, gcd, lcm, and, ior, and xor. Hrm. 2014-04-12T14:08:00Z stassats: so, it could be macroized to a type-check + calling an oprirate vop and falling back to a generic-generic function 2014-04-12T14:14:53Z davazp joined #sbcl 2014-04-12T14:45:09Z antoszka quit (Ping timeout: 250 seconds) 2014-04-12T14:46:03Z antoszka joined #sbcl 2014-04-12T15:06:23Z fridim_ joined #sbcl 2014-04-12T15:09:51Z eudoxia joined #sbcl 2014-04-12T15:09:54Z eudoxia quit (Client Quit) 2014-04-12T15:13:03Z eudoxia joined #sbcl 2014-04-12T15:38:01Z attila_lendvai joined #sbcl 2014-04-12T15:38:01Z attila_lendvai quit (Changing host) 2014-04-12T15:38:01Z attila_lendvai joined #sbcl 2014-04-12T15:53:13Z eudoxia quit (Quit: Lost terminal) 2014-04-12T16:32:29Z davazp quit (Remote host closed the connection) 2014-04-12T17:27:54Z fridim_ quit (Ping timeout: 240 seconds) 2014-04-12T17:28:31Z stassats quit (Ping timeout: 250 seconds) 2014-04-12T18:12:37Z nyef: Hrm. In a sigtrap handler, during cold-init, there's an internal error, it's just printed the error argument, then it somehow enters SIGSEGV during vfprintf()? 2014-04-12T18:23:24Z fiveop quit 2014-04-12T18:25:53Z nyef: Okay, so lose() dies after outputting the "fatal error encountered" message, but corruption_warning_and_maybe_lose() gets all the way to LDB. 2014-04-12T19:46:25Z crixus joined #sbcl 2014-04-12T19:59:16Z stassats joined #sbcl 2014-04-12T20:46:58Z angavrilov quit (Remote host closed the connection) 2014-04-12T21:14:53Z stassats quit (Ping timeout: 240 seconds) 2014-04-12T21:28:08Z nyef: ... The hell? If I comment out the first fprintf() and the vfrpintf() in print_message(), I get to LDB without issue. I enable either of them, and it's SIGSEGV city. 2014-04-12T21:28:29Z nyef: But it only dies that way from the SIGTRAP handler. 2014-04-12T21:28:33Z Posterdati quit (Read error: Connection reset by peer) 2014-04-12T21:29:21Z nyef: With the subsequent SIGSEGV, it prints just fine and enters LDB. 2014-04-12T21:29:56Z Posterdati joined #sbcl 2014-04-12T22:14:38Z attila_lendvai quit (Quit: Leaving.) 2014-04-12T22:23:43Z pkhuong: nyef: strange reentrancy badness? 2014-04-12T22:26:20Z nyef: I have no idea. 2014-04-12T22:26:39Z nyef: I think I'm going to give up and move on to hacking on something else. 2014-04-12T23:11:26Z davazp joined #sbcl 2014-04-12T23:16:29Z oleo quit (Ping timeout: 258 seconds) 2014-04-12T23:16:52Z oleo joined #sbcl 2014-04-13T01:10:44Z harovali quit (Ping timeout: 276 seconds) 2014-04-13T01:25:16Z harovali joined #sbcl 2014-04-13T01:35:19Z prxq_ joined #sbcl 2014-04-13T01:38:11Z prxq quit (Ping timeout: 252 seconds) 2014-04-13T02:23:01Z crixus quit (Ping timeout: 245 seconds) 2014-04-13T02:39:25Z christoph_debian quit (Ping timeout: 252 seconds) 2014-04-13T02:44:43Z brucem quit (Changing host) 2014-04-13T02:44:43Z brucem joined #sbcl 2014-04-13T02:52:51Z christoph_debian joined #sbcl 2014-04-13T03:06:17Z nyef quit (Quit: G'night all.) 2014-04-13T03:35:11Z pjb quit (*.net *.split) 2014-04-13T03:35:11Z White_Flame quit (*.net *.split) 2014-04-13T03:35:12Z edgar-rft quit (*.net *.split) 2014-04-13T03:35:13Z davazp quit (*.net *.split) 2014-04-13T03:35:15Z segv- quit (*.net *.split) 2014-04-13T03:35:15Z pkhuong quit (*.net *.split) 2014-04-13T03:35:17Z redline6561 quit (*.net *.split) 2014-04-13T03:35:20Z _8hzp quit (*.net *.split) 2014-04-13T03:35:20Z brucem quit (*.net *.split) 2014-04-13T03:35:21Z kanru quit (*.net *.split) 2014-04-13T03:35:23Z Krystof quit (*.net *.split) 2014-04-13T03:35:25Z Posterdati quit (*.net *.split) 2014-04-13T03:35:26Z kludge` quit (*.net *.split) 2014-04-13T03:35:27Z minion quit (*.net *.split) 2014-04-13T03:35:32Z scymtym quit (*.net *.split) 2014-04-13T03:35:35Z faheem quit (*.net *.split) 2014-04-13T03:35:37Z luis quit (*.net *.split) 2014-04-13T03:35:38Z xymox quit (*.net *.split) 2014-04-13T03:35:38Z ams quit (*.net *.split) 2014-04-13T03:35:41Z christoph_debian quit (*.net *.split) 2014-04-13T03:35:42Z oleo quit (*.net *.split) 2014-04-13T03:35:43Z jaimef quit (*.net *.split) 2014-04-13T03:35:43Z irsol quit (*.net *.split) 2014-04-13T03:35:44Z specbot quit (*.net *.split) 2014-04-13T03:35:45Z fikusz quit (*.net *.split) 2014-04-13T03:35:50Z prxq_ quit (*.net *.split) 2014-04-13T03:35:50Z harovali quit (*.net *.split) 2014-04-13T03:35:51Z drmeister quit (*.net *.split) 2014-04-13T03:35:51Z ASau quit (*.net *.split) 2014-04-13T03:35:53Z asedeno quit (*.net *.split) 2014-04-13T03:35:53Z Bike quit (*.net *.split) 2014-04-13T03:35:54Z antoszka quit (*.net *.split) 2014-04-13T03:35:56Z reb quit (*.net *.split) 2014-04-13T03:35:57Z DGASAU quit (*.net *.split) 2014-04-13T03:35:57Z |3b| quit (*.net *.split) 2014-04-13T03:35:57Z slyrus quit (*.net *.split) 2014-04-13T03:35:58Z tmh_ quit (*.net *.split) 2014-04-13T03:36:01Z nicdev quit (*.net *.split) 2014-04-13T03:36:04Z ferada quit (*.net *.split) 2014-04-13T03:36:04Z ivan`` quit (*.net *.split) 2014-04-13T03:36:05Z milosn quit (*.net *.split) 2014-04-13T03:36:06Z afleck quit (*.net *.split) 2014-04-13T03:36:08Z leoc quit (*.net *.split) 2014-04-13T03:36:09Z fe[nl]ix quit (*.net *.split) 2014-04-13T03:36:10Z Blkt quit (*.net *.split) 2014-04-13T03:36:10Z joshe quit (*.net *.split) 2014-04-13T03:36:11Z pchrist quit (*.net *.split) 2014-04-13T03:36:13Z Munksgaard quit (*.net *.split) 2014-04-13T03:36:17Z yauz quit (*.net *.split) 2014-04-13T03:36:17Z jsnell_ quit (*.net *.split) 2014-04-13T05:13:43Z edgar-rft joined #sbcl 2014-04-13T05:13:43Z White_Flame joined #sbcl 2014-04-13T05:13:43Z pjb joined #sbcl 2014-04-13T05:13:43Z tmh_ joined #sbcl 2014-04-13T05:13:43Z slyrus joined #sbcl 2014-04-13T05:13:43Z |3b| joined #sbcl 2014-04-13T05:13:43Z DGASAU joined #sbcl 2014-04-13T05:13:43Z reb joined #sbcl 2014-04-13T05:13:43Z antoszka joined #sbcl 2014-04-13T05:13:43Z prxq_ joined #sbcl 2014-04-13T05:13:43Z harovali joined #sbcl 2014-04-13T05:13:43Z drmeister joined #sbcl 2014-04-13T05:13:43Z ASau joined #sbcl 2014-04-13T05:13:43Z asedeno joined #sbcl 2014-04-13T05:13:43Z Bike joined #sbcl 2014-04-13T05:15:52Z faheem joined #sbcl 2014-04-13T05:15:57Z yauz joined #sbcl 2014-04-13T05:15:57Z jsnell_ joined #sbcl 2014-04-13T05:16:11Z segv- joined #sbcl 2014-04-13T05:16:11Z pkhuong joined #sbcl 2014-04-13T05:16:11Z redline6561 joined #sbcl 2014-04-13T05:16:14Z christoph_debian joined #sbcl 2014-04-13T05:16:14Z oleo joined #sbcl 2014-04-13T05:16:14Z jaimef joined #sbcl 2014-04-13T05:16:14Z irsol joined #sbcl 2014-04-13T05:16:14Z specbot joined #sbcl 2014-04-13T05:16:14Z fikusz joined #sbcl 2014-04-13T05:16:20Z milosn joined #sbcl 2014-04-13T05:16:20Z afleck joined #sbcl 2014-04-13T05:16:20Z leoc joined #sbcl 2014-04-13T05:17:29Z pchrist joined #sbcl 2014-04-13T05:17:29Z Posterdati joined #sbcl 2014-04-13T05:17:34Z fe[nl]ix joined #sbcl 2014-04-13T05:17:34Z Blkt joined #sbcl 2014-04-13T05:17:34Z joshe joined #sbcl 2014-04-13T05:17:34Z Munksgaard joined #sbcl 2014-04-13T05:17:46Z kludge` joined #sbcl 2014-04-13T05:17:46Z minion joined #sbcl 2014-04-13T05:17:46Z scymtym joined #sbcl 2014-04-13T05:17:57Z _8hzp joined #sbcl 2014-04-13T05:17:57Z brucem joined #sbcl 2014-04-13T05:17:57Z kanru joined #sbcl 2014-04-13T05:17:57Z Krystof joined #sbcl 2014-04-13T05:17:57Z kornbluth.freenode.net has set mode +o Krystof 2014-04-13T05:18:00Z xymox joined #sbcl 2014-04-13T05:18:00Z ams joined #sbcl 2014-04-13T05:25:02Z xymox quit (Read error: Connection reset by peer) 2014-04-13T05:25:38Z xymox joined #sbcl 2014-04-13T05:31:05Z ferada joined #sbcl 2014-04-13T05:54:26Z ivan`` joined #sbcl 2014-04-13T05:56:10Z crixus joined #sbcl 2014-04-13T06:26:44Z angavrilov joined #sbcl 2014-04-13T06:28:31Z crixxus joined #sbcl 2014-04-13T06:31:34Z crixus quit (Ping timeout: 258 seconds) 2014-04-13T06:34:49Z drmeister quit (Remote host closed the connection) 2014-04-13T06:37:24Z attila_lendvai joined #sbcl 2014-04-13T07:18:56Z attila_lendvai is now known as Guest24347 2014-04-13T07:24:32Z stassats joined #sbcl 2014-04-13T07:30:34Z xymox quit (Changing host) 2014-04-13T07:30:34Z xymox joined #sbcl 2014-04-13T07:47:24Z stassats: symbolp specifically checks for NIL, but (eql (sb-sys:sap-ref-word (sb-sys:int-sap 537919511) (- sb-vm:lowtag-mask)) sb-vm:symbol-header-widetag) => T 2014-04-13T07:47:44Z stassats: it may be efficient not to touch memory, but that's still a branch 2014-04-13T07:49:09Z stassats: ah, the problem is in the lowtag 2014-04-13T07:56:07Z crixxus quit (Ping timeout: 252 seconds) 2014-04-13T08:23:45Z drmeister joined #sbcl 2014-04-13T08:28:29Z drmeister quit (Ping timeout: 258 seconds) 2014-04-13T08:37:59Z stassats: my strategy: LEA EAX, [RCX-15] TEST AL, 7 => that way both other-lowtag and list-lowtag pass the test, then do CMP BYTE PTR [RCX-15], 69, that way both NIL and other symbols pass the test 2014-04-13T08:55:25Z segv- quit (Ping timeout: 250 seconds) 2014-04-13T09:00:54Z stassats: i think the problem would be random junk behind the first cons 2014-04-13T09:18:19Z ferada quit (*.net *.split) 2014-04-13T09:18:41Z stassats: yep, (list (make-array 2 :element-type '(unsigned-byte 64) :initial-element sb-vm:symbol-header-widetag)) breaks my test 2014-04-13T09:18:43Z stassats: bummer 2014-04-13T09:19:35Z ferada joined #sbcl 2014-04-13T09:37:25Z kludge` quit (Ping timeout: 252 seconds) 2014-04-13T09:39:14Z kludge` joined #sbcl 2014-04-13T10:00:55Z ams quit (Changing host) 2014-04-13T10:00:56Z ams joined #sbcl 2014-04-13T10:11:48Z fridim_ joined #sbcl 2014-04-13T10:12:30Z drmeister joined #sbcl 2014-04-13T10:16:53Z drmeister quit (Ping timeout: 250 seconds) 2014-04-13T10:25:59Z yacks quit (Ping timeout: 252 seconds) 2014-04-13T10:30:01Z yacks joined #sbcl 2014-04-13T10:37:07Z pnpuff joined #sbcl 2014-04-13T10:49:02Z antoszka quit (Quit: +++ killed by SIGSEGV +++) 2014-04-13T11:15:45Z luis joined #sbcl 2014-04-13T11:39:31Z pnpuff quit (Remote host closed the connection) 2014-04-13T11:43:08Z fridim_ quit (Ping timeout: 250 seconds) 2014-04-13T11:56:33Z oleo quit (Ping timeout: 245 seconds) 2014-04-13T11:57:23Z oleo joined #sbcl 2014-04-13T11:59:32Z davazp joined #sbcl 2014-04-13T12:01:09Z fridim_ joined #sbcl 2014-04-13T12:01:21Z drmeister joined #sbcl 2014-04-13T12:06:36Z drmeister quit (Ping timeout: 276 seconds) 2014-04-13T12:08:15Z fridim_ quit (Ping timeout: 250 seconds) 2014-04-13T12:09:33Z stassats quit (Ping timeout: 250 seconds) 2014-04-13T12:21:59Z stassats joined #sbcl 2014-04-13T12:22:08Z stassats quit (Remote host closed the connection) 2014-04-13T12:22:44Z antoszka joined #sbcl 2014-04-13T12:35:11Z drmeister joined #sbcl 2014-04-13T13:11:25Z harovali1 joined #sbcl 2014-04-13T13:11:46Z harovali quit (Ping timeout: 258 seconds) 2014-04-13T13:33:49Z ccl-logbot joined #sbcl 2014-04-13T13:33:49Z 2014-04-13T13:33:49Z names: ccl-logbot harovali1 drmeister antoszka davazp oleo luis yacks kludge` ferada Guest24347 angavrilov ivan`` xymox ams @Krystof kanru brucem _8hzp scymtym minion Munksgaard joshe Blkt fe[nl]ix Posterdati pchrist leoc afleck milosn fikusz specbot irsol jaimef christoph_debian mood flip214 psilord edgar-rft White_Flame pjb tmh_ slyrus |3b| DGASAU reb prxq_ ASau asedeno Bike faheem yauz jsnell_ pkhuong redline6561 2014-04-13T14:15:03Z nyef joined #sbcl 2014-04-13T14:15:37Z nyef: G'morning all. 2014-04-13T14:16:40Z oleo: morning 2014-04-13T14:56:35Z Guest24347 quit (Ping timeout: 245 seconds) 2014-04-13T15:00:13Z scymtym_ joined #sbcl 2014-04-13T15:09:45Z attila_lendvai joined #sbcl 2014-04-13T15:09:45Z attila_lendvai quit (Changing host) 2014-04-13T15:09:45Z attila_lendvai joined #sbcl 2014-04-13T15:38:07Z yacks quit (Ping timeout: 252 seconds) 2014-04-13T15:43:42Z nyef: What's *backend-register-save-penalty*, and why is it always set to 3? 2014-04-13T15:52:27Z drmeister quit (Remote host closed the connection) 2014-04-13T15:59:57Z drmeister joined #sbcl 2014-04-13T16:26:14Z pkhuong: we try to gauge the importance of giving each TN a register rather than a stack slot 2014-04-13T16:26:37Z pkhuong: a TN's score is decreased by *b-r-s-p* for each time it's live across a call and has to be spilled to the stack anyway 2014-04-13T16:26:44Z pkhuong: + loop depth trickery. 2014-04-13T16:33:16Z nyef: What's the effect of having it at 0? 2014-04-13T16:39:16Z yacks joined #sbcl 2014-04-13T16:39:45Z prxq_ is now known as prxq 2014-04-13T16:40:43Z pkhuong: nyef: b-r-s-p? 2014-04-13T16:41:06Z pkhuong: we'll give registers to TNs that spend more time getting spilled/unspilled that actually being worked on. 2014-04-13T16:53:45Z crixus joined #sbcl 2014-04-13T16:57:53Z nyef: So it's an optimization issue, not a correctness issue, right? 2014-04-13T16:59:05Z pkhuong: correct. 2014-04-13T16:59:17Z nyef: Thank you. 2014-04-13T16:59:38Z pkhuong: perhaps we could include a no-op allocator that only takes care of correctness. 2014-04-13T17:00:08Z nyef: The default value is zero, perhaps we should just bump it to three and remove the parameter from the backends? 2014-04-13T17:01:08Z pkhuong: ... On the one hand, I see the point in theory. On the other hand, I don't think any of us will have time to tweak this value for more than x86oids. 2014-04-13T17:01:44Z nyef: It's three across the board right now anyway, except for the ARM backend which doesn't set it. 2014-04-13T17:10:50Z nyef: Do we really need both GENERIC-+ and the static-fun VOP for TWO-ARG-+ ? 2014-04-13T17:12:22Z nyef: The GENERIC- versions seem to pick off one simple case and then fault to the TWO-ARG- version anyway, so why tell the compiler about both? 2014-04-13T17:13:08Z nyef: (Now, it could be that the TWO-ARG- version came first, and then someone added the GENERIC- version as a speed-hack, which would be both plausible and far enough back in time that we wouldn't have any authoritative answer.) 2014-04-13T17:20:14Z pnpuff joined #sbcl 2014-04-13T17:45:53Z ASau` joined #sbcl 2014-04-13T17:46:31Z Bike_ joined #sbcl 2014-04-13T17:46:34Z Bike quit (Read error: Connection reset by peer) 2014-04-13T17:49:18Z ASau quit (Ping timeout: 258 seconds) 2014-04-13T17:51:24Z pkhuong: pretty sure that's the point 2014-04-13T17:51:38Z pkhuong: save a teensy bit of space by calling out to generic-+ 2014-04-13T17:51:49Z ASau` is now known as ASau 2014-04-13T17:51:55Z pkhuong: and we only need a static fun but not a VOP for two-arg-+. 2014-04-13T17:52:35Z nyef: DEFINE-STATIC-FUN creates a VOP. 2014-04-13T17:55:09Z pnpuff left #sbcl 2014-04-13T17:55:19Z pkhuong: that sucks. 2014-04-13T17:58:08Z nyef: How so? How else would the compiler know that < with two parameters should be coverted to TWO-ARG-< ? 2014-04-13T17:58:33Z nyef: It's a VOP with a :TRANSLATE option, that inherits from one of the STATIC-FUN-TEMPLATE VOPs. 2014-04-13T17:59:00Z stassats joined #sbcl 2014-04-13T18:00:12Z Bike_ is now known as Bike 2014-04-13T18:00:15Z nyef: At any rate, it looks like as long as the two-arg-whatever functions can handle the FIXNUMxFIXNUM cases, I can defer all of the :FULL-CALL assembly routines. 2014-04-13T18:04:25Z pkhuong: nyef: but there's no need for two-arg-+ as a VOP 2014-04-13T18:06:01Z fridim_ joined #sbcl 2014-04-13T18:08:18Z nyef: You need it if you're going to use it as a static-fun directly, rather than via generic-+. 2014-04-13T18:10:19Z davazp quit (Remote host closed the connection) 2014-04-13T18:24:28Z harovali1 quit (Ping timeout: 245 seconds) 2014-04-13T18:30:16Z harovali joined #sbcl 2014-04-13T18:30:36Z harovali quit (Max SendQ exceeded) 2014-04-13T18:32:09Z harovali joined #sbcl 2014-04-13T18:48:40Z jaimef quit (Excess Flood) 2014-04-13T18:48:47Z jaimef joined #sbcl 2014-04-13T18:49:38Z nyef: Hrm. Looks like the static symbols take up #x470 octets... And no static-fdefns yet. 2014-04-13T18:50:15Z nyef: fdefns are #x10 octets each. 2014-04-13T18:52:44Z nyef: And our offsets are required to be ub-12. 2014-04-13T18:53:08Z nyef: Hunh. Guess we're in business, then. 2014-04-13T18:54:43Z eudoxia joined #sbcl 2014-04-13T19:03:21Z pchrist quit (Quit: leaving) 2014-04-13T19:04:19Z nyef quit (Quit: Gone for a couple hours.) 2014-04-13T19:07:30Z pchrist joined #sbcl 2014-04-13T19:37:15Z stassats: i guess i can't make symbolp VOP any better, but i found a way to shave one byte on x86 2014-04-13T19:38:08Z stassats: courtesy of CMP EAX having a shorter encoding, can't do that on x86-64 since something a cons which has higher bits the same as NIL 2014-04-13T19:40:57Z attila_lendvai quit (Quit: Leaving.) 2014-04-13T20:00:34Z stassats: and one byte off consp vop too 2014-04-13T20:03:50Z angavrilov quit (Remote host closed the connection) 2014-04-13T20:25:39Z stassats: a crazy idea for a symbolp test: make cons point to the CDR, make the low byte of the NIL address to be equal to symbol-header-widetag, that way CMP BYTE PTR [RCX-15], SYMBOL-WIDETAG will land up on the CAR of the NIL 2014-04-13T20:30:19Z stassats: now, can a widetag low bit be 1? right now it's all 0 2014-04-13T20:36:41Z pkhuong: stassats: wouldn't a low bit of 0 decode as a fixnum? 2014-04-13T20:37:17Z stassats: i'm talking about widetags 2014-04-13T20:40:20Z stassats: basically i want to make SYMBOL-WIDETAG to be the same as the first byte of NIL-VALUE, and shift the cons pointer to point to a CDR instead of a CAR 2014-04-13T20:40:21Z pkhuong: yes, but the gc must be able to tell the difference between a cons and a header. 2014-04-13T20:42:04Z stassats: well, cons will stay the same, just all pointers are shifted, and NIL doesn't need to be gced 2014-04-13T20:42:39Z stassats: i don't think that complication is worth it, but it's nice to try this 2014-04-13T20:45:48Z stassats: it could save about 10 bytes for SYMBOLP and make it marginally faster 2014-04-13T20:48:16Z stassats: well, i haven't thought that a cons with a NIL in its car will break this, another bummer 2014-04-13T20:53:35Z Krystof: it's OK, that never happens 2014-04-13T20:53:51Z Krystof: why would anyone want to use (NIL)? surely all uses can be replaced with just plain NIL 2014-04-13T20:56:08Z stassats quit (Ping timeout: 245 seconds) 2014-04-13T21:01:21Z fridim_ quit (Ping timeout: 258 seconds) 2014-04-13T22:07:59Z eudoxia quit (Quit: leaving) 2014-04-13T22:08:26Z nyef joined #sbcl 2014-04-13T22:08:38Z nyef: Hello all. 2014-04-13T22:18:24Z davazp joined #sbcl 2014-04-13T22:22:42Z slyrus quit (Ping timeout: 240 seconds) 2014-04-13T23:39:49Z crixus quit (Read error: Connection reset by peer) 2014-04-14T00:00:05Z oleo quit (Ping timeout: 252 seconds) 2014-04-14T00:00:48Z oleo joined #sbcl 2014-04-14T00:05:31Z nyef quit (Quit: G'night all.) 2014-04-14T00:29:18Z edgar-rft quit (Remote host closed the connection) 2014-04-14T00:30:10Z edgar-rft joined #sbcl 2014-04-14T01:35:55Z prxq_ joined #sbcl 2014-04-14T01:38:53Z prxq quit (Ping timeout: 258 seconds) 2014-04-14T02:34:14Z nyef joined #sbcl 2014-04-14T02:36:03Z nyef: minion: memo for fiveop: Your turn (dies from lack of a VOP to translate CL:> for two SINGLE-FLOATs): http://repo.or.cz/w/sbcl/nyef.git/shortlog/refs/heads/arm-port-2 2014-04-14T02:36:03Z minion: Remembered. I'll tell fiveop when he/she/it next speaks. 2014-04-14T02:37:36Z nyef quit (Client Quit) 2014-04-14T02:38:07Z zRecursive joined #sbcl 2014-04-14T02:38:13Z christoph_debian quit (Ping timeout: 245 seconds) 2014-04-14T02:38:16Z zRecursive: I forget how to fix "grovel-headers.c:549: error: 'page_index_t' undeclared (first use in this function)" when building sbcl on freebsd, any light ? 2014-04-14T02:46:36Z joshe: any other errors, like "gc.h not found" ? 2014-04-14T02:46:54Z zRecursive: no 2014-04-14T02:47:21Z joshe: what was the gcc command which make used to try and compile the file? 2014-04-14T02:48:01Z zRecursive: sh make.sh --prefix=/home/***/sbcl/ --xc-host="sbcl --disable-debugger --no-sysinit --no-userinit" 2014-04-14T02:48:41Z joshe: didn't make output some gcc command right before failing to compile grovel-headers? 2014-04-14T02:49:37Z zRecursive: pls wait, building it now ... 2014-04-14T02:50:05Z joshe: ok 2014-04-14T02:52:01Z christoph_debian joined #sbcl 2014-04-14T02:52:48Z zRecursive: cc -I/usr/local/include -fno-omit-frame-pointer -m32 -I../src/runtime -L/usr/local/lib grovel-headers.c -lutil -o grovel-headers 2014-04-14T02:54:21Z joshe: that looks fine 2014-04-14T02:55:34Z zRecursive: but reports "grovel-headers.c:549: error: 'page_index_t' undeclared (first use in this function)" 2014-04-14T02:56:00Z joshe: what about if you do (cd tools-for-build && cc -I/usr/local/include -fno-omit-frame-pointer -m32 -I../src/runtime grovel-headers.c -E) | grep page_index_t 2014-04-14T02:57:42Z zRecursive: { page_index_t foo; printf("(define-alien-type " "page-index-t" " (%s %d))\n", (((foo=-1)<0) ? "signed" : "unsigned"), (8 * (sizeof foo))); }; 2014-04-14T02:58:31Z joshe: hm 2014-04-14T02:59:34Z joshe: it this a git checkout or from a source tarball? 2014-04-14T02:59:57Z zRecursive: `git pull` the latest version 2014-04-14T03:00:16Z joshe: 'git status' shows no local changes? 2014-04-14T03:00:58Z zRecursive: yeah, i never change it 2014-04-14T03:01:28Z joshe: the latest revision you see is d299ab688be6bb155dad6277ee1a8b319398856d ? 2014-04-14T03:01:29Z zRecursive: just `git pull` and building it 2014-04-14T03:02:03Z zRecursive: commit d299ab688be6bb155dad6277ee1a8b319398856d Author: Stas Boukarev Date: Mon Apr 14 00:05:37 2014 +0400 Micro-optimize CONSP and SYMBOLP VOPs on x86. ... 2014-04-14T03:02:16Z joshe: I must be missing something obvious, I see no way you can be missing the page_index_t definition with an unmodified tree 2014-04-14T03:03:15Z zRecursive: IIRC, it seems page_index_t is in other header file on freebsd ... 2014-04-14T03:03:36Z joshe: oh, you mean freebsd has their own? 2014-04-14T03:04:03Z joshe: I would expect a different compiler error then 2014-04-14T03:04:12Z zRecursive: maybe , as i ever built it successful before 2014-04-14T03:04:20Z joshe: when was the last working build you did? 2014-04-14T03:05:24Z zRecursive: cannot find it now, but i rememeber just changing CFLAGS or something 2014-04-14T03:06:18Z joshe: you could look at /usr/ports/lang/sbcl and see what it does, perhaps there's a patch there which should be committed to sbcl 2014-04-14T03:07:51Z zRecursive: ok 2014-04-14T03:35:06Z attila_lendvai joined #sbcl 2014-04-14T03:35:06Z attila_lendvai quit (Changing host) 2014-04-14T03:35:06Z attila_lendvai joined #sbcl 2014-04-14T04:18:35Z scymtym_ quit (Ping timeout: 252 seconds) 2014-04-14T05:01:21Z davazp quit (Remote host closed the connection) 2014-04-14T05:03:26Z slyrus joined #sbcl 2014-04-14T05:23:44Z drmeister quit (Remote host closed the connection) 2014-04-14T05:23:45Z flip214: pkhuong: still awake? 2014-04-14T05:28:42Z oleo quit (Remote host closed the connection) 2014-04-14T06:12:27Z yacks quit (Ping timeout: 250 seconds) 2014-04-14T06:19:24Z prxq_ is now known as prxq 2014-04-14T06:28:25Z harovali quit (Read error: Connection reset by peer) 2014-04-14T06:31:54Z harovali joined #sbcl 2014-04-14T06:32:13Z harovali quit (Max SendQ exceeded) 2014-04-14T06:33:19Z harovali joined #sbcl 2014-04-14T06:33:37Z harovali quit (Max SendQ exceeded) 2014-04-14T06:35:36Z harovali joined #sbcl 2014-04-14T06:45:52Z yacks joined #sbcl 2014-04-14T06:55:00Z angavrilov joined #sbcl 2014-04-14T07:12:37Z drmeister joined #sbcl 2014-04-14T07:17:09Z drmeister quit (Ping timeout: 252 seconds) 2014-04-14T07:27:21Z stassats joined #sbcl 2014-04-14T07:31:41Z ASau quit (Ping timeout: 240 seconds) 2014-04-14T07:43:44Z angavrilov quit (Ping timeout: 276 seconds) 2014-04-14T07:45:00Z zRecursive left #sbcl 2014-04-14T07:51:51Z angavrilov joined #sbcl 2014-04-14T08:58:35Z stassats: more strong typechecks do not override more relaxed ones: (the vector (the array x)) checks first for array, then for vector 2014-04-14T08:58:46Z stassats: even though just checking for VECTOR would do 2014-04-14T09:01:27Z drmeister joined #sbcl 2014-04-14T09:04:21Z stassats: changing already existing casts doesn't sound like a good option, maybe an additional pass to link all the casts together 2014-04-14T09:05:10Z stassats: there's already ir1-merge-casts 2014-04-14T09:05:43Z drmeister quit (Ping timeout: 245 seconds) 2014-04-14T09:08:37Z stassats: it appears to be doing things too late, because it already receives calls to arrayp and vectorp 2014-04-14T09:14:28Z irsol quit (Ping timeout: 245 seconds) 2014-04-14T09:19:48Z hlavaty joined #sbcl 2014-04-14T09:28:45Z sdemarre joined #sbcl 2014-04-14T09:31:08Z kludge` quit (Ping timeout: 245 seconds) 2014-04-14T09:33:27Z sdemarre quit (Ping timeout: 258 seconds) 2014-04-14T09:35:56Z kludge` joined #sbcl 2014-04-14T09:57:45Z stassats quit (Ping timeout: 252 seconds) 2014-04-14T10:42:07Z DGASAU quit (Remote host closed the connection) 2014-04-14T10:43:13Z DGASAU joined #sbcl 2014-04-14T10:50:09Z drmeister joined #sbcl 2014-04-14T10:54:33Z drmeister quit (Ping timeout: 250 seconds) 2014-04-14T11:11:17Z irsol joined #sbcl 2014-04-14T12:59:07Z drmeister joined #sbcl 2014-04-14T13:04:55Z eudoxia joined #sbcl 2014-04-14T13:09:57Z psilord quit (Quit: Leaving.) 2014-04-14T13:16:49Z drmeister quit (Read error: Connection reset by peer) 2014-04-14T13:37:58Z drmeister joined #sbcl 2014-04-14T13:39:55Z drmeister quit (Remote host closed the connection) 2014-04-14T13:43:54Z nyef joined #sbcl 2014-04-14T13:44:06Z nyef: G'morning all. 2014-04-14T13:48:15Z oleo joined #sbcl 2014-04-14T13:53:33Z drmeister joined #sbcl 2014-04-14T13:57:18Z segv- joined #sbcl 2014-04-14T14:09:10Z psilord joined #sbcl 2014-04-14T14:11:46Z foom joined #sbcl 2014-04-14T14:13:18Z drmeister quit (Read error: No route to host) 2014-04-14T14:13:27Z drmeiste_ joined #sbcl 2014-04-14T14:16:22Z sdemarre joined #sbcl 2014-04-14T14:30:21Z sdemarre quit (Ping timeout: 250 seconds) 2014-04-14T14:34:17Z nyef: Does anyone know anything about the "### Note:" comments in compiler/ppc/vm about load-tns not working with element-size other than 1? 2014-04-14T14:36:00Z nyef: The impression I'm getting from the code is that either PPC has 64-bit float registers, or it has 64 float registers, and either way the :element-size thing is honored by LOAD-TNs. 2014-04-14T14:39:54Z oleo: http://publib.boulder.ibm.com/infocenter/pseries/v5r3/topic/com.ibm.aix.aixassem/doc/alangref/floating.htm 2014-04-14T14:40:55Z nyef: Okay, that fits with about half of what I'm seeing... Making the comment even MORE misleading, of course. 2014-04-14T14:40:57Z nyef: And thanks. 2014-04-14T14:41:21Z nyef: I seem to have misplaced my PPC archref, and I wasn't quite expecting to need it this week. /-: 2014-04-14T14:44:04Z oleo: http://gcc.gnu.org/onlinedocs/gcc/RS₀02f6000-and-PowerPC-Options.html maybe too.... 2014-04-14T14:45:06Z oleo: http://refspecs.linuxfoundation.org/ELF/ppc64/PPC-elf64abi-1.9.html seems to fit too 2014-04-14T14:45:57Z oleo: The 64-bit PowerPC ELF ABI is not the same as the 32-bit PowerPC ELF ABI, nor is it a simple extension. A system which supports the 64-bit PowerPC ELF ABI may, but need not, support the 32-bit PowerPC ELF ABI. 2014-04-14T14:46:01Z oleo: that part especially.... 2014-04-14T14:49:08Z oleo: The PowerPC standard specifies both 32-bit and 64-bit data paths. Early implementations were 32-bit (e.g. PowerPC 601); later higher-performance implementations were 64-bit (e.g. PowerPC 620). A PowerPC has 32 integer registers (32- or 64 bit) and 32 floating-point (IEEE standard 64 bit) floating-point registers. 2014-04-14T14:49:27Z oleo: from http://dictionary.reference.com/browse/powerpc 2014-04-14T14:50:54Z nyef: Yeah, fun and games. 2014-04-14T14:51:48Z nyef: Okay, finished reviewing the history of compiler/pack in SBCL and there's no sign of an issue with LOAD-TNs not honoring SC element sizes since day one. 2014-04-14T14:59:55Z stassats joined #sbcl 2014-04-14T15:00:00Z nyef: Found it! 1991-02-24 11:20:00, ram modified PACK, part of the commit message is "Changed load-TN packing to handle multi-location load-tns." and contains precisely the code that I had seen. This pre-dates the SBCL fork by just a scant few years. 2014-04-14T15:00:12Z nyef: (Like, eight of them or so.) 2014-04-14T15:04:37Z oleo: is that in cmu ? 2014-04-14T15:04:42Z nyef: Yeah. 2014-04-14T15:04:57Z oleo: i thought it does not work with 64bit archs..... 2014-04-14T15:05:26Z oleo: so they tried to port it and gave up ? 2014-04-14T15:05:30Z stassats: sbcl doesn't have 64-bit ppc backend either 2014-04-14T15:05:41Z nyef: ... so? No, this is for 32-bit backends with FP support. 2014-04-14T15:05:50Z oleo: ok 2014-04-14T15:06:35Z nyef: I found the source of the MIPS vm comment about the same, and it's from September 1990. And the MIPS backend is STILL set up to compensate for the issue today. 2014-04-14T15:07:23Z slyrus quit (Ping timeout: 245 seconds) 2014-04-14T15:07:42Z oleo: loop for i from 0 to 31 yes only 32bit 2014-04-14T15:08:11Z oleo: but that's the single-float register 2014-04-14T15:08:16Z oleo: look at the double-float one.... 2014-04-14T15:08:28Z oleo: :element-size 2 :alignment 2 2014-04-14T15:08:31Z nyef: Commented out. 2014-04-14T15:08:50Z oleo: ah you right bleh 2014-04-14T15:09:00Z nyef: The float registers are still 64 bits, which makes the comment make EVEN LESS SENSE. 2014-04-14T15:10:03Z nyef: And if the restriction against wide LOAD-TNs still applied, what's up with the complex float registers? 2014-04-14T15:11:08Z oleo: but the single-float register thing points to non-descriptor-stack.... 2014-04-14T15:11:23Z nyef: Which has it's own rules as an SC. 2014-04-14T15:11:29Z oleo: :unbounded 2014-04-14T15:11:31Z nyef thwaps himself. 2014-04-14T15:13:37Z oleo: so it gets extended..... 2014-04-14T15:14:22Z oleo: as far as i understand..... 2014-04-14T15:16:08Z nyef: By the rules of "you found it, you deal with it if you can", the PPC and MIPS VM cleanups are on me. And I now have some idea as to the design space for float-register SBs and SCs, so I can fix them up for the ARM port if necessary. 2014-04-14T15:16:50Z slyrus joined #sbcl 2014-04-14T15:17:35Z nyef: Oh, and I also (as of earlier this morning) have an angle for implementing software FP for an SBCL backend. 2014-04-14T15:18:10Z oleo: will that be slower ? 2014-04-14T15:18:46Z oleo: wait that's for when the arch lacks float registers ? 2014-04-14T15:18:47Z nyef: Slower than hardware FP? Absolutely. But if we ever decide to target an architecture without FP support... 2014-04-14T15:18:55Z oleo: ah ok 2014-04-14T15:19:17Z oleo: sounds good 2014-04-14T15:19:45Z stassats quit (Ping timeout: 250 seconds) 2014-04-14T15:24:21Z ltbarcly joined #sbcl 2014-04-14T15:25:04Z ltbarcly quit (Client Quit) 2014-04-14T15:35:09Z attila_lendvai quit (Quit: Leaving.) 2014-04-14T15:36:40Z fridim_ joined #sbcl 2014-04-14T15:38:47Z nicdev joined #sbcl 2014-04-14T15:59:41Z fridim_ quit (Ping timeout: 276 seconds) 2014-04-14T16:05:27Z stassats joined #sbcl 2014-04-14T16:05:53Z |3b|: nyef: could it do float sizes not supported by the hardware as well? 2014-04-14T16:06:46Z stassats: what about an architecture without MMU? 2014-04-14T16:10:07Z foom quit (Read error: Connection reset by peer) 2014-04-14T16:11:44Z asedeno quit (Ping timeout: 258 seconds) 2014-04-14T16:15:17Z drmeiste_ quit (Ping timeout: 276 seconds) 2014-04-14T16:19:01Z nyef: |3b|: Float sizes not supported by architecture? Plausibly, yes. 2014-04-14T16:19:37Z nyef: stassats: That's also doable, cheneygc, add a bounds-check in the allocation sequence. 2014-04-14T16:20:03Z nyef: Well, and pick your heap spaces carefully. 2014-04-14T16:20:12Z nyef: Honestly, the non-mmu options are pretty poor. 2014-04-14T16:20:29Z yacks quit (Ping timeout: 240 seconds) 2014-04-14T16:21:18Z drmeister joined #sbcl 2014-04-14T16:23:05Z fridim_ joined #sbcl 2014-04-14T16:25:02Z ltbarcly joined #sbcl 2014-04-14T16:25:42Z ltbarcly quit (Client Quit) 2014-04-14T16:25:58Z stassats: i don't think anybody wants to port to new architectures any time soon except for the ARM port 2014-04-14T16:26:05Z stassats: 64-bit soon could be nice too 2014-04-14T16:26:08Z stassats: 64-bit ppc 2014-04-14T16:26:53Z nyef: Maybe we can drum up interest in an Itanic port? 2014-04-14T16:27:36Z pjb: Yes, cloning the Titanic is in the air. 2014-04-14T16:27:50Z pjb: :-) sorry. 2014-04-14T16:31:40Z ltbarcly joined #sbcl 2014-04-14T16:31:41Z fridim_ quit (Ping timeout: 250 seconds) 2014-04-14T16:36:27Z slyrus quit (Ping timeout: 250 seconds) 2014-04-14T16:40:51Z slyrus joined #sbcl 2014-04-14T16:45:46Z eudoxia quit (Quit: lunchtime) 2014-04-14T16:54:57Z marcux joined #sbcl 2014-04-14T16:55:14Z marcux left #sbcl 2014-04-14T16:55:52Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-14T17:04:13Z sdemarre joined #sbcl 2014-04-14T17:08:31Z eudoxia joined #sbcl 2014-04-14T17:53:04Z ltbarcly joined #sbcl 2014-04-14T17:53:17Z kanru quit (Ping timeout: 240 seconds) 2014-04-14T17:58:13Z ltbarcly quit (Ping timeout: 245 seconds) 2014-04-14T17:59:26Z asedeno joined #sbcl 2014-04-14T18:01:27Z ltbarcly joined #sbcl 2014-04-14T18:03:30Z ASau joined #sbcl 2014-04-14T18:21:02Z drmeiste_ joined #sbcl 2014-04-14T18:21:24Z drmeister quit (Read error: Connection reset by peer) 2014-04-14T18:25:08Z stassats quit (Ping timeout: 258 seconds) 2014-04-14T18:30:59Z harovali quit (Read error: Connection reset by peer) 2014-04-14T18:33:00Z harovali joined #sbcl 2014-04-14T18:33:21Z harovali quit (Max SendQ exceeded) 2014-04-14T18:35:01Z harovali joined #sbcl 2014-04-14T18:35:23Z harovali quit (Max SendQ exceeded) 2014-04-14T18:36:31Z harovali joined #sbcl 2014-04-14T18:36:53Z harovali quit (Max SendQ exceeded) 2014-04-14T18:38:47Z sdemarre quit (Ping timeout: 252 seconds) 2014-04-14T18:38:55Z harovali joined #sbcl 2014-04-14T18:39:17Z harovali quit (Max SendQ exceeded) 2014-04-14T18:44:28Z drmeiste_ quit (Ping timeout: 245 seconds) 2014-04-14T18:55:48Z eudoxia quit (Quit: leaving) 2014-04-14T18:56:59Z harovali joined #sbcl 2014-04-14T18:57:22Z harovali quit (Max SendQ exceeded) 2014-04-14T18:58:01Z harovali joined #sbcl 2014-04-14T18:58:23Z harovali quit (Max SendQ exceeded) 2014-04-14T19:16:52Z harovali joined #sbcl 2014-04-14T19:17:14Z harovali quit (Max SendQ exceeded) 2014-04-14T19:18:20Z angavrilov quit (Remote host closed the connection) 2014-04-14T19:18:50Z harovali joined #sbcl 2014-04-14T19:19:11Z harovali quit (Max SendQ exceeded) 2014-04-14T19:20:17Z harovali joined #sbcl 2014-04-14T19:20:40Z harovali quit (Max SendQ exceeded) 2014-04-14T19:21:17Z irsol quit (Ping timeout: 240 seconds) 2014-04-14T19:21:37Z foom joined #sbcl 2014-04-14T19:22:50Z harovali joined #sbcl 2014-04-14T19:23:11Z harovali quit (Max SendQ exceeded) 2014-04-14T19:25:05Z harovali joined #sbcl 2014-04-14T19:25:26Z harovali quit (Max SendQ exceeded) 2014-04-14T19:26:22Z harovali joined #sbcl 2014-04-14T19:26:45Z harovali quit (Max SendQ exceeded) 2014-04-14T19:27:21Z harovali joined #sbcl 2014-04-14T19:27:44Z harovali quit (Max SendQ exceeded) 2014-04-14T19:28:22Z harovali joined #sbcl 2014-04-14T19:28:44Z harovali quit (Max SendQ exceeded) 2014-04-14T19:29:37Z harovali joined #sbcl 2014-04-14T19:29:58Z harovali quit (Max SendQ exceeded) 2014-04-14T19:32:02Z harovali joined #sbcl 2014-04-14T19:32:24Z harovali quit (Max SendQ exceeded) 2014-04-14T19:33:27Z harovali joined #sbcl 2014-04-14T19:33:50Z harovali quit (Max SendQ exceeded) 2014-04-14T19:35:41Z harovali joined #sbcl 2014-04-14T19:36:03Z harovali quit (Max SendQ exceeded) 2014-04-14T19:36:27Z harovali joined #sbcl 2014-04-14T19:36:49Z harovali quit (Max SendQ exceeded) 2014-04-14T19:38:26Z harovali joined #sbcl 2014-04-14T19:38:48Z harovali quit (Max SendQ exceeded) 2014-04-14T19:40:01Z harovali joined #sbcl 2014-04-14T19:40:23Z harovali quit (Max SendQ exceeded) 2014-04-14T19:45:12Z irsol joined #sbcl 2014-04-14T19:48:35Z drmeister joined #sbcl 2014-04-14T19:52:27Z fiveop joined #sbcl 2014-04-14T19:58:55Z harovali joined #sbcl 2014-04-14T19:59:17Z harovali quit (Max SendQ exceeded) 2014-04-14T20:17:42Z harovali joined #sbcl 2014-04-14T20:18:04Z harovali quit (Max SendQ exceeded) 2014-04-14T20:20:15Z harovali joined #sbcl 2014-04-14T20:20:38Z harovali quit (Max SendQ exceeded) 2014-04-14T20:21:22Z harovali joined #sbcl 2014-04-14T20:21:43Z harovali quit (Max SendQ exceeded) 2014-04-14T20:23:17Z harovali joined #sbcl 2014-04-14T20:23:40Z harovali quit (Max SendQ exceeded) 2014-04-14T20:24:27Z harovali joined #sbcl 2014-04-14T20:24:49Z harovali quit (Max SendQ exceeded) 2014-04-14T20:25:55Z harovali joined #sbcl 2014-04-14T20:26:17Z harovali quit (Max SendQ exceeded) 2014-04-14T20:31:59Z nyef: Okay, that's the float-registers SB/SC damage fixed up, including the instruction encodings as best as I can figure. 2014-04-14T20:32:44Z nyef: I guess the next step is MOVE VOPs or something. 2014-04-14T20:34:12Z eudoxia joined #sbcl 2014-04-14T20:42:59Z harovali joined #sbcl 2014-04-14T20:43:23Z harovali quit (Max SendQ exceeded) 2014-04-14T20:45:34Z harovali joined #sbcl 2014-04-14T20:45:35Z fridim_ joined #sbcl 2014-04-14T20:45:57Z harovali quit (Max SendQ exceeded) 2014-04-14T20:47:25Z jaimef quit (Excess Flood) 2014-04-14T20:47:36Z jaimef joined #sbcl 2014-04-14T20:47:51Z drmeister quit (Ping timeout: 252 seconds) 2014-04-14T20:47:59Z harovali joined #sbcl 2014-04-14T20:48:22Z harovali quit (Max SendQ exceeded) 2014-04-14T20:52:36Z fiveop: nyef: arm? 2014-04-14T20:52:36Z minion: fiveop, memo from nyef: Your turn (dies from lack of a VOP to translate CL:> for two SINGLE-FLOATs): http://repo.or.cz/w/sbcl/nyef.git/shortlog/refs/heads/arm-port-2 2014-04-14T20:53:05Z nyef: Yeah, ARM. 2014-04-14T20:54:28Z nyef: I've got a bit more to do just based on what was in my working tree earlier that I threw into a temporary commit, and I'm thinking about how to proceed from here. 2014-04-14T20:54:46Z fiveop: I just need to reconfigure my system (wiped everything, because windows kept thinking my linux partition is a broken ntfs partition) 2014-04-14T20:54:53Z fiveop: Then I'll pull again and get wrking :) 2014-04-14T20:55:02Z nyef: ... Ouch. Bad partition type code? 2014-04-14T20:55:24Z nyef: Of course, the first issue was trying to set up a dual-boot system or similar. 2014-04-14T20:55:38Z fiveop: Windows crashed once from overheating and after that, I didn't manage to convince it otherwise. 2014-04-14T20:55:44Z nyef: Ouch. 2014-04-14T20:55:44Z fiveop: So now it's just linux :) 2014-04-14T20:56:15Z fiveop: The only disadvantage is, that I'm now forced to use a wire. 2014-04-14T20:56:20Z nyef: Well, I'm on a bit of a roll right now, even if I have been scaring up longstanding issues with various other backends. 2014-04-14T20:56:47Z fiveop: I didn't expect there to still be wireless cards without linux support, when I bought the machine. 2014-04-14T20:57:10Z nyef: Heh. Oh, there are. 2014-04-14T20:57:18Z nyef: ndiswrapper? 2014-04-14T20:57:21Z fiveop: I probably have some time this weekend. 2014-04-14T20:58:12Z fiveop: We have two national holidays because of easter, so I hope I can take a look at the float stuff. 2014-04-14T20:58:59Z nyef: Mmm. I'm busy on the sunday, but I hope to be able to make some decent progress before then. 2014-04-14T20:59:07Z fiveop: good night (or day :)) 2014-04-14T21:01:55Z fiveop: I have some commits locally with vops for floats 2014-04-14T21:02:25Z fiveop: I'll push them on the weekend 2014-04-14T21:02:57Z fiveop quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2014-04-14T21:08:36Z pjb quit (Remote host closed the connection) 2014-04-14T21:09:05Z nyef quit (Quit: G'night all.) 2014-04-14T21:30:02Z eudoxia quit (Quit: leaving) 2014-04-14T21:30:21Z eudoxia joined #sbcl 2014-04-14T21:38:51Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-14T21:41:31Z drmeister joined #sbcl 2014-04-14T21:48:33Z eudoxia quit (Quit: leaving) 2014-04-14T22:02:49Z psilord quit (Quit: Leaving.) 2014-04-14T22:06:32Z eudoxia joined #sbcl 2014-04-14T22:15:17Z drmeister quit (Ping timeout: 240 seconds) 2014-04-14T22:26:19Z drmeister joined #sbcl 2014-04-14T22:32:22Z drmeister quit (Read error: Connection reset by peer) 2014-04-14T22:45:37Z jaimef quit (Excess Flood) 2014-04-14T22:46:00Z jaimef joined #sbcl 2014-04-14T22:48:17Z drmeister joined #sbcl 2014-04-14T22:53:59Z segv- quit (Ping timeout: 252 seconds) 2014-04-14T23:04:21Z ltbarcly joined #sbcl 2014-04-14T23:13:27Z drmeister quit (Remote host closed the connection) 2014-04-14T23:16:17Z drmeister joined #sbcl 2014-04-14T23:25:30Z davazp joined #sbcl 2014-04-14T23:32:37Z psilord joined #sbcl 2014-04-14T23:46:45Z scymtym_ joined #sbcl 2014-04-14T23:59:41Z zRecursive joined #sbcl 2014-04-15T00:05:43Z eudoxia quit (Quit: Lost terminal) 2014-04-15T00:08:11Z nicdev quit (Remote host closed the connection) 2014-04-15T00:17:12Z nicdev joined #sbcl 2014-04-15T00:24:53Z fridim_ quit (Ping timeout: 240 seconds) 2014-04-15T00:31:02Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-15T00:42:59Z nyef joined #sbcl 2014-04-15T00:49:16Z nyef: ... So, I have a chance to get some half-decent MIPS hardware. The only caveat being that the OS choices are irix and OpenBSD, and possibly not irix. 2014-04-15T00:50:02Z jaimef: which sgi box? 2014-04-15T00:50:10Z nyef: An Origin 350. 2014-04-15T00:50:18Z jaimef: 64bit? 2014-04-15T00:50:31Z nyef: I believe so, yes. 2014-04-15T00:51:02Z nyef: The issue being that it's IP35, which isn't supported by Linux. 2014-04-15T00:51:11Z nyef: Or FreeBSD or NetBSD. 2014-04-15T00:51:45Z nyef: And I just checked my local git: We only support MIPS on Linux right now. 2014-04-15T00:51:50Z jaimef: yeah openbsd beat us to IP35 :( 2014-04-15T00:51:56Z nyef: Which "us"? 2014-04-15T00:51:58Z jaimef: sbcl is not built on openbsd 2014-04-15T00:52:00Z jaimef: netbsd 2014-04-15T00:52:22Z jaimef: sbcl does not appear to support SGIMIPS64 2014-04-15T00:52:45Z nyef: ... Isn't there a 32-bit compatibility mode? 2014-04-15T00:54:11Z nyef: I'm SO not up for starting another new compiler backend right now. 2014-04-15T00:55:01Z jaimef: nyef: they do powerpc, i386, amd64 only for sbcl on openbsd 2014-04-15T00:55:15Z jaimef: actually I'm wrong 2014-04-15T00:59:54Z nicdev left #sbcl 2014-04-15T00:59:56Z zRecursive: i still cannot fix this http://paste.lisp.org/display/141374 2014-04-15T01:01:05Z nyef: zRecursive: Do a "sh ./make-host-1.sh" and tell me if it does anything. 2014-04-15T01:01:46Z nyef: zRecursive: And what are you running for a machine and OS? 2014-04-15T01:02:24Z zRecursive: freebsd 9.1 2014-04-15T01:03:07Z nyef: Architecture? 2014-04-15T01:04:05Z zRecursive: AMD Athlon(tm) 64 X2 Dual Core Processor 3600+ (1908.73-MHz 686-class CPU) 2014-04-15T01:05:22Z nyef: I'd say WFM, but according to git, I last tried it in April 2013. On FreeBSD 9.1-RELEASE, amd64. 2014-04-15T01:05:26Z zRecursive: nyef: beginning GENESIS, creating headers in "src/runtime/genesis" NIL, then "sh ./make-host-1.sh" ends 2014-04-15T01:06:10Z nyef: That's all the output? 2014-04-15T01:06:37Z zRecursive: there are may output before it 2014-04-15T01:06:47Z zRecursive: seems no problems 2014-04-15T01:06:50Z nyef: Hrm. 2014-04-15T01:07:56Z zRecursive: The last lines is: ; .../sbcl/obj/from-host/src/compiler/generic/genesis.fasl-tmp written\n ; compilation finished in 0:00:03.561 \nT \n* \n beginning GENESIS, creating headers in "src/runtime/genesis" \nNIL 2014-04-15T01:08:35Z nyef: Okay, so it's not the obvious minor thing. 2014-04-15T01:09:20Z zRecursive: IIRC, it is freebsd relative 2014-04-15T01:09:57Z nyef: Yeah, I have a FreeBSD 9.1-RELEASE VM handy because of a bughunt I did a while back for someone. 2014-04-15T01:11:30Z zRecursive: `find /usr/local/include/ -name '*.h'|xargs grep -i page_index_t` cannot find it ?! 2014-04-15T01:14:14Z nyef: It's in gc.h 2014-04-15T01:14:57Z nyef: (Just did a git grep page_index_t) 2014-04-15T01:16:37Z zRecursive: find /usr/local/include/ -name 'gc.h' => /usr/local/include/xorg/gc.h /usr/local/include/gc/gc.h /usr/local/include/libguile/gc.h /usr/local/include/gc.h 2014-04-15T01:16:59Z nyef: No, it's in SBCL's gc.h. 2014-04-15T01:17:30Z zRecursive: ./src/runtime/gc.h ? 2014-04-15T01:17:39Z nyef: That's the one. 2014-04-15T01:17:55Z nyef: Real near the top, too. 2014-04-15T01:18:03Z nyef: Probably line 24 or so. 2014-04-15T01:18:53Z zRecursive: it is there, but why "grovel-headers.c:547: error: 'page_index_t' undeclared (first use in this function)" ? 2014-04-15T01:19:34Z nyef: I just did a git pull, and am doing a build now to see if I can replicate the error. 2014-04-15T01:19:44Z zRecursive: ok 2014-04-15T01:22:12Z nyef: Bad news: It's into host-2 already, which means that grovel-headers already ran. 2014-04-15T01:24:44Z nyef: Hrm. 2014-04-15T01:26:44Z nyef: ... Oh! You have a /usr/local/include/gc.h. 2014-04-15T01:28:35Z nyef: And the CFLAGS in your environment probably also supplies -I/usr/local/include doesn't it? 2014-04-15T01:29:21Z nyef: Yup. That's it, right there. 2014-04-15T01:29:36Z nyef: Anything more I can do for you tonight? (-: 2014-04-15T01:32:04Z zRecursive: yeah, there is -I/usr/local/include in ~/.bashrc 2014-04-15T01:33:52Z zRecursive: nyef: thanks! i will rebuild it after `export CFLAGS=` 2014-04-15T01:34:13Z nyef: It's the small things that get you, right? 2014-04-15T01:34:34Z zRecursive: indeed small 2014-04-15T01:35:35Z ltbarcly joined #sbcl 2014-04-15T01:36:16Z zRecursive: maybe sbcl should search gc.h of its own before searching /usr/local/include ? 2014-04-15T01:36:29Z prxq_ joined #sbcl 2014-04-15T01:37:32Z nyef: The problem is that it appends to CFLAGS, allowing you to specify your own. 2014-04-15T01:37:45Z nyef: So since your -I comes before SBCL's -I, your -I wins. 2014-04-15T01:37:56Z zRecursive: i see 2014-04-15T01:38:33Z nyef: I really don't know what, if anything, to do about that. At a certain point one just throws one's hands up and bitches about weird build environments. 2014-04-15T01:39:25Z prxq quit (Ping timeout: 250 seconds) 2014-04-15T01:39:51Z zRecursive: it is not so weird though 2014-04-15T01:40:57Z nyef: Feel free to open a discussion on sbcl-devel, then. I'm sufficiently rusty with my make-fu that I'm not about to make an attempt to "improve" the situation. 2014-04-15T01:41:45Z zRecursive: ok 2014-04-15T01:45:45Z nyef: So, I can get an IP35 machine that might work, but I'd be looking at 90-120 minute build times, plus having to either port to OpenBSD or IRIX, or do quite a bit of Linux kernel hacking to get things operational. 2014-04-15T01:46:10Z nyef: And I'd far rather be hacking SBCL than Linux in my "copious spare time". 2014-04-15T01:46:24Z joshe: oh, that's annoying 2014-04-15T01:47:04Z joshe: perhaps the groveler makefile should prepend the include path 2014-04-15T01:47:06Z nyef: Oh, and because it's a MIPS64 machine and SBCL is thoroughly MIPS32, I might be looking at further problems. 2014-04-15T01:47:38Z joshe: especially if you go with openbsd, which doesn't provide a 32-bit ABI on any 64-bit kernels 2014-04-15T01:47:39Z nyef: joshe: I expect that it'd have to cover the runtime as well. 2014-04-15T01:48:01Z joshe: I thought the runtime already did the right thing 2014-04-15T01:48:04Z nyef: It... WHAT? 2014-04-15T01:48:26Z joshe: having to do something horrible to get /usr/include/util.h instead of sbcl's util.h 2014-04-15T01:48:28Z nyef: Okay, just recovered my brain from a snapshot. 2014-04-15T01:48:38Z joshe: hah 2014-04-15T01:48:55Z nyef: Still, WTF? No 32-bit support on 64-bit OpenBSD? 2014-04-15T01:49:09Z joshe: correct 2014-04-15T01:49:53Z nyef takes the LORD's name in vain, in conjunction with a reference to copulation. 2014-04-15T01:50:03Z joshe: in some special cases like i386, hppa or ppc you can run a 32-bit kernel on the hardware, but I'm pretty sure there's no mips32 port 2014-04-15T01:52:26Z nyef: For SGI? No, there isn't. The lowest OpenBSD goes to is IP20, and that's R4000. 2014-04-15T01:53:06Z joshe: right, and the non-sgi mips hardware is all 64-bit as well 2014-04-15T01:53:08Z nyef: So, I'm looking at having to spend just north of a grand to acquire a hacking project. 2014-04-15T01:53:56Z joshe: hppa is probably more affordable 2014-04-15T01:54:05Z nyef: Two grand for a decent machine. 2014-04-15T01:54:23Z nyef: Same as alpha. Nice try, though. 2014-04-15T01:54:27Z joshe: this one is halfway decent: http://www.ebay.com/itm/HP-A6057A-Visualize-C3700-73GB-4GB-CD-FX5-HP-UX-1-YEAR-WARRANTY-/330888655335?pt=COMP_EN_Servers&hash=item4d0a8031e7 2014-04-15T01:54:38Z joshe: it's what I have, except mine doesn't power on 2014-04-15T01:55:14Z nyef: You win, that's a lot better than what I was seeing. 2014-04-15T01:56:08Z joshe: I just look stuff up on openpa.net and openbsd.org/hppa.html 2014-04-15T01:56:21Z joshe: then search on ebay 2014-04-15T01:57:21Z nyef: It'd be somewhat amusing to get HPPA to be a more-supported platform than MIPS or ALPHA. 2014-04-15T01:57:30Z joshe: indeed 2014-04-15T01:57:47Z nyef: But at the same time, those are the three least-supported platforms for SBCL. 2014-04-15T01:57:47Z joshe: next: vax 2014-04-15T01:58:18Z nyef: VAX if-and-only-if there's an 800MHz-or-faster SMP system that runs Linux. 2014-04-15T01:59:55Z nicdev joined #sbcl 2014-04-15T02:00:19Z joshe: I don't think they got past 200mhz, and those are quite rare 2014-04-15T02:01:35Z oleo quit (Ping timeout: 276 seconds) 2014-04-15T02:01:49Z nyef: Good. That means that I don't have to worry about another new compiler backend for that. (-: 2014-04-15T02:03:56Z nyef: Yes, by this point I probably AM the local expert on writing a new compiler backend. That doesn't mean that I want to do so again. 2014-04-15T02:05:00Z joshe: so that means no itanium port anytime soon, eh? 2014-04-15T02:07:02Z nyef: We don't even have a working ARM port yet! 2014-04-15T02:07:14Z joshe: details, details 2014-04-15T02:07:42Z nyef: Maybe we should float the idea of an Itanic port for next year's GSoC? 2014-04-15T02:08:21Z joshe: to see how fast it can sink? 2014-04-15T02:10:21Z nyef: Yeah, exactly. 2014-04-15T02:12:25Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-15T02:13:51Z oleo joined #sbcl 2014-04-15T02:14:29Z Hydan joined #sbcl 2014-04-15T02:15:23Z nyef: Hrm. So, $1,100 for a Tezro machine supposedly in working order, running IRIX. 2014-04-15T02:15:29Z christoph_debian quit (Ping timeout: 258 seconds) 2014-04-15T02:15:41Z White_Flame quit (Remote host closed the connection) 2014-04-15T02:16:05Z nyef: Upsides, "hey, 600MHz dual-CPU MIPS". Downsides, doesn't work on any OS worth actually using. 2014-04-15T02:16:42Z nyef: Further downsides, very likely that SBCL would require substantial work to get running on any OS that will run on said hardware. 2014-04-15T02:18:31Z nyef: Alternately, Linux would require substantial work to get running on said hardware, but would likely run SBCL fairly easily. 2014-04-15T02:18:52Z nyef: Modulo the bit where SBCL only barely runs on MIPS as it is. 2014-04-15T02:20:28Z joshe: would you rather port linux to the hardware or porting sbcl to openbsd/mips64 :) 2014-04-15T02:20:42Z joshe: s/ing // 2014-04-15T02:21:11Z nyef: I'd rather not have to deal with yet another instance of yak shaving. 2014-04-15T02:21:40Z nyef: The name of the game is always and forever SBCL maintenance. 2014-04-15T02:21:54Z joshe: yak shaving is in the eye of the beholder 2014-04-15T02:22:20Z nyef: ... Which is a lot funnier as soon as you remember the D&D monster. 2014-04-15T02:22:26Z joshe: as strange as it might seem, some people would place "porting a lisp compiler" in that category 2014-04-15T02:26:03Z attila_lendvai joined #sbcl 2014-04-15T02:26:03Z attila_lendvai quit (Changing host) 2014-04-15T02:26:03Z attila_lendvai joined #sbcl 2014-04-15T02:29:29Z christoph_debian joined #sbcl 2014-04-15T02:34:40Z kanru joined #sbcl 2014-04-15T02:39:17Z christoph_debian quit (Ping timeout: 276 seconds) 2014-04-15T02:39:18Z davazp quit (Remote host closed the connection) 2014-04-15T02:40:01Z nyef: Actually, what I'd really like is to have the opportunity to run the existing MIPS32 build, a threaded MIPS32 build (tested against a multi-CPU system, the REAL minimal test for a theads port) AND a MIPS64 port with both single-threaded and multi-threaded versions. 2014-04-15T02:40:09Z nyef: Preferably on the same hardware. 2014-04-15T02:52:02Z christoph_debian joined #sbcl 2014-04-15T03:03:07Z nyef: Ah, well. I give up for tonight. Should have put the light out an hour ago. 2014-04-15T03:03:11Z nyef quit (Quit: G'night all.) 2014-04-15T03:23:02Z zRecursive quit (Remote host closed the connection) 2014-04-15T03:23:16Z zRecursive joined #sbcl 2014-04-15T03:53:01Z yacks joined #sbcl 2014-04-15T04:22:21Z scymtym_ quit (Ping timeout: 250 seconds) 2014-04-15T04:23:55Z drmeister quit (Remote host closed the connection) 2014-04-15T04:29:03Z ltbarcly joined #sbcl 2014-04-15T05:00:56Z oleo quit (Remote host closed the connection) 2014-04-15T05:29:53Z Hydan quit (Ping timeout: 245 seconds) 2014-04-15T05:38:12Z pranavrc joined #sbcl 2014-04-15T05:38:12Z pranavrc quit (Changing host) 2014-04-15T05:38:12Z pranavrc joined #sbcl 2014-04-15T05:50:55Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-15T05:54:23Z angavrilov joined #sbcl 2014-04-15T06:12:50Z drmeister joined #sbcl 2014-04-15T06:17:29Z drmeister quit (Ping timeout: 265 seconds) 2014-04-15T06:22:04Z tmh__ joined #sbcl 2014-04-15T07:24:00Z prxq_ is now known as prxq 2014-04-15T07:35:28Z tomterl joined #sbcl 2014-04-15T07:48:34Z zRecursive quit (Remote host closed the connection) 2014-04-15T07:51:40Z sdemarre joined #sbcl 2014-04-15T08:01:37Z drmeister joined #sbcl 2014-04-15T08:06:14Z drmeister quit (Ping timeout: 265 seconds) 2014-04-15T08:14:22Z davazp joined #sbcl 2014-04-15T08:45:43Z ASau quit (Ping timeout: 245 seconds) 2014-04-15T09:12:23Z sdemarre quit (Ping timeout: 245 seconds) 2014-04-15T09:19:10Z pranavrc quit (Remote host closed the connection) 2014-04-15T09:19:26Z pranavrc joined #sbcl 2014-04-15T09:19:26Z pranavrc quit (Changing host) 2014-04-15T09:19:26Z pranavrc joined #sbcl 2014-04-15T09:32:06Z kludge` quit (Ping timeout: 258 seconds) 2014-04-15T09:36:15Z kludge` joined #sbcl 2014-04-15T09:50:22Z drmeister joined #sbcl 2014-04-15T09:54:43Z drmeister quit (Ping timeout: 252 seconds) 2014-04-15T09:59:35Z fridim_ joined #sbcl 2014-04-15T11:09:34Z attila_lendvai quit (Quit: Leaving.) 2014-04-15T11:31:43Z sdemarre joined #sbcl 2014-04-15T11:39:11Z drmeister joined #sbcl 2014-04-15T11:43:37Z drmeister quit (Ping timeout: 252 seconds) 2014-04-15T12:10:26Z eudoxia joined #sbcl 2014-04-15T12:20:41Z LiamH joined #sbcl 2014-04-15T12:37:14Z flip214 quit (Ping timeout: 252 seconds) 2014-04-15T12:37:23Z sdemarre quit (Ping timeout: 265 seconds) 2014-04-15T12:38:54Z flip214 joined #sbcl 2014-04-15T12:38:54Z flip214 quit (Changing host) 2014-04-15T12:38:54Z flip214 joined #sbcl 2014-04-15T13:04:51Z drmeister joined #sbcl 2014-04-15T13:09:03Z pranavrc quit 2014-04-15T13:12:19Z flip214 quit (Ping timeout: 250 seconds) 2014-04-15T13:12:33Z flip214 joined #sbcl 2014-04-15T13:12:33Z flip214 quit (Changing host) 2014-04-15T13:12:33Z flip214 joined #sbcl 2014-04-15T13:15:25Z LiamH quit (Quit: Leaving.) 2014-04-15T13:43:53Z oleo joined #sbcl 2014-04-15T13:59:27Z segv- joined #sbcl 2014-04-15T14:04:05Z tmh__ quit (Remote host closed the connection) 2014-04-15T14:14:32Z oleo quit (Ping timeout: 265 seconds) 2014-04-15T14:14:59Z oleo joined #sbcl 2014-04-15T14:34:36Z nyef joined #sbcl 2014-04-15T14:34:50Z nyef: G'morning all. 2014-04-15T14:34:51Z minion: nyef, memo from hlavaty: cl-olefs updated, you can pull from http://logand.com/git/cl-olefs.git and thanks for the sb-kernel:make-double-float hint; i'm looking forward to your mac encoding bug report :-) 2014-04-15T14:37:37Z jaimef quit (Excess Flood) 2014-04-15T14:40:06Z jaimef joined #sbcl 2014-04-15T15:35:07Z attila_lendvai joined #sbcl 2014-04-15T15:35:07Z attila_lendvai quit (Changing host) 2014-04-15T15:35:07Z attila_lendvai joined #sbcl 2014-04-15T15:47:01Z tomterl quit (Ping timeout: 250 seconds) 2014-04-15T16:01:11Z sdemarre joined #sbcl 2014-04-15T16:04:00Z davazp quit (Remote host closed the connection) 2014-04-15T16:24:59Z fiveop joined #sbcl 2014-04-15T16:27:29Z eudoxia quit (Quit: leaving) 2014-04-15T16:39:20Z attila_lendvai quit (Quit: Leaving.) 2014-04-15T16:51:59Z ltbarcly joined #sbcl 2014-04-15T17:04:03Z ltbarcly quit (Ping timeout: 258 seconds) 2014-04-15T17:06:23Z sdemarre quit (Ping timeout: 276 seconds) 2014-04-15T17:07:01Z ltbarcly joined #sbcl 2014-04-15T17:22:25Z jaimef quit (Excess Flood) 2014-04-15T17:22:32Z jaimef joined #sbcl 2014-04-15T17:22:35Z attila_lendvai joined #sbcl 2014-04-15T17:22:35Z attila_lendvai quit (Changing host) 2014-04-15T17:22:35Z attila_lendvai joined #sbcl 2014-04-15T18:12:22Z drmeister quit (Remote host closed the connection) 2014-04-15T18:35:11Z attila_lendvai quit (Quit: Leaving.) 2014-04-15T19:41:10Z nyef: Oh, lovely. Floating-point load/store instructions use an offset scaled in words, so I can't use them to untag a boxed pointer. /-: 2014-04-15T19:42:33Z nyef: Means I have to burn an interior-reg or up to two unboxed-regs. 2014-04-15T19:45:53Z ASau joined #sbcl 2014-04-15T19:46:43Z scymtym_ joined #sbcl 2014-04-15T19:49:20Z drmeister joined #sbcl 2014-04-15T20:06:01Z angavrilov quit (Remote host closed the connection) 2014-04-15T20:15:05Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-15T20:36:23Z eudoxia joined #sbcl 2014-04-15T20:39:35Z Bicyclidine joined #sbcl 2014-04-15T20:43:38Z ltbarcly joined #sbcl 2014-04-15T20:46:38Z nyef quit (Quit: G'night lal.) 2014-04-15T21:11:57Z yacks quit (Quit: Leaving) 2014-04-15T21:23:38Z fiveop quit (Remote host closed the connection) 2014-04-15T22:00:52Z drmeister quit (Remote host closed the connection) 2014-04-15T22:01:44Z michael_lee joined #sbcl 2014-04-15T22:02:29Z davazp joined #sbcl 2014-04-15T22:19:39Z drmeister joined #sbcl 2014-04-15T22:44:28Z segv- quit (Ping timeout: 245 seconds) 2014-04-15T23:03:40Z drmeister quit (Ping timeout: 240 seconds) 2014-04-15T23:08:11Z fridim_ quit (Ping timeout: 252 seconds) 2014-04-15T23:12:16Z michael_lee quit (Quit: Ex-Chat) 2014-04-15T23:27:34Z eudoxia quit (Quit: leaving) 2014-04-15T23:32:17Z ltbarcly quit (Read error: Connection reset by peer) 2014-04-15T23:32:24Z ltbarcly_ joined #sbcl 2014-04-15T23:34:22Z nyef joined #sbcl 2014-04-15T23:35:16Z nyef: So, fun and games: I just got an error about there being no :MOVE-ARG VOP to go from a SINGLE-REG to a DESCRIPTOR-REG. 2014-04-15T23:35:34Z zRecursive joined #sbcl 2014-04-15T23:36:34Z nyef: I'm willing to acknowledge that I've screwed up somewhere, somehow, but I don't see any evidence that such a VOP would be required. 2014-04-15T23:41:24Z ltbarcly_ quit (Quit: Computer has gone to sleep.) 2014-04-15T23:41:40Z drmeister joined #sbcl 2014-04-16T00:06:56Z nyef: Ugh. Found the evidence for what I was missing. A completely uncommented DEFINE-MOVE-VOP in each backend that sets things up. 2014-04-16T00:20:52Z nyef: Clearly, I need to add an explanatory comment and declare the VOP to be its own toplevel section. 2014-04-16T00:58:52Z Bicyclidine quit (Quit: leaving) 2014-04-16T01:22:01Z ASau quit (Ping timeout: 265 seconds) 2014-04-16T01:35:59Z prxq quit (Ping timeout: 276 seconds) 2014-04-16T01:44:09Z ASau joined #sbcl 2014-04-16T01:48:01Z prxq joined #sbcl 2014-04-16T02:09:33Z ltbarcly joined #sbcl 2014-04-16T02:31:41Z scymtym_ quit (Ping timeout: 252 seconds) 2014-04-16T02:32:52Z nyef quit (Quit: G'night all.) 2014-04-16T02:37:49Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-16T02:39:26Z christoph_debian quit (Ping timeout: 258 seconds) 2014-04-16T02:48:31Z ltbarcly joined #sbcl 2014-04-16T02:53:37Z christoph_debian joined #sbcl 2014-04-16T02:54:00Z oleo quit (Ping timeout: 258 seconds) 2014-04-16T02:54:55Z oleo joined #sbcl 2014-04-16T02:55:48Z davazp quit (Remote host closed the connection) 2014-04-16T03:19:11Z ltbarcly_ joined #sbcl 2014-04-16T03:21:33Z ltbarcly quit (Ping timeout: 252 seconds) 2014-04-16T03:23:28Z ltbarcly joined #sbcl 2014-04-16T03:25:40Z ltbarcly_ quit (Ping timeout: 240 seconds) 2014-04-16T03:36:21Z ltbarcly_ joined #sbcl 2014-04-16T03:36:57Z ltbarcly quit (Ping timeout: 252 seconds) 2014-04-16T03:37:26Z ltbarcly_ quit (Read error: Connection reset by peer) 2014-04-16T03:37:43Z ltbarcly joined #sbcl 2014-04-16T03:39:10Z ltbarcly quit (Read error: Connection reset by peer) 2014-04-16T03:39:57Z ltbarcly joined #sbcl 2014-04-16T03:40:23Z attila_lendvai joined #sbcl 2014-04-16T03:40:23Z attila_lendvai quit (Changing host) 2014-04-16T03:40:23Z attila_lendvai joined #sbcl 2014-04-16T04:01:34Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-16T04:31:17Z yacks joined #sbcl 2014-04-16T05:06:27Z oleo quit (Remote host closed the connection) 2014-04-16T05:36:48Z drmeister quit (Remote host closed the connection) 2014-04-16T06:04:14Z pranavrc joined #sbcl 2014-04-16T06:04:14Z pranavrc quit (Changing host) 2014-04-16T06:04:14Z pranavrc joined #sbcl 2014-04-16T06:52:28Z hlavaty quit (Ping timeout: 240 seconds) 2014-04-16T07:25:38Z drmeister joined #sbcl 2014-04-16T07:27:08Z zRecursive quit (Remote host closed the connection) 2014-04-16T07:29:53Z drmeister quit (Ping timeout: 245 seconds) 2014-04-16T07:34:26Z angavrilov joined #sbcl 2014-04-16T07:39:27Z ltbarcly joined #sbcl 2014-04-16T07:39:28Z yacks quit (Ping timeout: 245 seconds) 2014-04-16T07:55:24Z tidux joined #sbcl 2014-04-16T07:55:54Z tidux: is there any chance of this ARM port being upstreamed? 2014-04-16T07:55:56Z tidux: https://github.com/fiveop/sbcl-arm 2014-04-16T07:56:16Z Krystof: as soon as it executes more than about three forms of lisp code 2014-04-16T07:57:28Z tidux: my only ARM system at the moment only has 512MB RAM so SBCL wouldn't even build 2014-04-16T07:57:30Z tidux: guess it's not ready 2014-04-16T07:57:40Z tidux: thanks for the quick response 2014-04-16T07:57:42Z tidux left #sbcl 2014-04-16T08:07:48Z yacks joined #sbcl 2014-04-16T08:39:08Z ASau quit (Ping timeout: 276 seconds) 2014-04-16T09:14:27Z drmeister joined #sbcl 2014-04-16T09:19:09Z drmeister quit (Ping timeout: 250 seconds) 2014-04-16T09:33:13Z kludge` quit (Ping timeout: 245 seconds) 2014-04-16T09:36:03Z kludge` joined #sbcl 2014-04-16T09:55:54Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-16T10:00:12Z ltbarcly joined #sbcl 2014-04-16T10:32:15Z ltbarcly quit (Quit: Textual IRC Client: www.textualapp.com) 2014-04-16T10:37:20Z attila_lendvai quit (Quit: Leaving.) 2014-04-16T11:03:17Z drmeister joined #sbcl 2014-04-16T11:07:55Z drmeister quit (Ping timeout: 250 seconds) 2014-04-16T11:26:53Z fridim_ joined #sbcl 2014-04-16T11:31:43Z segv- joined #sbcl 2014-04-16T12:10:18Z eudoxia joined #sbcl 2014-04-16T12:52:35Z pranavrc quit 2014-04-16T13:16:06Z drmeister joined #sbcl 2014-04-16T13:23:16Z drmeister quit (Ping timeout: 240 seconds) 2014-04-16T13:28:22Z oleo joined #sbcl 2014-04-16T13:41:23Z drmeister joined #sbcl 2014-04-16T13:42:08Z nyef joined #sbcl 2014-04-16T13:42:16Z nyef: G'morning all. 2014-04-16T13:42:46Z oleo: morning 2014-04-16T14:06:33Z drmeister quit (Ping timeout: 245 seconds) 2014-04-16T14:11:55Z drmeister joined #sbcl 2014-04-16T14:20:22Z fridim_ quit (Ping timeout: 276 seconds) 2014-04-16T14:27:29Z nyef: Okay, object is not of type SIMPLE-VECTOR from %MAKE-HASH-TABLE means that the :TABLE argument is wrong. 2014-04-16T14:27:58Z nyef: That it's a LIST (34 . %EMPTY-HT-SLOT%) also implies something. 2014-04-16T14:30:12Z nyef: Misaligned allocation pointer? 2014-04-16T14:37:38Z nyef: Yeah, misaligned allocation pointer. Ouch. 2014-04-16T14:54:27Z fridim_ joined #sbcl 2014-04-16T15:06:52Z oleo quit (Ping timeout: 240 seconds) 2014-04-16T15:14:49Z ltbarcly joined #sbcl 2014-04-16T15:15:43Z fridim_ quit (Ping timeout: 245 seconds) 2014-04-16T15:20:28Z oleo joined #sbcl 2014-04-16T15:26:19Z Krystof: but getting that far means there must be semi-working floats 2014-04-16T15:26:34Z nyef: Yup! 2014-04-16T15:27:07Z nyef: And I have a possible lead on the SIGSEGV from SIGTRAP issue that I'd been having. 2014-04-16T15:27:44Z nyef: (Which might require that I re-work the P-A traps again, unfortunately.) 2014-04-16T15:33:38Z nyef: Hrm. "/entering ERROR, argument list=..", repeated until the control stack is blown. 2014-04-16T15:34:28Z hlavaty joined #sbcl 2014-04-16T15:54:08Z attila_lendvai joined #sbcl 2014-04-16T15:54:08Z attila_lendvai quit (Changing host) 2014-04-16T15:54:08Z attila_lendvai joined #sbcl 2014-04-16T16:12:40Z attila_lendvai quit (Quit: Leaving.) 2014-04-16T16:15:39Z attila_lendvai joined #sbcl 2014-04-16T16:15:39Z attila_lendvai quit (Changing host) 2014-04-16T16:15:39Z attila_lendvai joined #sbcl 2014-04-16T16:20:12Z attila_lendvai1 joined #sbcl 2014-04-16T16:20:12Z attila_lendvai quit (Disconnected by services) 2014-04-16T16:20:12Z attila_lendvai1 quit (Changing host) 2014-04-16T16:20:12Z attila_lendvai1 joined #sbcl 2014-04-16T16:26:17Z drmeister quit (Ping timeout: 258 seconds) 2014-04-16T16:49:28Z slyrus quit (Ping timeout: 245 seconds) 2014-04-16T16:59:16Z kludge` quit (Ping timeout: 240 seconds) 2014-04-16T17:05:25Z kludge` joined #sbcl 2014-04-16T17:29:44Z eudoxia quit (Quit: leaving) 2014-04-16T17:44:34Z ltbarcly quit (Read error: Connection reset by peer) 2014-04-16T17:45:10Z ltbarcly joined #sbcl 2014-04-16T18:03:31Z jdz joined #sbcl 2014-04-16T18:05:07Z nyef: Ugh. SIMPLE-TYPE-ERROR :DATUM # ... 2014-04-16T18:09:50Z nyef: ... Immediately after single-word BIGNUM 0x004c3491 ? Something isn't quite right... 2014-04-16T18:14:32Z angavrilov quit (Remote host closed the connection) 2014-04-16T18:20:11Z attila_lendvai joined #sbcl 2014-04-16T18:20:11Z attila_lendvai quit (Changing host) 2014-04-16T18:20:11Z attila_lendvai joined #sbcl 2014-04-16T18:24:07Z attila_lendvai1 quit (Ping timeout: 276 seconds) 2014-04-16T18:26:18Z slyrus joined #sbcl 2014-04-16T18:47:52Z jdz quit (Remote host closed the connection) 2014-04-16T18:48:24Z jdz joined #sbcl 2014-04-16T19:18:07Z DGASAU quit (Remote host closed the connection) 2014-04-16T19:19:32Z DGASAU joined #sbcl 2014-04-16T19:39:11Z drmeister joined #sbcl 2014-04-16T20:28:45Z Posterdati quit (Excess Flood) 2014-04-16T20:29:08Z Posterdati joined #sbcl 2014-04-16T20:34:48Z eudoxia joined #sbcl 2014-04-16T20:35:13Z nyef quit (Quit: Bye all, I'll be back in a bit.) 2014-04-16T21:00:29Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-16T21:23:53Z ltbarcly joined #sbcl 2014-04-16T21:30:29Z fridim_ joined #sbcl 2014-04-16T21:32:04Z jdz quit (Quit: Leaving...) 2014-04-16T21:32:27Z nyef joined #sbcl 2014-04-16T21:44:12Z attila_lendvai quit (Quit: Leaving.) 2014-04-16T22:17:11Z drmeister quit (Remote host closed the connection) 2014-04-16T22:23:02Z drmeister joined #sbcl 2014-04-16T22:24:24Z drmeister quit (Remote host closed the connection) 2014-04-16T22:25:52Z ASau joined #sbcl 2014-04-16T22:31:15Z segv- quit (Ping timeout: 265 seconds) 2014-04-16T22:34:51Z drmeister joined #sbcl 2014-04-16T22:36:11Z ltbarcly quit (Ping timeout: 252 seconds) 2014-04-16T22:56:08Z fridim_ quit (Ping timeout: 258 seconds) 2014-04-16T23:06:13Z ASau quit (Remote host closed the connection) 2014-04-16T23:06:48Z ASau joined #sbcl 2014-04-16T23:07:31Z drmeister quit (Ping timeout: 276 seconds) 2014-04-16T23:17:16Z drmeister joined #sbcl 2014-04-16T23:20:57Z ASau quit (Remote host closed the connection) 2014-04-16T23:22:54Z ASau joined #sbcl 2014-04-16T23:37:34Z ASau quit (Remote host closed the connection) 2014-04-16T23:38:11Z ASau joined #sbcl 2014-04-16T23:41:39Z zRecursive joined #sbcl 2014-04-16T23:51:20Z ASau quit (Remote host closed the connection) 2014-04-16T23:51:54Z ASau joined #sbcl 2014-04-17T00:41:16Z eudoxia quit (Quit: leaving) 2014-04-17T00:42:27Z nyef: Damned GDB. If I tell it to pass SIGTRAP, the program under test terminates basically immediately. 2014-04-17T00:44:28Z nyef: In fact, if I mess with the SIGTRAP handling at all, it becomes useless. 2014-04-17T01:12:23Z slyrus quit (Ping timeout: 252 seconds) 2014-04-17T01:22:06Z ams quit (Ping timeout: 240 seconds) 2014-04-17T01:32:33Z prxq_ joined #sbcl 2014-04-17T01:35:29Z prxq quit (Ping timeout: 252 seconds) 2014-04-17T02:01:31Z pchrist quit (Ping timeout: 252 seconds) 2014-04-17T02:03:50Z pchrist joined #sbcl 2014-04-17T02:07:40Z irsol quit (Ping timeout: 240 seconds) 2014-04-17T02:15:26Z irsol joined #sbcl 2014-04-17T02:33:37Z zRecursive quit (Read error: Connection reset by peer) 2014-04-17T02:34:04Z zRecursive joined #sbcl 2014-04-17T02:38:43Z christoph_debian quit (Ping timeout: 250 seconds) 2014-04-17T02:46:49Z nyef quit (Quit: G'night all.) 2014-04-17T02:48:57Z ehaliewicz joined #sbcl 2014-04-17T02:52:23Z christoph_debian joined #sbcl 2014-04-17T03:03:52Z faheem quit (Read error: Connection reset by peer) 2014-04-17T03:16:03Z faheem joined #sbcl 2014-04-17T03:24:52Z oleo quit (Ping timeout: 240 seconds) 2014-04-17T03:38:03Z oleo joined #sbcl 2014-04-17T03:52:28Z drmeister quit (Ping timeout: 240 seconds) 2014-04-17T04:47:13Z Bike quit (Quit: leaving) 2014-04-17T04:47:34Z Bike joined #sbcl 2014-04-17T04:57:15Z krzysz00 joined #sbcl 2014-04-17T05:00:59Z kvsari joined #sbcl 2014-04-17T05:01:04Z kvsari left #sbcl 2014-04-17T05:06:55Z oleo quit (Quit: Leaving) 2014-04-17T05:36:55Z pranavrc joined #sbcl 2014-04-17T05:36:55Z pranavrc quit (Changing host) 2014-04-17T05:36:55Z pranavrc joined #sbcl 2014-04-17T05:49:50Z attila_lendvai joined #sbcl 2014-04-17T05:49:50Z attila_lendvai quit (Changing host) 2014-04-17T05:49:50Z attila_lendvai joined #sbcl 2014-04-17T05:54:57Z slyrus joined #sbcl 2014-04-17T06:03:56Z krzysz00 quit (Ping timeout: 258 seconds) 2014-04-17T06:26:27Z angavrilov joined #sbcl 2014-04-17T06:27:30Z prxq_ is now known as prxq 2014-04-17T07:27:22Z zRecursive quit (Remote host closed the connection) 2014-04-17T07:27:45Z redline6561 quit (Ping timeout: 250 seconds) 2014-04-17T07:28:41Z redline6561 joined #sbcl 2014-04-17T07:35:31Z fridim_ joined #sbcl 2014-04-17T07:45:31Z ASau quit (Ping timeout: 250 seconds) 2014-04-17T08:05:04Z fridim_ quit (Ping timeout: 258 seconds) 2014-04-17T08:41:29Z ehaliewicz quit (Remote host closed the connection) 2014-04-17T08:57:11Z pranavrc quit (Remote host closed the connection) 2014-04-17T09:01:21Z ams joined #sbcl 2014-04-17T09:23:31Z pranavrc joined #sbcl 2014-04-17T09:23:31Z pranavrc quit (Changing host) 2014-04-17T09:23:32Z pranavrc joined #sbcl 2014-04-17T09:30:43Z kludge` quit (Ping timeout: 245 seconds) 2014-04-17T09:36:09Z kludge` joined #sbcl 2014-04-17T09:37:24Z DGASAU quit (Remote host closed the connection) 2014-04-17T09:38:36Z fridim_ joined #sbcl 2014-04-17T10:06:55Z DGASAU joined #sbcl 2014-04-17T10:27:23Z xymox quit (Ping timeout: 245 seconds) 2014-04-17T10:27:39Z rudi_ joined #sbcl 2014-04-17T10:27:58Z rudi_ quit (Client Quit) 2014-04-17T10:28:16Z rudi_ joined #sbcl 2014-04-17T10:28:43Z xymox joined #sbcl 2014-04-17T10:28:59Z rudi_ is now known as rudi 2014-04-17T10:35:08Z rudi quit (Quit: Client exciting.) 2014-04-17T10:39:17Z xymox quit (Quit: leaving) 2014-04-17T10:46:11Z xymox joined #sbcl 2014-04-17T11:00:49Z krzysz00 joined #sbcl 2014-04-17T11:14:14Z pchrist quit (Quit: leaving) 2014-04-17T11:14:57Z pchrist joined #sbcl 2014-04-17T12:47:42Z davazp joined #sbcl 2014-04-17T12:47:58Z krzysz00 quit (Ping timeout: 258 seconds) 2014-04-17T13:00:26Z davazp quit (Remote host closed the connection) 2014-04-17T13:00:34Z pranavrc quit 2014-04-17T13:08:19Z eudoxia joined #sbcl 2014-04-17T13:34:37Z attila_lendvai quit (Quit: Leaving.) 2014-04-17T13:41:06Z oleo joined #sbcl 2014-04-17T13:41:48Z segv- joined #sbcl 2014-04-17T13:52:27Z nyef joined #sbcl 2014-04-17T13:52:35Z nyef: G'morning all. 2014-04-17T13:58:16Z oleo: morning 2014-04-17T15:25:53Z slyrus quit (Remote host closed the connection) 2014-04-17T15:31:26Z slyrus joined #sbcl 2014-04-17T15:34:29Z DGASAU` joined #sbcl 2014-04-17T15:35:52Z DGASAU quit (Ping timeout: 258 seconds) 2014-04-17T15:36:42Z DGASAU` is now known as DGASAU 2014-04-17T15:38:07Z oleo quit (Ping timeout: 276 seconds) 2014-04-17T15:45:34Z michael_lee joined #sbcl 2014-04-17T15:47:52Z michael_lee quit (Max SendQ exceeded) 2014-04-17T15:49:02Z michael_lee joined #sbcl 2014-04-17T15:50:35Z oleo joined #sbcl 2014-04-17T15:53:25Z eudoxia quit (Quit: leaving) 2014-04-17T15:54:51Z attila_lendvai joined #sbcl 2014-04-17T15:54:51Z attila_lendvai quit (Changing host) 2014-04-17T15:54:51Z attila_lendvai joined #sbcl 2014-04-17T16:25:58Z hlavaty quit (Remote host closed the connection) 2014-04-17T17:01:55Z ltbarcly joined #sbcl 2014-04-17T17:19:23Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-17T17:38:22Z attila_lendvai quit (Quit: Leaving.) 2014-04-17T17:47:41Z eudoxia joined #sbcl 2014-04-17T17:52:42Z attila_lendvai joined #sbcl 2014-04-17T17:52:42Z attila_lendvai quit (Changing host) 2014-04-17T17:52:42Z attila_lendvai joined #sbcl 2014-04-17T18:05:18Z ASau joined #sbcl 2014-04-17T18:18:01Z attila_lendvai quit (Ping timeout: 276 seconds) 2014-04-17T18:21:10Z michael_lee quit (Quit: Ex-Chat) 2014-04-17T18:30:03Z ltbarcly joined #sbcl 2014-04-17T18:37:03Z cmack joined #sbcl 2014-04-17T18:45:51Z angavrilov quit (Remote host closed the connection) 2014-04-17T18:57:23Z eudoxia quit (Quit: leaving) 2014-04-17T19:01:31Z fiveop joined #sbcl 2014-04-17T19:03:47Z davazp joined #sbcl 2014-04-17T19:34:35Z scymtym_ joined #sbcl 2014-04-17T19:40:34Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-17T19:43:53Z slyrus quit (Ping timeout: 258 seconds) 2014-04-17T19:59:44Z slyrus joined #sbcl 2014-04-17T20:01:52Z krzysz00 joined #sbcl 2014-04-17T20:08:25Z davazp quit (Ping timeout: 258 seconds) 2014-04-17T20:17:12Z davazp joined #sbcl 2014-04-17T20:31:35Z nyef quit (Quit: Offline for a bit.) 2014-04-17T20:43:32Z eudoxia joined #sbcl 2014-04-17T20:47:17Z ltbarcly joined #sbcl 2014-04-17T21:02:50Z slyrus_ joined #sbcl 2014-04-17T21:04:28Z slyrus quit (Ping timeout: 245 seconds) 2014-04-17T21:04:30Z slyrus_ is now known as slyrus 2014-04-17T21:11:41Z fikusz quit (Quit: Leaving) 2014-04-17T21:18:13Z fiveop quit (Quit: bye) 2014-04-17T21:18:43Z fikusz joined #sbcl 2014-04-17T21:23:12Z slyrus_ joined #sbcl 2014-04-17T21:24:35Z slyrus quit (Ping timeout: 252 seconds) 2014-04-17T21:24:36Z slyrus_ is now known as slyrus 2014-04-17T21:24:49Z yacks quit (Quit: Leaving) 2014-04-17T21:32:42Z prxq quit (Quit: Leaving) 2014-04-17T21:45:29Z eudoxia quit (Quit: leaving) 2014-04-17T21:45:30Z drmeister joined #sbcl 2014-04-17T21:46:04Z slyrus quit (Ping timeout: 240 seconds) 2014-04-17T21:49:29Z ehaliewicz joined #sbcl 2014-04-17T21:55:03Z slyrus joined #sbcl 2014-04-17T22:00:35Z drmeister quit (Read error: Connection reset by peer) 2014-04-17T22:08:21Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-17T22:10:00Z krzysz00 quit (Quit: Lost terminal) 2014-04-17T22:10:21Z drmeister joined #sbcl 2014-04-17T22:21:44Z cmack` joined #sbcl 2014-04-17T22:23:51Z cmack quit (Ping timeout: 265 seconds) 2014-04-17T22:34:26Z cmack`` joined #sbcl 2014-04-17T22:36:25Z cmack` quit (Ping timeout: 265 seconds) 2014-04-17T22:49:44Z drmeister quit (Remote host closed the connection) 2014-04-17T22:51:58Z drmeister joined #sbcl 2014-04-17T22:52:35Z segv- quit (Ping timeout: 252 seconds) 2014-04-17T22:57:13Z nyef joined #sbcl 2014-04-17T22:57:25Z nyef: Hello all. 2014-04-17T23:24:05Z slyrus: hey nyef 2014-04-17T23:25:48Z reb: Hey nyef, are you by chance the person who wrote chipz? 2014-04-17T23:26:22Z slyrus: that's xach, right? 2014-04-17T23:26:56Z drmeister quit (Remote host closed the connection) 2014-04-17T23:29:21Z nyef: The name "chipz" rings a vague bell, but I'm fairly sure it's not one of my projects. 2014-04-17T23:30:36Z slyrus: chipz 'n salza I think are xach's 2014-04-17T23:30:50Z ltbarcly joined #sbcl 2014-04-17T23:31:03Z nyef: Okay, that DOES sound like something that Xach would do, and ISTR salza being his. 2014-04-17T23:31:30Z nyef: ... there was also the joke about him going into business with water... 2014-04-17T23:32:41Z reb: Thanks! 2014-04-17T23:36:26Z nyef: reb: Do I remember rightly that you have something to do with ccl? 2014-04-17T23:36:39Z nyef: Or am I mixing you up with someone else? 2014-04-17T23:44:05Z fridim_ quit (Ping timeout: 265 seconds) 2014-04-17T23:45:44Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-17T23:48:21Z ltbarcly joined #sbcl 2014-04-17T23:49:44Z zRecursive joined #sbcl 2014-04-18T00:00:47Z pkhuong: nyef: rme 2014-04-18T00:00:59Z nyef: Hello pkhuong. 2014-04-18T00:05:00Z reb: nyef: I use ccl at home on my Raspberry Pi. 2014-04-18T00:05:06Z reb: ... I'm just a user. 2014-04-18T00:05:21Z nyef: Ah, okay. 2014-04-18T00:08:10Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-18T00:15:36Z cmack``` joined #sbcl 2014-04-18T00:15:40Z ltbarcly joined #sbcl 2014-04-18T00:17:25Z cmack`` quit (Ping timeout: 250 seconds) 2014-04-18T00:21:47Z drmeister joined #sbcl 2014-04-18T00:25:04Z ehaliewicz quit (Remote host closed the connection) 2014-04-18T00:31:03Z nyef: Hrm. I'm getting some sort of simple-type-error where the datum is an other-pointer with header 0xfffffffa. 2014-04-18T00:31:38Z nyef: I'm somewhat at a loss at this point. 2014-04-18T00:43:11Z slyrus: hmm... that 0xfffffffa rings a bell 2014-04-18T00:44:03Z nyef: Something to do with either a hash table problem or a bignum problem, I hope? 2014-04-18T00:46:55Z nyef: This is an improvement over the header 0x00000102 that I ran into at one point. 2014-04-18T00:59:18Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-18T01:58:32Z nyef quit (Quit: G'night all.) 2014-04-18T02:11:34Z Hydan_ joined #sbcl 2014-04-18T02:32:43Z Hydan_ is now known as Hydan` 2014-04-18T02:39:03Z christoph_debian quit (Ping timeout: 265 seconds) 2014-04-18T02:46:02Z krzysz00 joined #sbcl 2014-04-18T02:52:12Z christoph_debian joined #sbcl 2014-04-18T02:57:50Z Hydan` is now known as Hydan_ 2014-04-18T03:01:20Z attila_lendvai joined #sbcl 2014-04-18T03:01:20Z attila_lendvai quit (Changing host) 2014-04-18T03:01:20Z attila_lendvai joined #sbcl 2014-04-18T03:23:43Z ltbarcly joined #sbcl 2014-04-18T03:28:03Z ltbarcly quit (Client Quit) 2014-04-18T03:28:45Z yacks joined #sbcl 2014-04-18T03:29:24Z davazp quit (Remote host closed the connection) 2014-04-18T03:34:36Z drmeister quit (Quit: Leaving...) 2014-04-18T03:41:30Z drmeister joined #sbcl 2014-04-18T04:22:55Z oleo quit (Ping timeout: 258 seconds) 2014-04-18T04:24:49Z oleo joined #sbcl 2014-04-18T04:26:10Z Hydan_ is now known as Hydan` 2014-04-18T04:39:29Z Hydan` is now known as Hydan_ 2014-04-18T04:40:30Z Hydan_ is now known as Hydan` 2014-04-18T04:40:48Z Hydan` is now known as Hydan_ 2014-04-18T05:09:12Z yacks quit (Read error: Connection reset by peer) 2014-04-18T05:19:43Z scymtym_ quit (Ping timeout: 276 seconds) 2014-04-18T05:31:44Z pranavrc joined #sbcl 2014-04-18T05:31:44Z pranavrc quit (Changing host) 2014-04-18T05:31:44Z pranavrc joined #sbcl 2014-04-18T05:41:05Z Hydan_ quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-18T05:47:50Z yacks joined #sbcl 2014-04-18T05:57:26Z drmeister quit (Remote host closed the connection) 2014-04-18T05:59:05Z angavrilov joined #sbcl 2014-04-18T06:04:18Z krzysz00 quit (Ping timeout: 240 seconds) 2014-04-18T06:26:02Z fridim_ joined #sbcl 2014-04-18T06:33:38Z fridim_ quit (Ping timeout: 245 seconds) 2014-04-18T06:36:33Z ehaliewicz joined #sbcl 2014-04-18T07:17:47Z segv- joined #sbcl 2014-04-18T07:21:37Z zRecursive left #sbcl 2014-04-18T07:46:30Z drmeister joined #sbcl 2014-04-18T07:51:17Z drmeister quit (Ping timeout: 265 seconds) 2014-04-18T09:21:11Z fridim_ joined #sbcl 2014-04-18T09:31:07Z kludge` quit (Ping timeout: 258 seconds) 2014-04-18T09:34:32Z Bike quit (Quit: leaving) 2014-04-18T09:35:12Z drmeister joined #sbcl 2014-04-18T09:35:52Z kludge` joined #sbcl 2014-04-18T09:36:25Z ehaliewi` joined #sbcl 2014-04-18T09:39:15Z ehaliewicz quit (Ping timeout: 240 seconds) 2014-04-18T09:39:56Z drmeister quit (Ping timeout: 258 seconds) 2014-04-18T09:56:59Z ehaliewi` quit (Ping timeout: 252 seconds) 2014-04-18T10:18:36Z attila_lendvai quit (Quit: Leaving.) 2014-04-18T10:37:13Z ASau` joined #sbcl 2014-04-18T10:40:53Z ASau quit (Ping timeout: 258 seconds) 2014-04-18T11:01:18Z krzysz00 joined #sbcl 2014-04-18T11:19:11Z ASau` is now known as ASau 2014-04-18T11:21:15Z yacks quit (Ping timeout: 240 seconds) 2014-04-18T11:23:48Z drmeister joined #sbcl 2014-04-18T11:28:47Z drmeister quit (Ping timeout: 265 seconds) 2014-04-18T11:47:12Z jdz joined #sbcl 2014-04-18T11:48:32Z yacks joined #sbcl 2014-04-18T11:53:02Z drmeister joined #sbcl 2014-04-18T12:26:35Z cmack``` is now known as cmack 2014-04-18T13:01:56Z pranavrc quit 2014-04-18T13:28:03Z nyef joined #sbcl 2014-04-18T13:28:15Z nyef: G'morning all. 2014-04-18T13:30:44Z slyrus: morning nyef 2014-04-18T13:32:37Z yacks quit (Ping timeout: 258 seconds) 2014-04-18T13:47:48Z fridim_ quit (Ping timeout: 245 seconds) 2014-04-18T13:53:26Z Krystof: morning 2014-04-18T13:57:06Z nyef: So, I'm still at a loss with regards to the odd type error that I've been getting where the header is 0xfffffffa. 2014-04-18T14:07:08Z jdz_ joined #sbcl 2014-04-18T14:07:53Z jdz quit (Read error: Connection reset by peer) 2014-04-18T14:07:54Z jdz_ quit (Read error: Connection reset by peer) 2014-04-18T14:09:18Z jdz joined #sbcl 2014-04-18T14:10:07Z Krystof: is that the inverse or the negative of something interesting? 2014-04-18T14:10:19Z Krystof: is there something near it in memory? 2014-04-18T14:10:59Z nyef: It seems to be in the middle of a bunch of bignums, IIRC. I don't actually have the system set up right now. 2014-04-18T14:12:06Z jdz_ joined #sbcl 2014-04-18T14:12:33Z yacks joined #sbcl 2014-04-18T14:15:03Z jdz quit (Ping timeout: 265 seconds) 2014-04-18T14:16:16Z stassats joined #sbcl 2014-04-18T14:18:28Z fiveop joined #sbcl 2014-04-18T14:18:36Z nyef: http://paste.lisp.org/display/142205 2014-04-18T14:19:43Z nyef: Ignore the CORRUPTION WARNING for now, it's something to do with using fprintf() or vfprintf() with actual parameters from within a SIGTRAP handler, which I have yet to figure out. 2014-04-18T14:20:25Z fiveop: Nice 2014-04-18T14:23:24Z nyef: And just pushed my current changes. My working tree also contains static-funs for two-arg-+ and two-arg-and, disabling compilation of ERROR (so that I get an undefined fun trap instead), and an experiment in switching the form of the breakpoint used to generate SIGTRAP (which didn't help the vfprintf() issue, but could be useful elsewhere anyway). 2014-04-18T14:26:26Z stassats: #xfffffffa could be a sign extended high part of a bignum 2014-04-18T14:28:39Z nyef: Annotated with backtrace. 2014-04-18T14:29:20Z fiveop: I have a few things to do, organizing my wedding, then I should finally have time to pull :) 2014-04-18T14:30:05Z Krystof: what's lowtag a? Bignum? 2014-04-18T14:30:10Z nyef: fiveop: Sounds like congratulations are in order, then. 2014-04-18T14:30:32Z nyef: Yes, bignum. 2014-04-18T14:30:40Z stassats: bignum #xfffffffa #x31da0170 would be -24933432976 2014-04-18T14:30:45Z Krystof: it looks to me like #xfffffffa "ought to be" #x0000010a 2014-04-18T14:31:10Z Krystof: (compare it with the four words above beginning at address 0x5152F4E0) 2014-04-18T14:32:14Z nyef: Hrm. So, a bignum allocated at two or three digits, then shrunk to one, and the shrink gets blown on one of the later instances? 2014-04-18T14:32:31Z Krystof: there are some other odd bits around there: lots of ".....20a" with only one word following 2014-04-18T14:32:47Z Krystof: e.g. at 0x5152F4A0 2014-04-18T14:32:57Z nyef: Mmm. 2014-04-18T14:33:22Z stassats: bignum lowtag? shouldn't it be a widetag? 2014-04-18T14:33:41Z nyef: Actually, nothing in memory makes any kind of sense after about 5152F4F0. 2014-04-18T14:33:59Z nyef: Oh, unless it's CONS allocation. 2014-04-18T14:34:28Z nyef: And it is CONS allocation, the list structure is obvious now that I'm looking for it. 2014-04-18T14:35:02Z Krystof: right 2014-04-18T14:35:44Z Krystof: but anyway I'd be suspicious of bignum code 2014-04-18T14:36:28Z nyef: I was afraid of that. I've already desk-checked (and in several cases, fixed) the bignum VOPs. 2014-04-18T14:36:43Z Krystof: I'm not sure how to explain the fffffffa (I'd have expected ffffff0a) 2014-04-18T14:37:05Z stassats: maybe it's a most-positive-bignum? overflown size 2014-04-18T14:37:11Z Krystof: but the "20a"s look very wrong 2014-04-18T14:37:23Z nyef: Also went over the arithmetic conditional VOPs, and those SEEM right. 2014-04-18T14:37:43Z stassats: 20a is a bignum of size 2, isn't it? 2014-04-18T14:37:57Z nyef: stassats: Yes. Two digits, plus a padding word. 2014-04-18T14:38:16Z nyef: (Plus the header, of course.) 2014-04-18T14:38:30Z Krystof: nyef: maybe construct some bignum arithmetic tests to run early in cold-init with /show decoration? 2014-04-18T14:39:15Z nyef: That might be what I end up doing. And that early in cold-init I can still use gdb. 2014-04-18T14:39:43Z stassats: at 0x5152F4A0, it says 0x0000020a, then 0x004c3491 follows, then 0x0000010a, which looks like the next bignum 2014-04-18T14:40:20Z nyef: stassats: Yes. Which suggests that it was originally a one-word bignum, possibly from MAKE-SMALL-BIGNUM that was then resized incorrectly. 2014-04-18T14:40:54Z stassats: bignums usually grow smaller, not larger 2014-04-18T14:41:10Z nyef: Hence "incorrectly". 2014-04-18T14:41:52Z stassats: 0x5152F478 starts a similar sequence 2014-04-18T14:42:00Z stassats: three bignums, 10a 20a 10a 2014-04-18T14:44:19Z Bike joined #sbcl 2014-04-18T14:45:59Z stassats: maybe something is wrong with the set-header-data VOP? 2014-04-18T14:46:01Z nyef: That list is (X NUMBER #), we're pre-package-init, so finding packages for the symbols isn't likely. 2014-04-18T14:47:13Z nyef: VOP source annotated. 2014-04-18T14:48:54Z nyef: Nothing too obviously wrong with it, AFAICT. 2014-04-18T14:52:18Z jdz_ quit (Ping timeout: 240 seconds) 2014-04-18T14:58:00Z Krystof: it looks plausible at least 2014-04-18T14:58:07Z Krystof: "to fool the maintenance programmer..>" 2014-04-18T14:58:58Z Krystof: I assume that your MOVE is move dest src 2014-04-18T14:59:21Z Krystof: and that storew is store src dest 2014-04-18T15:02:55Z Krystof: but yes, some bignum tests early enough that single-stepping through gdb would be what I would look at 2014-04-18T15:04:41Z stassats: maybe something due to not zeroed memory? 2014-04-18T15:04:47Z stassats: is it zeroed? 2014-04-18T15:16:22Z drmeister quit (Remote host closed the connection) 2014-04-18T15:18:26Z jdz joined #sbcl 2014-04-18T15:21:47Z nyef: stassats: Should be zeroed to start, yes. 2014-04-18T15:22:28Z stassats: is 0800000b NIL? 2014-04-18T15:22:35Z nyef: Yes, that's NIL. 2014-04-18T15:23:24Z drmeister joined #sbcl 2014-04-18T15:24:37Z stassats: why would %SXHASH-SIMPLE-STRING be adding numbers from a list? 2014-04-18T15:27:16Z jdz quit (Quit: Leaving...) 2014-04-18T15:30:04Z nyef: Where do you see it adding numbers from a list? 2014-04-18T15:30:28Z stassats: well, problematic number is in a list 2014-04-18T15:31:06Z stassats: and it breaks on TWO-ARG-+, so, that's what i assumed 2014-04-18T15:31:47Z nyef: And if that list is from TWO-ARG-+ putting together a SIMPLE-TYPE-ERROR? 2014-04-18T15:32:32Z stassats: ok 2014-04-18T15:32:32Z nyef: Yeah, that's from the error code generated by NUMBER-DISPATCH. 2014-04-18T15:33:35Z stassats: perhaps that's due to missing modular arithmetic? 2014-04-18T15:33:49Z stassats: the presence of bignums, that is 2014-04-18T15:34:11Z nyef: Perhaps, but mere presence of bignums shouldn't break things this way. 2014-04-18T15:34:55Z stassats: it would be intersting to know what string is being hashed 2014-04-18T15:34:59Z nyef: Is modular arithmetic supported by all backends? 2014-04-18T15:35:17Z krzysz00 quit (Ping timeout: 258 seconds) 2014-04-18T15:35:52Z stassats: should be, but you need some vops here and there and some there's some bookkeeping for what can be modularized 2014-04-18T15:36:36Z nyef: So it's plausibly not enabled for all backends. 2014-04-18T15:37:08Z stassats: define-modular-fun 2014-04-18T15:37:22Z stassats: x86-64 has more define-modular-fun than ppc 2014-04-18T15:38:47Z nyef: Hrm. Looks like there's a decent amount of coverage, even for HPPA. 2014-04-18T15:40:38Z ASau quit (Remote host closed the connection) 2014-04-18T15:41:08Z ASau joined #sbcl 2014-04-18T15:42:09Z stassats: so, clearly all those bignums are created during %sxhash-substring 2014-04-18T15:42:46Z stassats: you do not have alloc-bignum routines, how do you produce bignums? 2014-04-18T15:43:29Z stassats: just something in move? 2014-04-18T15:44:25Z nyef: The usual stuff in compiler/target/move. 2014-04-18T15:44:35Z stassats: i guess x86 are unusual 2014-04-18T15:44:48Z nyef: Yes, once again, it's the odd platform out. 2014-04-18T15:48:34Z stassats: is move-from-signed really that simple? 2014-04-18T15:49:32Z attila_lendvai joined #sbcl 2014-04-18T15:49:32Z attila_lendvai quit (Changing host) 2014-04-18T15:49:32Z attila_lendvai joined #sbcl 2014-04-18T15:49:41Z stassats: ok, it is, since it deals only with one word 2014-04-18T15:58:40Z stassats: nyef: in move-from-unsigned, shouldn't it be (storew PA-FLAG y 0 other-pointer-lowtag :pl) and not TEMP? 2014-04-18T15:59:57Z nyef: ... Yes. Yes, it should. Good catch, thank you. 2014-04-18T16:00:22Z fiveop: nyef: I'm currently comparing your code in arm/float.lisp against mine (unfortunately, I didnt commit those changes, and there is a lot of duplication). What I noticed is, that we differ in that you often multiply addresses by n-word-bytes, while I don't. If I understood it correctly, the float instructions take word aligned addresses. (I also don't use your @ macro/function, and didn't look at your instruction definitions yet, so I 2014-04-18T16:00:22Z fiveop: might miss something fundamental :)) 2014-04-18T16:01:17Z nyef: fiveop: Yeah, I declared that "we have a precedent for memory access that roughly mirrors what's in the ARM ARM, so we should follow it", and adjusted the instruction definitions to suit. 2014-04-18T16:01:45Z nyef: Also means that I can define a function to return a memory-operand and be able to use it as-is, which should help for number-stack access. 2014-04-18T16:04:22Z stassats: i should get me an arm 2014-04-18T16:04:41Z stassats: or just start porting sbcl to android 2014-04-18T16:05:20Z nyef: I figure porting to android can wait until we've got the basic system working. 2014-04-18T16:08:05Z fiveop: Our FLOAT-ROUND-TO- constants still differ. 2014-04-18T16:10:02Z nyef: Where did you get yours from? 2014-04-18T16:11:04Z ltbarcly joined #sbcl 2014-04-18T16:13:21Z fiveop: the 2005 ARM ARM file 2014-04-18T16:13:42Z ltbarcly_ joined #sbcl 2014-04-18T16:13:53Z nyef: Hrm. Well, might be a little further. http://paste.lisp.org/display/142205#3 2014-04-18T16:13:54Z fiveop: where it says "0b01 Indicates Round towards Plus Infinity (RP) mode" 2014-04-18T16:16:58Z ltbarcly quit (Ping timeout: 245 seconds) 2014-04-18T16:20:03Z nyef: I see it. You're right, my set of FLOAT-ROUND-TO- constants is wrong. 2014-04-18T16:23:22Z fiveop: done merging our changes, there is still some new stuff left, after I replaced most of it with your code ... :) 2014-04-18T16:23:52Z nyef: Let me guess... mostly complex float stuff, plus anything floaty for arrays, SAPs, or unboxed structure slots? 2014-04-18T16:26:55Z fiveop: stuff in arm/c-call and I removed a few more #-arm conditionals in various files, because stuff worked with the new float vops etc. 2014-04-18T16:27:20Z fiveop: Next up: get NFS setup, so that I can run it on my pi again :) 2014-04-18T16:30:37Z nyef: I just rsync up and down. 2014-04-18T16:33:46Z fiveop: When it's setup its very convenient. 2014-04-18T16:34:15Z fiveop: mount when start working, unmount when shutting down pi 2014-04-18T16:36:49Z stassats: i got tired of forgetting to unmount NFS when disconnecting/shutting down and everything hanging 2014-04-18T16:37:05Z fiveop: you shouldn't forget that 2014-04-18T16:37:18Z fiveop: but there probably is a way to configure a hook 2014-04-18T16:37:48Z stassats: well, i now use scp/rsync, no forgetting required 2014-04-18T16:42:18Z oleo quit (Ping timeout: 240 seconds) 2014-04-18T16:43:16Z oleo joined #sbcl 2014-04-18T16:58:32Z scymtym_ joined #sbcl 2014-04-18T17:09:07Z davazp joined #sbcl 2014-04-18T17:15:21Z fiveop quit (Quit: bye) 2014-04-18T17:55:39Z Hydan joined #sbcl 2014-04-18T18:04:17Z cmack` joined #sbcl 2014-04-18T18:06:05Z cmack quit (Ping timeout: 265 seconds) 2014-04-18T18:25:24Z davazp quit (Remote host closed the connection) 2014-04-18T18:30:14Z ehaliewicz joined #sbcl 2014-04-18T18:34:06Z angavrilov quit (Remote host closed the connection) 2014-04-18T18:36:12Z kanru` joined #sbcl 2014-04-18T18:37:03Z ASau` joined #sbcl 2014-04-18T18:37:55Z ASau quit (Remote host closed the connection) 2014-04-18T18:46:34Z ASau` quit (Remote host closed the connection) 2014-04-18T18:47:58Z ASau` joined #sbcl 2014-04-18T18:51:29Z Krystof is now known as sbcl|Krystof 2014-04-18T18:55:07Z ASau` quit (Remote host closed the connection) 2014-04-18T18:55:43Z ASau` joined #sbcl 2014-04-18T18:58:13Z ASau` is now known as ASau 2014-04-18T18:58:14Z attila_lendvai quit (Read error: Connection reset by peer) 2014-04-18T18:58:20Z attila_lendvai joined #sbcl 2014-04-18T18:58:21Z attila_lendvai quit (Changing host) 2014-04-18T18:58:21Z attila_lendvai joined #sbcl 2014-04-18T19:00:05Z attila_lendvai1 joined #sbcl 2014-04-18T19:00:05Z attila_lendvai quit (Disconnected by services) 2014-04-18T19:00:05Z attila_lendvai1 quit (Changing host) 2014-04-18T19:00:05Z attila_lendvai1 joined #sbcl 2014-04-18T19:02:02Z attila_lendvai1 quit (Client Quit) 2014-04-18T19:02:10Z attila_lendvai joined #sbcl 2014-04-18T19:02:11Z attila_lendvai quit (Changing host) 2014-04-18T19:02:11Z attila_lendvai joined #sbcl 2014-04-18T19:21:57Z sbcl|Krystof is now known as Krystof 2014-04-18T19:27:23Z kanru` quit (Ping timeout: 245 seconds) 2014-04-18T19:28:21Z segv- quit (Ping timeout: 258 seconds) 2014-04-18T20:00:34Z scymtym_: Krystof: is there anything left to do for finalizing the ELS submission? 2014-04-18T20:01:35Z davazp joined #sbcl 2014-04-18T20:03:09Z Krystof: I'm just working on it 2014-04-18T20:03:15Z attila_lendvai quit (Ping timeout: 240 seconds) 2014-04-18T20:03:43Z Krystof: I will make minor revisions to the existing paper, and probably not attempt to address the bigger questions 2014-04-18T20:04:57Z scymtym_: can i help? 2014-04-18T20:05:09Z Krystof: I'm certainly not going to try at this stage to mix sections 2 and 3; that would go horribly wrong 2014-04-18T20:05:27Z Krystof: I think the planned changes that I have are straightforward 2014-04-18T20:06:00Z Krystof: hopefully I'll get them done tonight -- assuming I do, if you have a moment to glance at them and check that I haven't mangled anything that would be helpful 2014-04-18T20:06:32Z scymtym_: sure 2014-04-18T20:07:28Z edgar-rft quit (Quit: execution expired into permanent bleeding) 2014-04-18T20:23:34Z lucasaiu joined #sbcl 2014-04-18T20:25:17Z davazp quit (Ping timeout: 265 seconds) 2014-04-18T20:51:22Z ehaliewicz quit (Remote host closed the connection) 2014-04-18T20:51:34Z ehaliewicz joined #sbcl 2014-04-18T20:56:58Z nyef quit (Quit: Offline for a while) 2014-04-18T21:00:44Z davazp joined #sbcl 2014-04-18T21:06:30Z Bike_ joined #sbcl 2014-04-18T21:07:42Z Bike_ is now known as Bicyclidine 2014-04-18T21:31:01Z yacks quit (Ping timeout: 258 seconds) 2014-04-18T21:35:24Z segv- joined #sbcl 2014-04-18T21:39:04Z stassats quit (Ping timeout: 258 seconds) 2014-04-18T21:54:01Z ehaliewicz quit (Remote host closed the connection) 2014-04-18T21:54:12Z krzysz00 joined #sbcl 2014-04-18T22:03:30Z ltbarcly_ quit (Quit: Computer has gone to sleep.) 2014-04-18T22:03:51Z eudoxia joined #sbcl 2014-04-18T23:09:20Z White_Flame joined #sbcl 2014-04-18T23:49:45Z Hydan is now known as Hydan` 2014-04-18T23:57:57Z Hydan` is now known as Hydan 2014-04-19T00:12:14Z eudoxia quit (Quit: leaving) 2014-04-19T00:36:08Z Bicyclidine quit (Ping timeout: 245 seconds) 2014-04-19T00:46:58Z Hydan is now known as Hydan` 2014-04-19T01:01:07Z faheem quit (Read error: Connection reset by peer) 2014-04-19T01:01:12Z faheem_ joined #sbcl 2014-04-19T01:06:11Z Bike quit (Read error: Connection reset by peer) 2014-04-19T01:06:34Z Bike joined #sbcl 2014-04-19T01:15:38Z Hydan` quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-19T01:42:33Z drmeister quit (Remote host closed the connection) 2014-04-19T01:50:03Z zRecursive joined #sbcl 2014-04-19T01:52:06Z ltbarcly joined #sbcl 2014-04-19T02:03:44Z segv- quit (Remote host closed the connection) 2014-04-19T02:11:14Z drmeister joined #sbcl 2014-04-19T02:16:18Z kanru` joined #sbcl 2014-04-19T02:23:22Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-19T02:38:39Z christoph_debian quit (Ping timeout: 250 seconds) 2014-04-19T02:41:53Z ltbarcly joined #sbcl 2014-04-19T02:51:57Z christoph_debian joined #sbcl 2014-04-19T03:05:29Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-19T03:05:53Z zRecursive quit (Remote host closed the connection) 2014-04-19T03:19:51Z kanru` quit (Ping timeout: 258 seconds) 2014-04-19T03:46:38Z davazp quit (Remote host closed the connection) 2014-04-19T04:16:36Z drmeister quit (Remote host closed the connection) 2014-04-19T04:20:37Z edgar-rft joined #sbcl 2014-04-19T04:24:45Z drmeister joined #sbcl 2014-04-19T04:43:26Z oleo quit (Read error: Connection reset by peer) 2014-04-19T04:44:43Z oleo joined #sbcl 2014-04-19T05:09:34Z Hydan_ joined #sbcl 2014-04-19T05:13:59Z pranavrc joined #sbcl 2014-04-19T05:13:59Z pranavrc quit (Changing host) 2014-04-19T05:13:59Z pranavrc joined #sbcl 2014-04-19T05:22:48Z yacks joined #sbcl 2014-04-19T06:03:30Z krzysz00 quit (Ping timeout: 240 seconds) 2014-04-19T06:14:23Z attila_lendvai joined #sbcl 2014-04-19T06:14:23Z attila_lendvai quit (Changing host) 2014-04-19T06:14:23Z attila_lendvai joined #sbcl 2014-04-19T06:17:03Z angavrilov joined #sbcl 2014-04-19T06:49:05Z zRecursive joined #sbcl 2014-04-19T07:11:59Z Hydan joined #sbcl 2014-04-19T07:13:08Z Hydan_ quit (Ping timeout: 258 seconds) 2014-04-19T07:13:09Z psilord quit (Ping timeout: 258 seconds) 2014-04-19T07:13:38Z psilord joined #sbcl 2014-04-19T07:22:37Z attila_lendvai quit (Ping timeout: 265 seconds) 2014-04-19T07:27:54Z yacks quit (Ping timeout: 240 seconds) 2014-04-19T07:57:03Z attila_lendvai joined #sbcl 2014-04-19T07:57:03Z attila_lendvai quit (Changing host) 2014-04-19T07:57:03Z attila_lendvai joined #sbcl 2014-04-19T07:59:07Z stassats joined #sbcl 2014-04-19T08:26:16Z zRecursive quit (Remote host closed the connection) 2014-04-19T08:33:05Z momo-reina joined #sbcl 2014-04-19T08:37:19Z momo-reina quit (Remote host closed the connection) 2014-04-19T08:53:15Z lucasaiu quit (Ping timeout: 240 seconds) 2014-04-19T08:58:20Z yacks joined #sbcl 2014-04-19T09:29:21Z momo-reina joined #sbcl 2014-04-19T09:33:54Z scymtym_ quit (Ping timeout: 240 seconds) 2014-04-19T09:34:52Z momo-reina quit (Remote host closed the connection) 2014-04-19T09:35:18Z momo-reina joined #sbcl 2014-04-19T09:36:21Z kludge` quit (Ping timeout: 252 seconds) 2014-04-19T09:39:40Z kludge` joined #sbcl 2014-04-19T09:39:40Z kludge` quit (Changing host) 2014-04-19T09:39:40Z kludge` joined #sbcl 2014-04-19T09:58:39Z stassats: i know get hangs in threaded tests, isn't that grand 2014-04-19T09:58:45Z momo-reina quit (Remote host closed the connection) 2014-04-19T11:00:28Z krzysz00 joined #sbcl 2014-04-19T11:36:24Z oleo quit (Quit: Leaving) 2014-04-19T11:56:49Z oleo joined #sbcl 2014-04-19T12:11:18Z pchrist_ joined #sbcl 2014-04-19T12:13:01Z Hydan_ joined #sbcl 2014-04-19T12:14:51Z Hydan quit (*.net *.split) 2014-04-19T12:14:53Z pchrist quit (*.net *.split) 2014-04-19T12:19:18Z attila_lendvai quit (Quit: Leaving.) 2014-04-19T12:53:33Z pranavrc quit 2014-04-19T13:04:43Z eudoxia joined #sbcl 2014-04-19T13:10:51Z nyef joined #sbcl 2014-04-19T13:11:21Z nyef: G'morning all. 2014-04-19T13:19:26Z scymtym_ joined #sbcl 2014-04-19T13:35:09Z ASau quit (Ping timeout: 250 seconds) 2014-04-19T13:37:03Z nicdev` joined #sbcl 2014-04-19T13:40:51Z nicdev quit (Ping timeout: 240 seconds) 2014-04-19T13:41:00Z stassats: got to optimize array typechecks by better widetag grouping => simple-array simple-vectors simple-string complex-strings complex-vectors complex-array 2014-04-19T13:41:18Z stassats: so that all kinds of checks are contiguous 2014-04-19T13:41:54Z nyef: Don't forget (array nil). 2014-04-19T13:42:13Z stassats: didn't 2014-04-19T13:42:25Z stassats: grouped it with strings 2014-04-19T13:42:37Z nyef: Good. 2014-04-19T13:43:15Z stassats: so, tests for vector, simple-vectors, strings, simple-strings etc. are just a single range comparison 2014-04-19T13:45:34Z stassats: the hardest part: restoring the proper constant values in the comments 2014-04-19T13:46:14Z davazp joined #sbcl 2014-04-19T13:46:39Z nyef: Wasn't there something about having a distinct relation between simple and complex specialized array types with the same element type? 2014-04-19T13:46:44Z leuler joined #sbcl 2014-04-19T13:48:30Z stassats: can't see any 2014-04-19T13:48:32Z leuler: Re widetag comments: I have been thinking recently to drop the comments in the code and instead automatically generate a table of the widetag values in the internals manual. Didn't get to it yet. 2014-04-19T13:48:53Z stassats: having them in comments is semi-nice, if only the were always correct 2014-04-19T13:49:10Z Hydan_ is now known as Hydan` 2014-04-19T13:49:46Z stassats is cooking up an emacs macro to sort them out 2014-04-19T13:49:47Z nyef: It was something like they were guaranteed to only differ by a single bit, IIRC. 2014-04-19T14:02:20Z stassats: a bit of slime-eval + C-x ( and it's easily updated 2014-04-19T14:23:16Z ASau joined #sbcl 2014-04-19T14:23:35Z nyef: stassats: Would you happen to know when typecheckfuns stopped getting used? 2014-04-19T14:24:58Z stassats: after my modifications to defstruct 2014-04-19T14:25:46Z stassats: it no longer uses closures for accessors, so no need for the roundabout way to check types 2014-04-19T14:26:08Z nyef: Ah, okay. 2014-04-19T14:26:25Z nyef: So I can't just kill it in my local tree. 2014-04-19T14:29:36Z krzysz00 quit (Quit: leaving) 2014-04-19T14:34:02Z Hydan` is now known as Hydan_ 2014-04-19T14:34:37Z Hydan_ is now known as Hydan` 2014-04-19T14:38:41Z eudoxia quit (Quit: wow such lunchtime) 2014-04-19T14:55:10Z kanru` joined #sbcl 2014-04-19T15:18:15Z Hydan` quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-19T15:26:23Z michael_lee joined #sbcl 2014-04-19T15:44:51Z davazp quit (Ping timeout: 240 seconds) 2014-04-19T15:50:23Z Hydan joined #sbcl 2014-04-19T15:54:45Z Hydan quit (Ping timeout: 252 seconds) 2014-04-19T16:04:06Z Krystof: there definitely was a relationship between the array type widetags 2014-04-19T16:04:25Z Krystof: single-bit differences between single/complex 2014-04-19T16:04:48Z Krystof: it may be that the new ordering is strictly superior anyway, I haven't checked 2014-04-19T16:08:13Z nyef: And I'm not likely to check soon unless it breaks something I'm doing, which is unlikely to happen this weekend. 2014-04-19T16:12:03Z stassats: it doesn't look like it was the case at the time, at least for (format t "~@{~8,'0b~%~}" complex-vector-nil-widetag simple-array-nil-widetag) 2014-04-19T16:12:08Z stassats: 11101001 and 10011001 2014-04-19T16:12:42Z nyef: 32-bit or 64-bit? Did you check the other specialized array types? 2014-04-19T16:12:49Z Krystof: not to worry 2014-04-19T16:13:00Z stassats: neither for complex-bit-vector-widetag and simple-bit-vector-widetag 2014-04-19T16:13:55Z Krystof: when I did this, there was a desire to optimize type tests for (vector t), simple-array, and (simple-array * (*)) 2014-04-19T16:14:12Z Krystof: as well as the string types 2014-04-19T16:14:15Z Krystof: and bit vectors 2014-04-19T16:18:56Z stassats: simple vectormy change optimized tests for simple-string, string, vector, simple vector 2014-04-19T16:19:29Z stassats: though, (typep x '(simple-array * (*))) doesn't look like it uses type vops 2014-04-19T16:20:30Z Krystof: ok. I propose not to worry about it unless someone comes saying that it's messed up their work 2014-04-19T16:21:31Z stassats: i didn't see any logical grouping, except for the complex/simple split, but i didn't check the bit patterns 2014-04-19T16:24:14Z Krystof: cd875f8c1fb306067521330fbf84411713b7c20d is where I (thought I) did this 2014-04-19T16:24:36Z Krystof: ... Mon Mar 1 2004 2014-04-19T16:24:44Z nyef: Ugh. Bloody branch-happy code. I mean, it's obviously correct code, but it's not in the least bit matched to the CPU. /-: 2014-04-19T16:33:47Z stassats: i missed one more grouping, but not sure which to prefer, base-strings or character-strings 2014-04-19T16:38:44Z stassats: can't really decide, will go with character-strings 2014-04-19T16:42:34Z nyef: Ahh... Bug in FAST-ASH/SIGNED=>SIGNED. Let's see if this changes things. (-: 2014-04-19T16:43:25Z davazp joined #sbcl 2014-04-19T16:44:11Z stassats: and things like (typep x '(or base-string (array character (*)))) are not rolled into one type check 2014-04-19T16:45:28Z nyef: Let's not start in on deficiencies in array typechecks, please? That TYPECASE thing was bad enough. 2014-04-19T16:46:05Z nyef: Oh, wait. That was the point of this discussion wasn't it? Nevermind. 2014-04-19T16:47:51Z Krystof: in my glorious future array-set-type representation, this will all be solved 2014-04-19T16:48:08Z stassats: i guess there needs to be a general array type check vop, and just dump any union types to it, and it will fetch and figure out all the widetag groupings 2014-04-19T16:49:33Z nyef: It'd have to know the inferred type as well, so that it wouldn't duplicate tests when it's already known to be an (AND ARRAY (NOT SIMPLE-ARRAY))... 2014-04-19T16:50:43Z nyef: And there's that lovely case where we can make an array type that won't round-trip through the type system (where specifier-type of type-specifier isn't type= to the original input). 2014-04-19T16:53:30Z nyef: Hrm. No change, still dies with some insane bignum header not being a number. 2014-04-19T16:54:08Z stassats: i'm also bothered by things like (the vector (the array x)) typechecking twice 2014-04-19T16:54:23Z Hydan joined #sbcl 2014-04-19T16:54:44Z nyef: Which... would be covered by your magic VOP if it knew the already asserted/inferred type, surely? 2014-04-19T16:54:45Z stassats: wasn't the move-from-unsigned not the culprit? 2014-04-19T16:55:05Z nyef: It was the previous culprit. 2014-04-19T16:55:16Z stassats: well, (the fixnum (the integer x)) isn't optimized either 2014-04-19T16:55:46Z stassats: that have to be done somewhere during IR1 2014-04-19T17:04:50Z jdz joined #sbcl 2014-04-19T17:17:40Z attila_lendvai joined #sbcl 2014-04-19T17:17:40Z attila_lendvai quit (Changing host) 2014-04-19T17:17:40Z attila_lendvai joined #sbcl 2014-04-19T17:25:15Z lucasaiu joined #sbcl 2014-04-19T17:37:32Z Hydan is now known as Hydan` 2014-04-19T17:39:57Z oleo quit (Read error: Connection reset by peer) 2014-04-19T17:41:49Z oleo joined #sbcl 2014-04-19T17:42:30Z davazp quit (Read error: Connection reset by peer) 2014-04-19T17:42:35Z yacks quit (Ping timeout: 250 seconds) 2014-04-19T17:56:45Z michael_lee quit (Quit: Ex-Chat) 2014-04-19T17:57:30Z stassats quit (Ping timeout: 240 seconds) 2014-04-19T18:06:55Z ehaliewicz joined #sbcl 2014-04-19T18:17:39Z yacks joined #sbcl 2014-04-19T18:19:13Z kanru` quit (Ping timeout: 252 seconds) 2014-04-19T18:39:23Z slyrus quit (Ping timeout: 252 seconds) 2014-04-19T18:48:11Z nyef: Got it! Slapped a breakpoint on *, and then also on SXHASH, checked the register contents when they tripped, compared with the source for SXHASH. The SYMBOL-HASH VOP is broken. 2014-04-19T18:54:33Z ASau` joined #sbcl 2014-04-19T18:55:28Z ASau quit (Remote host closed the connection) 2014-04-19T18:57:42Z ASau` is now known as ASau 2014-04-19T19:07:09Z Hydan` is now known as Hydan 2014-04-19T19:09:46Z ltbarcly joined #sbcl 2014-04-19T19:11:55Z ltbarcly_ joined #sbcl 2014-04-19T19:12:01Z nyef: Hrm. Failed aver in %PUTHASH... Or tail-called from something called from %PUTHASH. 2014-04-19T19:13:41Z ltbarcly_ quit (Read error: Connection reset by peer) 2014-04-19T19:14:20Z ltbarcly_ joined #sbcl 2014-04-19T19:14:42Z ltbarcly quit (Ping timeout: 240 seconds) 2014-04-19T19:15:15Z nyef: Hrm. (NOT (ZEROP FREE-KV-SLOT)). That's from %%PUTHASH. 2014-04-19T19:20:37Z nyef: And it really is a CANTHAPPEN scenario. 2014-04-19T19:23:17Z attila_lendvai quit (Quit: Leaving.) 2014-04-19T19:46:33Z stassats joined #sbcl 2014-04-19T19:58:32Z ltbarcly_ quit (Quit: Computer has gone to sleep.) 2014-04-19T20:03:16Z ltbarcly joined #sbcl 2014-04-19T20:05:39Z LiamH joined #sbcl 2014-04-19T20:06:41Z stassats: i got something mucking up pure tests on ppc, that's just what i need 2014-04-19T20:07:48Z ltbarcly quit (Client Quit) 2014-04-19T20:10:06Z angavrilov quit (Remote host closed the connection) 2014-04-19T20:26:18Z stassats quit (Ping timeout: 240 seconds) 2014-04-19T20:30:06Z fiveop joined #sbcl 2014-04-19T20:30:22Z nyef: Hello fiveop. 2014-04-19T20:33:12Z fiveop: Hi 2014-04-19T20:38:36Z nyef: Looks like it dies on the 32nd call to PUTHASH. Which means the second time the table needs resizing, IIUC. 2014-04-19T20:40:41Z fiveop: So the first time might leave behind some garbage 2014-04-19T20:42:16Z nyef: Hrm, no, rehash-size is 1.5. 2014-04-19T20:42:39Z fiveop: What size do we start with? 2014-04-19T20:43:02Z nyef: 16. 2014-04-19T20:43:50Z nyef: Double-checked, no special parameters to M-H-T other than that it's an EQUAL table. 2014-04-19T20:52:25Z nyef: Double-checking: Making a rehash dump some output, so I know when it gets triggered. 2014-04-19T21:02:30Z nyef: Okay, it IS just-post-second-rehash. 2014-04-19T21:04:03Z nyef: Rehash occurs on the 17th insert and the 33rd insert? Hrm. 2014-04-19T21:07:43Z lucasaiu left #sbcl 2014-04-19T21:22:35Z nyef: I should probably sort another seven commits out of my working tree before I dig too much deeper into this. 2014-04-19T21:24:00Z fiveop: I'm still setting up nfs (right now, my pi's doing a dist upgrade, which takes for ages) 2014-04-19T21:32:33Z eudoxia joined #sbcl 2014-04-19T21:35:08Z fiveop: That's done. Tomorrow, I should be able to do some actual work :) 2014-04-19T21:35:31Z nyef: Oh, good. I'm probably not going to be able to do much on this tomorrow. 2014-04-19T21:53:59Z fiveop: good night 2014-04-19T21:54:01Z fiveop quit (Quit: bye) 2014-04-19T21:57:59Z nyef: And pushed. 2014-04-19T22:07:17Z jdz: what exactly is supposed to be the state of threading on OSX (Mavericks)? 2014-04-19T22:07:36Z jdz: i'm getting one unexpected success, and one failure on tests 2014-04-19T22:08:17Z nyef: I... have no idea. What are the success and failure? 2014-04-19T22:08:49Z jdz: Unexpected success: threads.pure.lisp / (SEMAPHORE-NOTIFICATION WAIT-ON-SEMAPHORE LP-1038034) 2014-04-19T22:09:02Z jdz: Failure: threads.impure.lisp / FP-MODE-INHERITANCE-THREADS 2014-04-19T22:09:59Z jdz: on sbcl-1.1.17-84-gcc56b59 2014-04-19T22:10:50Z nyef: Hrm. Those don't sound too-too bad, all things considered. 2014-04-19T22:12:13Z fe[nl]ix quit (Remote host closed the connection) 2014-04-19T22:12:13Z Blkt quit (Remote host closed the connection) 2014-04-19T22:12:22Z Blkt joined #sbcl 2014-04-19T22:12:24Z fe[nl]ix joined #sbcl 2014-04-19T22:14:24Z jdz: well, just let me know if I may be of assistance in fixing those 2014-04-19T22:16:49Z jdz: hmm, weird, where did git describe get the initial letter g for commit hash? 2014-04-19T22:17:11Z jdz: oh wait, that's what it does 2014-04-19T22:17:13Z jdz: nvm 2014-04-19T22:27:34Z nyef: Hash table fails to grow the second time. Oddly, the first time the old-size is 0x00000044 and the new size is 0x00000080, suggesting something wrong as far back as table creation. 2014-04-19T22:31:31Z nyef: Possibly just something weird with table creation. 2014-04-19T22:37:16Z leuler quit (Quit: ERC Version 5.1.2 $Revision: 1.796.2.6 $ (IRC client for Emacs)) 2014-04-19T22:49:49Z nyef: ... (truncate (* 32 1.5)) => 24 ? What? 2014-04-19T22:50:04Z nyef: No, that's not 24, is it? 2014-04-19T22:50:20Z nyef: 18. 2014-04-19T22:50:30Z nyef: That's 18. Even better. /-: 2014-04-19T22:52:02Z nyef: Time to get a trace file. 2014-04-19T23:06:47Z nyef: ... Wha...? FMSR OCFP, NARGS / FUITOS F1, OCFP?!? That's not how that's supposed to go. 2014-04-19T23:08:02Z nyef: Ouch on two fronts: First is that I screwed up the VOP definition. Second is that the instruction definitions don't do an SC or SB check on the TNs involved. 2014-04-19T23:15:10Z nyef: And a full call for %UNARY-TRUNCATE/SINGLE-FLOAT. Lovely. 2014-04-19T23:24:51Z nyef: Oh, neat. x86-64 does the full call there, too. 2014-04-19T23:25:25Z nyef: Did someone say THE where they should have said TRULY-THE, or is it intended to take a full call and throw an error if the result doesn't fit? 2014-04-19T23:41:00Z davazp joined #sbcl 2014-04-19T23:55:30Z jdz quit (Remote host closed the connection) 2014-04-20T00:09:04Z eudoxia quit (Quit: wow such dinner) 2014-04-20T00:13:41Z nyef: And %UNARY-TRUNCATE/SINGLE-FLOAT appears to be equivalent to (CONSTANTLY 18), for all three known input values (17, 25.5, and 48). 2014-04-20T00:14:59Z nyef: That might not be an entirely bad situation, you'd be legal for sex and to vote on this side of the pond, AIUI, but having a drink would be a bit of a problem. 2014-04-20T00:32:17Z joshe quit (Ping timeout: 246 seconds) 2014-04-20T00:33:21Z joshe joined #sbcl 2014-04-20T00:37:27Z pchrist_ quit (Quit: leaving) 2014-04-20T00:38:11Z pchrist joined #sbcl 2014-04-20T00:44:59Z eudoxia joined #sbcl 2014-04-20T01:10:30Z eudoxia quit (Quit: leaving) 2014-04-20T01:17:29Z nyef sighs. 2014-04-20T01:17:35Z nyef: Parameters are backwards on FMRS. 2014-04-20T01:34:03Z nyef: ... And now it locks up in !RANDOM-COLD-INIT... Unless it's taking a stupidly long time. 2014-04-20T02:07:17Z Hydan quit (Quit: Textual IRC Client: www.textualapp.com) 2014-04-20T02:07:47Z Hydan joined #sbcl 2014-04-20T02:12:45Z zRecursive joined #sbcl 2014-04-20T02:39:24Z christoph_debian quit (Ping timeout: 258 seconds) 2014-04-20T02:53:26Z christoph_debian joined #sbcl 2014-04-20T02:58:39Z momo-reina joined #sbcl 2014-04-20T02:58:44Z nyef quit (Quit: G'night all.) 2014-04-20T03:13:41Z ehaliewicz quit (Read error: No route to host) 2014-04-20T03:14:07Z momo-reina quit (Ping timeout: 245 seconds) 2014-04-20T03:18:09Z momo-reina joined #sbcl 2014-04-20T03:38:33Z momo-reina quit (Remote host closed the connection) 2014-04-20T03:56:42Z joshe quit (Ping timeout: 240 seconds) 2014-04-20T04:02:18Z joshe joined #sbcl 2014-04-20T04:09:06Z zRecursive quit (Remote host closed the connection) 2014-04-20T04:21:14Z ehaliewicz joined #sbcl 2014-04-20T04:36:08Z pranavrc joined #sbcl 2014-04-20T04:36:08Z pranavrc quit (Changing host) 2014-04-20T04:36:08Z pranavrc joined #sbcl 2014-04-20T05:00:19Z davazp quit (Remote host closed the connection) 2014-04-20T05:03:01Z slyrus joined #sbcl 2014-04-20T05:24:57Z pranavrc quit (Remote host closed the connection) 2014-04-20T05:48:51Z LiamH quit (Ping timeout: 240 seconds) 2014-04-20T05:50:50Z attila_lendvai joined #sbcl 2014-04-20T05:50:50Z attila_lendvai quit (Changing host) 2014-04-20T05:50:50Z attila_lendvai joined #sbcl 2014-04-20T06:11:04Z drmeister quit (Remote host closed the connection) 2014-04-20T06:12:32Z DGASAU quit (Ping timeout: 258 seconds) 2014-04-20T06:22:15Z oleo quit (Ping timeout: 276 seconds) 2014-04-20T06:24:05Z DGASAU joined #sbcl 2014-04-20T06:35:24Z oleo joined #sbcl 2014-04-20T07:04:44Z angavrilov joined #sbcl 2014-04-20T07:19:15Z slyrus quit (Ping timeout: 240 seconds) 2014-04-20T07:20:41Z stassats joined #sbcl 2014-04-20T07:25:09Z slyrus joined #sbcl 2014-04-20T07:27:47Z milosn quit (Quit: Lost terminal) 2014-04-20T07:28:14Z milosn joined #sbcl 2014-04-20T07:37:24Z DGASAU quit (Remote host closed the connection) 2014-04-20T07:37:39Z slyrus quit (Ping timeout: 240 seconds) 2014-04-20T07:43:59Z slyrus joined #sbcl 2014-04-20T07:59:58Z drmeister joined #sbcl 2014-04-20T08:04:27Z drmeister quit (Ping timeout: 240 seconds) 2014-04-20T08:14:03Z ehaliewicz quit (Ping timeout: 240 seconds) 2014-04-20T08:51:48Z attila_lendvai quit (Read error: Connection reset by peer) 2014-04-20T09:30:55Z jdz joined #sbcl 2014-04-20T09:39:32Z kludge` quit (Ping timeout: 258 seconds) 2014-04-20T09:40:27Z kludge` joined #sbcl 2014-04-20T09:48:41Z drmeister joined #sbcl 2014-04-20T09:53:06Z drmeister quit (Ping timeout: 240 seconds) 2014-04-20T10:13:11Z oleo quit (Ping timeout: 250 seconds) 2014-04-20T10:32:20Z stassats: running mop.pure before clos.pure is what's causing ppc failures 2014-04-20T10:33:12Z oleo joined #sbcl 2014-04-20T10:37:43Z ASau` joined #sbcl 2014-04-20T10:40:47Z ASau quit (Ping timeout: 252 seconds) 2014-04-20T10:40:52Z stassats: well, it does fail everywhere, and it's just a peculiritaty of my ppc system that it runs mop.pure before clos.pure 2014-04-20T10:41:09Z stassats: mop.pure does (make-instance 'standard-class), and clos.pure has (defun standardized-class-p (c) (eq (class-name c) (find-symbol (symbol-name (class-name c)) "CL"))) 2014-04-20T10:41:44Z Krystof: ha 2014-04-20T10:41:51Z Krystof: and then a map-all-classes? 2014-04-20T10:41:54Z Krystof: that's... brilliant 2014-04-20T10:41:55Z stassats: yes 2014-04-20T10:42:43Z Krystof: I'd say that standardized-class-p should be fixed to ignore anonymous classes 2014-04-20T10:43:08Z Krystof: but maybe also it's not "pure" to make new classes 2014-04-20T10:46:19Z stassats: well, nothing is really pure, interning symbols for one 2014-04-20T10:48:13Z stassats: i think the test could be done with using a prototype 2014-04-20T10:48:17Z Strigoides joined #sbcl 2014-04-20T10:48:54Z stassats: nope, a prototype doesn't have c-direct-superclasses 2014-04-20T10:52:36Z stassats: and (eq (class-name c) (find-symbol (symbol-name (class-name c)) "CL")) is a roundabout way to say (eql (symbol-package (class-name c)) (find-package :cl)) 2014-04-20T11:05:18Z stassats: (not for cases when the CL package reexports some other package, but that's not the case for sbcl) 2014-04-20T11:06:41Z Strigoides quit (Quit: leaving) 2014-04-20T11:26:28Z stassats: and regarding one-bit differing widetags, where they do match on x86, they don't on x86-64 2014-04-20T11:37:31Z drmeister joined #sbcl 2014-04-20T11:42:42Z drmeister quit (Ping timeout: 276 seconds) 2014-04-20T12:27:38Z leuler joined #sbcl 2014-04-20T12:52:34Z Krystof: right. I now have a pi and a beaglebone in my home network 2014-04-20T12:53:18Z Krystof: only one of them having previously been doused in coffee 2014-04-20T12:53:52Z LiamH joined #sbcl 2014-04-20T13:06:32Z davazp joined #sbcl 2014-04-20T13:26:17Z drmeister joined #sbcl 2014-04-20T13:30:47Z drmeister quit (Ping timeout: 250 seconds) 2014-04-20T13:31:34Z ehaliewicz joined #sbcl 2014-04-20T13:36:03Z ehaliewicz quit (Ping timeout: 240 seconds) 2014-04-20T14:05:41Z jdz_ joined #sbcl 2014-04-20T14:08:17Z jdz quit (Ping timeout: 245 seconds) 2014-04-20T14:29:11Z ltbarcly joined #sbcl 2014-04-20T14:30:27Z drmeister joined #sbcl 2014-04-20T14:40:27Z attila_lendvai joined #sbcl 2014-04-20T14:40:27Z attila_lendvai quit (Changing host) 2014-04-20T14:40:27Z attila_lendvai joined #sbcl 2014-04-20T14:51:29Z fiveop joined #sbcl 2014-04-20T14:53:15Z ltbarcly_ joined #sbcl 2014-04-20T14:55:15Z ltbarcly quit (Ping timeout: 240 seconds) 2014-04-20T15:01:07Z ltbarcly_ quit (Quit: Computer has gone to sleep.) 2014-04-20T15:44:50Z yacks quit (Quit: Leaving) 2014-04-20T16:12:57Z pkhuong: so, Krystof/scymtym: tips on responding to peer review without throwing one's laptop out the window? 2014-04-20T16:20:21Z Krystof: "it's (probably) not personal" 2014-04-20T16:20:47Z Krystof: preferably throw your laptop at the idiot reviewer 2014-04-20T16:22:59Z pkhuong: some are easy. "you should cite this not exactly relevant paper" Sure, *wink wink nudge nudge*. Others, not so much, especially when, two years after writing the thing, I'm not sure I disagree ;) 2014-04-20T16:36:16Z scymtym_: i'm the last person who should give advice regarding publication matters, sorry 2014-04-20T16:50:51Z Krystof: I suppose the question I'd ask is what do you most care about? objective/subjective quality vs publication expedience? 2014-04-20T17:05:19Z nyef joined #sbcl 2014-04-20T17:05:31Z nyef: Hello all. 2014-04-20T17:09:59Z fiveop: Am I at the point you left off, when one of my cores is running at 100% and nothing happens? 2014-04-20T17:10:02Z fiveop: Hi 2014-04-20T17:10:32Z fiveop: I pushed all (the few) changes I still lying around. 2014-04-20T17:15:07Z nyef: If you build with SB-SHOW, it should do that in !RANDOM-COLD-INIT, yes. 2014-04-20T17:15:27Z fiveop: yes it does 2014-04-20T17:15:30Z stassats: probably the lack of modular arithmetic 2014-04-20T17:15:45Z attila_lendvai1 joined #sbcl 2014-04-20T17:15:45Z attila_lendvai quit (Disconnected by services) 2014-04-20T17:15:45Z attila_lendvai1 quit (Changing host) 2014-04-20T17:15:45Z attila_lendvai1 joined #sbcl 2014-04-20T17:15:59Z nyef: stassats: Shouldn't be an issue, we're lacking just as much modular arithmetic as some of the other backends, AIUI. 2014-04-20T17:16:22Z stassats: last i looked, there were almost none 2014-04-20T17:20:03Z stassats: you need to have all the [c]/unsigned=>unsigned vops first 2014-04-20T17:20:13Z nyef: The other thing is, modular arithmetic is supposed to be an optimization, not a requirement. 2014-04-20T17:20:50Z stassats: well, it will certainly make a thing like PRNG quite slow 2014-04-20T17:21:02Z pkhuong: or hashing 2014-04-20T17:21:13Z nyef: Slow is one thing. Taking forever to initialize a random state is another. 2014-04-20T17:22:08Z ltbarcly joined #sbcl 2014-04-20T17:23:53Z stassats: it requires only 624 iterations, so, it should still be fast 2014-04-20T17:24:06Z nyef: See? 2014-04-20T17:24:32Z attila_lendvai1 quit (Ping timeout: 245 seconds) 2014-04-20T17:24:38Z attila_lendvai joined #sbcl 2014-04-20T17:24:38Z attila_lendvai quit (Changing host) 2014-04-20T17:24:38Z attila_lendvai joined #sbcl 2014-04-20T17:24:39Z nyef: I think that we need to get through !COLD-INIT and remove as many arm-specific hacks as we can before we worry about optimizations. 2014-04-20T17:25:13Z stassats: have you tried sb-qshow? 2014-04-20T17:25:21Z stassats: maybe it's a GC issue, or something 2014-04-20T17:26:39Z nyef: GC isn't enabled yet. 2014-04-20T17:27:03Z nyef: And gdb would catch if the trigger gets hit, anyway. 2014-04-20T17:27:46Z stassats tries to get a debian chroot on android 2014-04-20T17:30:28Z nyef: If you're going to use gdb, please note that calling runtime functions while "in lisp" will corrupt the control stack. 2014-04-20T17:37:59Z stassats: do you compile on the target? 2014-04-20T17:38:03Z attila_lendvai quit (Ping timeout: 252 seconds) 2014-04-20T17:41:12Z nyef: We're not far enough through cold-init to run the compiler, if that's what you're asking. 2014-04-20T17:41:18Z nyef: We run target-1 on the target, though. 2014-04-20T17:41:56Z nyef: Run config on the target, host-1 on a host, target-1 on the target, host-2 on a host, then watch the cold-core crash, lock up, or whatever on the target. 2014-04-20T17:42:30Z stassats: well, ccl can be used to bootstrap 2014-04-20T17:43:14Z nyef: That's fair, I guess, but I haven't tried that. 2014-04-20T17:43:20Z stassats: though, that will make it even slower 2014-04-20T17:44:27Z nyef: Yeah, I at least have a a few hundred MHz and a few gigs of RAM to throw around on my host machine. 2014-04-20T17:46:27Z slyrus quit (Ping timeout: 240 seconds) 2014-04-20T17:54:59Z stassats: i think i chose too little space with 512MB, 219MB left 2014-04-20T17:55:49Z stassats: should be enough for gcc and gdb 2014-04-20T17:56:53Z stassats: bloody gdb fetches python with it 2014-04-20T17:57:28Z nyef: Really? That's... messed up. 2014-04-20T17:57:42Z stassats: i guess setting up nfs would be easier for running make-config and friends 2014-04-20T17:58:28Z nyef: I set up a small number of one-line shell-scripts to do rsync. 2014-04-20T18:03:42Z stassats quit (Ping timeout: 245 seconds) 2014-04-20T18:04:30Z stassats joined #sbcl 2014-04-20T18:06:47Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-20T18:08:22Z stassats: nfs client doesn't even work, so, rsync it is 2014-04-20T18:18:33Z fiveop: we have the constant mt19937-n set to 624 and sprinkled over the code we have the literal 627 for the size of the random state, while the two depend on each other. 2014-04-20T18:18:36Z fiveop: that's ugly 2014-04-20T18:20:12Z nyef: My plan for resolving this one is to take trace-files of target-random and random, along with my existing trace-files for float and bignum, then use the map file to find the functions involved so that I can set breakpoints in gdb and to disassemble the code involved. 2014-04-20T18:20:52Z nyef: I suspect a chain of tail-calls, one of which should be open-coded instead, or some similar damage. 2014-04-20T18:21:09Z nyef: Alternately, there's an accidental infinite loop due to data corruption somehow. 2014-04-20T18:22:06Z nyef: I should have time to dig into it properly either this evening or tomorrow. 2014-04-20T18:25:10Z ASau` is now known as ASau 2014-04-20T18:26:27Z stassats: ok, time to target make-target-1.sh 2014-04-20T18:26:30Z stassats: run 2014-04-20T18:27:03Z fiveop: What device are you using? 2014-04-20T18:27:36Z stassats: galaxy s2 2014-04-20T18:43:07Z stassats: ok, i got mixed up with rsync and trailing "/" more than i should 2014-04-20T18:43:12Z stassats: back to make-host-2 2014-04-20T18:45:26Z nyef: Getting mixed up with rsync is why I ended up making simple shell scripts for it. (-: 2014-04-20T18:45:47Z stassats: i got mixed while making the scripts 2014-04-20T18:46:31Z fiveop: It hangs the first time it runs the then clause of the loop in init-random-state 2014-04-20T18:46:37Z stassats: why, now i have a loop too! 2014-04-20T18:47:11Z stassats: and the phone is getting warmer 2014-04-20T18:50:26Z stassats: at least i have now something else than just "think hard" 2014-04-20T18:51:07Z nyef: Okay, I'm going to sign off in about ten minutes. I should be back about three hours later, give or take. 2014-04-20T18:51:36Z nyef: At that point I can probably start digging into this, if nobody has figured it out by then. 2014-04-20T18:55:28Z fiveop: almost there (not as sophisticated as the method you explained above, but I'll start looking at the trace files, afterwards :)) https://gist.github.com/anonymous/584ba84b0519983446e1 2014-04-20T18:55:47Z oleo quit (Ping timeout: 252 seconds) 2014-04-20T18:58:18Z oleo joined #sbcl 2014-04-20T19:00:21Z fiveop: I get to see "step 5" but it does not enter the second iteration 2014-04-20T19:03:34Z nyef: So, possibly something with LOGAND? 2014-04-20T19:05:52Z nyef: So, we have bignum, fixnum-or-bignum, and the operation is always in tail position, and if it's the fixnum-fixnum case is converted as a full call which should blow the stack. 2014-04-20T19:06:34Z fiveop: if I read the trace file right, the compiler optimizes all my calls between the (/show0) forms out 2014-04-20T19:06:40Z fiveop: so this doesn't tell us anything 2014-04-20T19:07:02Z nyef: Okay, wrap each form in /hexstr 2014-04-20T19:08:37Z fiveop: (I should kill the sbcl process, before going back to emacs.) 2014-04-20T19:10:44Z stassats: hm, built with sb-show and now get a fault at #xFFFFFFFC 2014-04-20T19:13:24Z fiveop: How are trace files organized with respect to the lisp file? 2014-04-20T19:13:40Z nyef: Linearly. 2014-04-20T19:13:46Z Krystof: each top-level thingy gets an IR1 then IR2 then disassembly block 2014-04-20T19:13:47Z stassats: fiveop: in the normal order? 2014-04-20T19:13:49Z fiveop: Is it always IR1 of TLF1, IR2 of TLF1, IR1 of TLF2 etc? 2014-04-20T19:13:55Z Krystof: yes 2014-04-20T19:14:05Z fiveop: so there are three 2014-04-20T19:14:10Z nyef: Looks like I've got an extra 15 minutes or so, my ride is late. 2014-04-20T19:14:52Z fiveop: and ";;;; component top level form" is the separator? 2014-04-20T19:15:19Z ltbarcly joined #sbcl 2014-04-20T19:15:39Z Krystof: no, maybe ";;;; component" 2014-04-20T19:15:49Z Krystof: that is there might well be ";;;; component XXX" also 2014-04-20T19:15:55Z stassats: so, i get a loop without sb-show, no idea where, and sb-show faults at -4, no idea where 2014-04-20T19:16:06Z Krystof: apart from that, everything's working 2014-04-20T19:16:23Z fiveop: stassats: when did you pull from nyef's repo last? 2014-04-20T19:16:37Z stassats: tonight? 2014-04-20T19:17:04Z fiveop: No idea what that means. Which timezone do you reside in? 2014-04-20T19:17:05Z nyef: HEAD is "fix single-reg-transfer-fp-instructions"? 2014-04-20T19:17:13Z stassats: yes 2014-04-20T19:17:22Z nyef: Hrm. 2014-04-20T19:17:43Z nyef: Maybe I should find the bits to re-set-up the debian chroot on my note 10.1. 2014-04-20T19:19:58Z nyef: stassats: Your next step is gdb, I guess. /-: 2014-04-20T19:20:17Z stassats: and sometimes i get a lot of bonus faults after the -4 one 2014-04-20T19:20:45Z nyef: SBCL plays worse stack games than usual on ARM, and if the platform depends on sane stack handling then things will break quickly. 2014-04-20T19:21:19Z stassats: well, the fault happens during call_into_c 2014-04-20T19:22:08Z nyef: ... Thumb? 2014-04-20T19:22:36Z nyef: Having a thumb runtime / libc might do that. 2014-04-20T19:23:14Z fiveop: so its step 3 (the multiplication) 2014-04-20T19:23:45Z fiveop: I get the result of the xor, see "step 3" and it runs away 2014-04-20T19:24:08Z stassats: nyef: how do i figure that? 2014-04-20T19:24:47Z nyef: stassats: Objdump the runtime and libc, see if the instructions are 16-bits or 32-bits wide. 2014-04-20T19:24:58Z nyef: Objdump --disassemble, that is. 2014-04-20T19:25:23Z stassats: looks 32-bit to me 2014-04-20T19:25:38Z nyef: Hrm. 2014-04-20T19:25:48Z nyef: Dunno then, I'm afraid. 2014-04-20T19:26:12Z stassats: gdb says it fails at line 200, that is ldr sp, STATIC_SYMBOL_VALUE(NUMBER_STACK_POINTER) 2014-04-20T19:26:47Z nyef: Is reg_NULL still pointing correctly at that point? 2014-04-20T19:27:13Z Krystof: what platforms are you on, nyef / fiveop? 2014-04-20T19:27:18Z stassats: i can't run context 2014-04-20T19:27:55Z fiveop: raspberry pi 2014-04-20T19:28:05Z nyef: Raspberry Pi here as well for now. 2014-04-20T19:28:28Z nyef: I also have a Galaxy Note 10.1 and a BeagleBone Black, the latter not yet set up. 2014-04-20T19:28:38Z stassats: i can run, but it greets me with an endless CORRUPTION WARNING, CORRUPTION WARNING 2014-04-20T19:28:50Z Krystof: I'll work on the BBB, then 2014-04-20T19:28:52Z fiveop: once I found the IR1 block for a piece of code I'm interested in, is there an easo way to find the relevant passages in IR2 and disassemble output? 2014-04-20T19:28:59Z nyef: stassats: Okay, that sounds familiar. Try doing --lose-on-corruption. 2014-04-20T19:29:21Z nyef: fiveop: Match up the block number from IR1 to the IR2 level, then look for surrounding labels. 2014-04-20T19:29:36Z nyef: The labels would match up to the disassembly, along with the VOP information. 2014-04-20T19:29:52Z stassats: by the time gdb gets there, r10 is #x944 2014-04-20T19:31:15Z stassats: with --lose-on-corruption, con says "no interrupt context" 2014-04-20T19:31:42Z nyef: Yeah, but my next question is, was it printing something else when the CORRUPTION WARNING arrived? 2014-04-20T19:32:07Z stassats: and backtrace is different 2014-04-20T19:33:31Z nyef: And my ride is here, so I've got to go. I'll be back in a few hours. 2014-04-20T19:33:34Z nyef quit (Quit: Bye all.) 2014-04-20T19:33:40Z stassats: (and i don't get no output from sb-show whatsoever) 2014-04-20T19:34:50Z stassats: ok, let's build without sb-show again 2014-04-20T19:44:12Z stassats: i know get that without sb-show too 2014-04-20T19:44:25Z stassats: now 2014-04-20T19:52:35Z stassats: oh goddamn, i've been syncinc the core into the wrong place 2014-04-20T19:53:46Z edgar-rft quit (Quit: experience vanished because no future found) 2014-04-20T19:54:10Z stassats: ok, back to the loop, and back to enabling sb-show, sigh 2014-04-20T19:54:50Z _8hzp` joined #sbcl 2014-04-20T19:56:03Z _8hzp quit (Ping timeout: 240 seconds) 2014-04-20T19:58:32Z stassats: yay, entering !random-cold-init 2014-04-20T19:59:00Z Krystof: hm, where's my job control? 2014-04-20T20:02:01Z fiveop: so we call two-arg-* 2014-04-20T20:02:33Z Krystof: with what arguments? 2014-04-20T20:03:00Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-20T20:03:38Z fiveop: 1812433253 (constant) and 5489 2014-04-20T20:03:42Z stassats: ok, i just return an uninitialized array, and it fails at !package-cold-init 2014-04-20T20:04:01Z fiveop: That's what's supposed to happen. I'm reading the trace file 2014-04-20T20:04:15Z fiveop: the result of the multiplication is a bignum 2014-04-20T20:08:38Z Krystof: is this the first call to two-arg-*? 2014-04-20T20:12:30Z Krystof: does anyone know off the top of their head how to translate from addresses in cold-sbcl.map to first instruction in functions? 2014-04-20T20:14:22Z stassats: fiveop: how to enable tracing for make-host-2? 2014-04-20T20:15:38Z Krystof: no, it isn't the first time we call two-arg-* 2014-04-20T20:15:51Z angavrilov quit (Remote host closed the connection) 2014-04-20T20:16:08Z stassats: just :trace in build-order? ok 2014-04-20T20:17:16Z stassats: probably the first time bignum*fixnum? 2014-04-20T20:18:09Z drmeister quit (Ping timeout: 276 seconds) 2014-04-20T20:20:17Z Krystof: the registers agree with the trace file 2014-04-20T20:20:41Z Krystof: and yes, from sampling the previous hits they were all fixnum*fixnum 2014-04-20T20:20:50Z stassats: calling multiply-bignum-and-fixnum directly does hang too 2014-04-20T20:22:24Z stassats: there's a loop which does bignum-length things, it can get out of hand quickly if bignum lengths are damaged like that fffffffa thing 2014-04-20T20:22:35Z Krystof sets a breakpoint at %multiply-and-add 2014-04-20T20:23:42Z Krystof: oh, wait, that's implemented in the backend 2014-04-20T20:24:40Z stassats: or not implemented 2014-04-20T20:25:02Z stassats: there's your loop 2014-04-20T20:25:24Z Krystof: whoops 2014-04-20T20:25:43Z Krystof: I wonder how long it would have taken me to spot that 2014-04-20T20:26:12Z stassats: minion: chant 2014-04-20T20:26:12Z minion: MORE ALWAYSTRANSLATABLE 2014-04-20T20:27:36Z Krystof: in fact there's quite a lot that's not implemented 2014-04-20T20:27:40Z Krystof: digit-0-or-plusp 2014-04-20T20:28:51Z fiveop: stassats: :trace-file 2014-04-20T20:29:23Z stassats: figured that, i though there was an option to generate trace files for everything at once 2014-04-20T20:31:21Z fiveop: Krystof: I just read what you wrote the last quarter hour. So we are just missing vops etc? 2014-04-20T20:31:46Z stassats: i think someone should go through circular definitions and add ALWAYS-TRANSLATABLE, lest to repeat the pain 2014-04-20T20:32:38Z fiveop: what is ALWAYS-TRANSLATEABLE? 2014-04-20T20:33:48Z pkhuong: fiveop: a ir1 attribute so that the compiler will scream at you if a call to a given function doesn't match a VOP 2014-04-20T20:34:08Z stassats: prevents generating full calls to what should always be a VOP 2014-04-20T20:35:56Z Krystof: stassats: I'm doing these ones at least 2014-04-20T20:36:08Z stassats: i'm not brave enough to convert ppc's %multiply-and-add to arm 2014-04-20T20:36:11Z Krystof: I'm also cargo-culting vop definitions 2014-04-20T20:36:20Z Krystof: because what could go wrong? 2014-04-20T20:36:32Z Krystof: I've done this on multiple platforms in the past 2014-04-20T20:36:39Z Krystof: who needs instruction reference manuals? 2014-04-20T20:36:45Z stassats: (waking up dreaming of VOPs may be considered wrong) 2014-04-20T20:38:20Z stassats: four people on the ARM port? i guess that's good 2014-04-20T20:38:41Z stassats: (in other news, intel is doomed) 2014-04-20T20:39:39Z Krystof: just today I was looking for a nice arm desktop to replace my 10-year-old noisebox machine 2014-04-20T20:39:42Z Krystof: I didn't find one :-/ 2014-04-20T20:39:54Z Krystof: my basic requirement is "should run debian" 2014-04-20T20:40:56Z stassats: i guess you can only get all in ones or laptopy/tablety things 2014-04-20T20:41:27Z Krystof: the cubox-i was the closest, but there's at least a month's lead time 2014-04-20T20:42:01Z pkhuong: Krystof: Novena ;) 2014-04-20T20:42:57Z Krystof: stop distracting me! 2014-04-20T20:43:43Z fiveop: pkhuong: is it done, as in are they being sold? 2014-04-20T20:44:16Z pkhuong: fiveop: it's pretty clear the campaign will reach its funding goals for a first batch. 2014-04-20T20:44:19Z stassats: i think there needs to be some better way to share arm progress, maybe a branch on sf git? 2014-04-20T20:44:53Z fiveop: but someone who's here more often than nyef should 'maintain' it 2014-04-20T20:44:59Z pkhuong: stassats: github will be easier to grant write access. 2014-04-20T20:45:15Z Krystof: meh, this fix and we're done 2014-04-20T20:45:19Z Krystof: ha ha ha ha ha 2014-04-20T20:47:37Z stassats: re instruction manuals, google brought me to http://www.heyrick.co.uk/armwiki/Main_Page and i found it helpful 2014-04-20T20:47:52Z Krystof: I think bignum-mult-and-add-3-arg is 2014-04-20T20:47:54Z Krystof: (inst umull temp hi x y) 2014-04-20T20:47:54Z Krystof: (inst adcs lo temp carry-in) 2014-04-20T20:47:54Z Krystof: (inst mov :cs temp 1) 2014-04-20T20:47:58Z Krystof: (inst mov :cc temp 0) 2014-04-20T20:47:58Z Krystof: (inst adcs hi hi temp))) 2014-04-20T20:48:17Z Krystof: please critique 2014-04-20T20:48:33Z Krystof: this is my first^Wsecond ever arm code (first was the one-line fix to arm-assem.S a few months back) 2014-04-20T20:48:38Z stassats: "know jack of arm" 2014-04-20T20:48:47Z Krystof: tell me again about how much intel is doomed 2014-04-20T20:49:00Z pkhuong: currently blocked on trying to open http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.dui0497a/BABFFEJF.html 2014-04-20T20:49:31Z stassats: why are the intel manuals so good? 2014-04-20T20:49:46Z Krystof: I can't even keep the different arm versions in my head 2014-04-20T20:50:09Z pkhuong: ADCS: "Rd and Rn must specify the same register." 2014-04-20T20:50:11Z Krystof: I see v9, v15, v7, v8, A9, ... and simply don't even understand what they're talking about 2014-04-20T20:50:22Z Krystof: darn 2014-04-20T20:50:33Z pkhuong: and R0-R7 only 2014-04-20T20:50:53Z Krystof: but wait, I see (inst adcs result a b) 2014-04-20T20:50:57Z pkhuong: this might just be 64 bit. 2014-04-20T20:50:59Z Krystof: in code that was not written by me 2014-04-20T20:51:10Z pkhuong: Krystof: lucky targeting? 2014-04-20T20:53:54Z pkhuong: so, some version of cortex have a hardware multiply and add 2014-04-20T20:57:30Z fiveop: Krystof: my version https://gist.github.com/anonymous/bf1be0da3961e3690ff1 2014-04-20T20:57:56Z Krystof: see, that's just moon language 2014-04-20T20:57:58Z stassats: does it work? or is it just think-hard version? 2014-04-20T20:58:35Z fiveop: the latter 2014-04-20T20:58:58Z pkhuong: Krystof: umull only works on -M ARMs, it seems 2014-04-20T20:59:02Z fiveop: But it uses umlal (unsigned multiply and accumulate long) 2014-04-20T20:59:08Z pkhuong: but they also have UMLAL 2014-04-20T20:59:10Z pkhuong: righ 2014-04-20T20:59:21Z milosn quit (Remote host closed the connection) 2014-04-20T20:59:30Z fiveop: I currently look at http://www.ic.unicamp.br/~ranido/mc404/arm/arm-instructionset.pdf 2014-04-20T20:59:53Z fiveop: I have the proper specs here, but 60 pages is easier to navigate than 2xxx pages 2014-04-20T21:00:38Z Krystof: it looks plausible. I don't see what the temporaries add to the vop, and I have no real intuition over the lifetimes 2014-04-20T21:00:57Z slyrus joined #sbcl 2014-04-20T21:01:24Z Krystof: but yes, if we have umlal let's use it 2014-04-20T21:01:26Z fiveop: Me neither, even in general (other sbcl platforms) 2014-04-20T21:01:36Z milosn joined #sbcl 2014-04-20T21:01:54Z fiveop: But doesn't (move ...) cause the compiler to only move if it is necessary. 2014-04-20T21:02:03Z stassats: "try randomly until generates something sane" 2014-04-20T21:02:04Z fiveop: While (inst mov ...) is a move. 2014-04-20T21:02:08Z Krystof: yes, right 2014-04-20T21:02:21Z Krystof: the temporaries might not cost anything, but I don't understand what they're adding 2014-04-20T21:02:25Z stassats: and move can't move immediates 2014-04-20T21:03:02Z pkhuong: there's no point to the temporaries 2014-04-20T21:03:50Z slyrus thought that said "if we have urinal, let's use it". sound advice. 2014-04-20T21:03:51Z pkhuong: everything but carry-in conflicts 2014-04-20T21:04:54Z fiveop: https://gist.github.com/anonymous/bbebe7988d4e8d0aa7b0 2014-04-20T21:05:06Z stassats: is eor better than moving in a 0? 2014-04-20T21:05:41Z pkhuong: right. seems like both would be the same # of bytes. 2014-04-20T21:05:53Z stassats: and latency wise? 2014-04-20T21:06:31Z fiveop: do we have a null reg on arm? 2014-04-20T21:06:59Z stassats: shouldn't mov r, #0 do the right thing? 2014-04-20T21:07:01Z Krystof: we have a NULL but not a ZERO 2014-04-20T21:07:38Z stassats: and i'm running with the above paste 2014-04-20T21:07:53Z Krystof: there's the four-arg version yet 2014-04-20T21:07:58Z pkhuong: it's wrong though. incoming fix. 2014-04-20T21:08:18Z pkhuong: (says the dev without any arm hardware) 2014-04-20T21:08:35Z davazp quit (Read error: Connection reset by peer) 2014-04-20T21:09:00Z stassats: arm hardware only avoids having to think wrong things too hard 2014-04-20T21:09:27Z Krystof: stassats: you might also need digit-0-or-plus 2014-04-20T21:09:29Z Krystof: (inst cmp digit 0) 2014-04-20T21:09:31Z Krystof: (inst b (if not-p :ge :lt) target) 2014-04-20T21:09:32Z Krystof: I think 2014-04-20T21:09:45Z Krystof: no! backwards 2014-04-20T21:10:04Z Krystof: (inst cmp digit 0) 2014-04-20T21:10:05Z Krystof: (inst b (if not-p :lt :ge) target 2014-04-20T21:10:07Z Krystof: ) 2014-04-20T21:10:29Z stassats: ok, it hangs anyhow and i better go to sleep instead 2014-04-20T21:10:34Z Krystof: pkhuong is clearly thinking hard 2014-04-20T21:10:54Z pkhuong: https://gist.github.com/pkhuong/11125384 2014-04-20T21:11:57Z pkhuong: the nice thing is, we could use our "free" shift to fixnumize/unfixnumize (: 2014-04-20T21:13:43Z Krystof: I think it's lovely that this vop is laughing at me 2014-04-20T21:13:44Z Krystof: "hi hi hi" 2014-04-20T21:15:17Z fiveop: :) 2014-04-20T21:16:40Z fiveop: pkhuong: could you explain to me, why that works and the other shouldn't? 2014-04-20T21:16:57Z stassats: and digit-0-or-plus is implemented in lisp 2014-04-20T21:17:45Z Krystof: so it is 2014-04-20T21:18:59Z Krystof: so, 4-arg version: 2014-04-20T21:19:04Z Krystof: (inst adds lo prev carry-in) 2014-04-20T21:19:04Z Krystof: (inst mov :cs hi 1) 2014-04-20T21:19:04Z Krystof: (inst mov :cc hi 0) 2014-04-20T21:19:05Z Krystof: (inst umlal lo hi x y) 2014-04-20T21:19:23Z fiveop: https://gist.github.com/anonymous/b8d59639521da2176fb9 2014-04-20T21:19:28Z Krystof: if I have correctly understood what these things are doing 2014-04-20T21:19:58Z fiveop: I think I like your version better :) 2014-04-20T21:21:32Z pkhuong: Krystof: looking for a 1-instruction way to get the carry bit 2014-04-20T21:21:58Z Krystof: we could add 0 to hi with carry 2014-04-20T21:22:07Z Krystof: ? 2014-04-20T21:22:19Z pkhuong: yeah, but not sure I see the point 2014-04-20T21:22:20Z Krystof: oh, no re_zero 2014-04-20T21:22:23Z fiveop: but then hi needs to be 0 2014-04-20T21:22:32Z Krystof: yeah, we don't have a handy zero 2014-04-20T21:22:59Z Krystof: we could add minus NIL to reg_NIL with carry :-) 2014-04-20T21:23:19Z fiveop: since nyef used the same two mov instructions in sub w borrow and add w carry, I'd leave it like that for now :) 2014-04-20T21:24:03Z pkhuong: we could also zero out hi and only use a single conditional move 2014-04-20T21:24:19Z Krystof: rsbs 2014-04-20T21:24:28Z Krystof: no, nonsense 2014-04-20T21:24:50Z Krystof: ok, let's see what happens 2014-04-20T21:24:50Z fiveop: full vop https://gist.github.com/anonymous/9937aea986a07b4cfab2 does that look better this time, with respect to lifetimes? 2014-04-20T21:25:30Z pkhuong: https://gist.github.com/pkhuong/11125384 ? 2014-04-20T21:25:53Z stassats: i'm no longer thinking, just building 2014-04-20T21:26:26Z pkhuong: fiveop: it does 2014-04-20T21:26:28Z fiveop: pkhuong: how is that better than two conditional moves? 2014-04-20T21:27:26Z pkhuong: fiveop: conditional moves are a liability on out of order uarch 2014-04-20T21:27:36Z pkhuong: but we're really not there yet. 2014-04-20T21:29:04Z stassats: still hangs 2014-04-20T21:29:06Z pkhuong: Re lifetime, the problem with https://gist.github.com/anonymous/bbebe7988d4e8d0aa7b0 is that hi/lo need to conflict with x/y, and hi must conflict with carry-in 2014-04-20T21:31:23Z fiveop: thanks. now, I get what life times are actually for :) 2014-04-20T21:32:11Z pkhuong: it would probably be easier to define a conflict graph than to try and devise a set of lifetime ranges that does what we want 2014-04-20T21:32:16Z pkhuong: but this is what we have. 2014-04-20T21:33:03Z pkhuong: nyef has a document that sketches out how the phases relate to one another 2014-04-20T21:33:17Z Krystof: hm 2014-04-20T21:33:23Z Krystof: we might just have run out of registers 2014-04-20T21:33:47Z pkhuong: out of unboxed regs, yes. 2014-04-20T21:34:01Z pkhuong: carry-in could be boxed. 2014-04-20T21:35:19Z Krystof: the message I get says that NFP is component-live 2014-04-20T21:35:38Z pkhuong: can't pack component live? 2014-04-20T21:35:42Z fiveop: "ocfp nfp nargs nl2 nl3" are our non-descriptor regs 2014-04-20T21:36:24Z Krystof: unable to pack a Load-TN in SC UNSIGNED-REG for the third argument to 2014-04-20T21:36:25Z Krystof: the SB!VM::BIGNUM-MULT-AND-ADD-4-ARG VOP 2014-04-20T21:36:28Z pkhuong: well then, we're screwed. 2014-04-20T21:36:31Z Krystof: 9: t2[NFP] (component live) 2014-04-20T21:37:05Z pkhuong: we clearly need 4 unsigned regs for mulal 2014-04-20T21:37:54Z Krystof: "RdLo, RdHi, and Rm must all be different registers." 2014-04-20T21:38:06Z Krystof: could we not get away with 3? 2014-04-20T21:38:29Z pkhuong: ... if Rn = 0 or 1 2014-04-20T21:40:36Z fiveop: I only see "RdLo, RdHi must be different registers." 2014-04-20T21:41:02Z pkhuong: it doesn't matter. No collision between the 4 arguments makes sense. 2014-04-20T21:41:05Z stassats: smart me, not running make-host-1 2014-04-20T21:41:50Z pkhuong: It's RdHi:RdLo += Rn * Rm. All four values are read. 2014-04-20T21:42:55Z stassats: another way to get !random-cold-init to work is to implement modular arithmetic 2014-04-20T21:43:00Z fiveop: But we could split it into RdHi:RdLo = Rn(x) * Rm(y), RdHi:RdLo += prev + carry-in 2014-04-20T21:43:13Z pkhuong: yes we could. Not with umlal. 2014-04-20T21:43:19Z fiveop: no not with umlal :) 2014-04-20T21:43:56Z fiveop: Why does it work on my machine? 2014-04-20T21:44:05Z Krystof: "work" 2014-04-20T21:44:07Z fiveop: /returning from !RANDOM-COLD-INIT 2014-04-20T21:44:07Z fiveop: 2014-04-20T21:44:18Z stassats: ok, i get past multiply-bignum-and-fixnum 2014-04-20T21:44:37Z stassats: to be faulted at -4 during !package-cold-init 2014-04-20T21:44:39Z fiveop: /about to loop over *!INITIAL-SYMBOLS* to make packages 2014-04-20T21:44:39Z fiveop: CORRUPTION WARNING in SBCL pid 1400: 2014-04-20T21:44:39Z fiveop: 2014-04-20T21:44:42Z stassats: what's whit those -4 faults 2014-04-20T21:44:50Z Krystof: if you manage not to have register conflicts, you might get the wrong answer form bignum-mult-and-add 2014-04-20T21:45:00Z Krystof: doesn't matter much for initializing the random state vector 2014-04-20T21:45:10Z stassats: fiveop: what's your fault address? 2014-04-20T21:45:36Z Krystof: any bignums will do 2014-04-20T21:45:40Z fiveop: -4 :) 2014-04-20T21:45:55Z stassats: it's magical 2014-04-20T21:46:15Z Krystof: ok, it's well past my bedtime 2014-04-20T21:46:16Z fiveop: Krystof: but we have 5 non-descriptor regs. even if nfp is component life, there are 4 left. 2014-04-20T21:46:32Z Krystof: I expect to be able to multiply 2 and 3 tomorrow 2014-04-20T21:46:40Z fiveop: can you post your whole vops ? 2014-04-20T21:46:48Z fiveop: before you go to sleep :) 2014-04-20T21:47:06Z pkhuong: ocfp is probably component live too. 2014-04-20T21:47:20Z jdz_ quit (Quit: Leaving...) 2014-04-20T21:47:31Z Krystof: ocfp was "second result load tn" 2014-04-20T21:47:39Z Krystof: my vops are pkhuong's vops with less style 2014-04-20T21:48:04Z pkhuong: we might be able to do something by letting carry-in be a fixnum 2014-04-20T21:48:29Z pkhuong: and untagging via the free shift. 2014-04-20T21:48:30Z Krystof: http://paste.lisp.org/display/142226 2014-04-20T21:49:44Z fiveop: carry-in doesn't have lo as target, no idea whether that's a problem 2014-04-20T21:50:20Z pkhuong: fiveop: targeting shouldn't make us completely fail at packing. 2014-04-20T21:51:14Z fiveop: How can he get the errors, while stassats and I don't? 2014-04-20T21:51:44Z stassats: i used pkhuong's vops 2014-04-20T21:54:53Z Krystof: see, told you 2014-04-20T21:55:04Z Krystof: no style 2014-04-20T21:55:20Z Krystof: sheesh my study is full of blinkenlights now 2014-04-20T21:55:25Z Krystof: sleep time 2014-04-20T21:56:24Z fiveop: gn8 2014-04-20T21:56:53Z fiveop: it was fun, thank you for all the explanations 2014-04-20T21:56:56Z fiveop quit (Quit: bye) 2014-04-20T21:57:20Z stassats: i get the correct results from multiply-bignum-and-fixnum 2014-04-20T21:58:23Z stassats: the package thing is for the morrow 2014-04-20T21:58:47Z davazp joined #sbcl 2014-04-20T21:58:58Z pkhuong: http://paste.lisp.org/display/142226#2 ? 2014-04-20T21:59:25Z pkhuong: also, the variant above will probably work better with our representation selection. 2014-04-20T22:00:57Z pkhuong: Krystof's version of -4-arg needs 6 untagged registers (: 2014-04-20T22:01:13Z pkhuong: (because of really conservative/arguably buggy lifetimes) 2014-04-20T22:03:49Z pkhuong: http://paste.lisp.org/display/142226#3 2014-04-20T22:05:39Z stassats quit (Ping timeout: 240 seconds) 2014-04-20T22:08:24Z pkhuong: and.. nope. the fourth argument is usually not a fixnum 2014-04-20T22:10:59Z nyef joined #sbcl 2014-04-20T22:11:04Z pkhuong: add-with-carry/sub-with-borrow will benefit though. 2014-04-20T22:11:09Z nyef: Hello all. 2014-04-20T22:15:36Z nyef: ... OCFP can't be component-live, it's the P-A flag register. 2014-04-20T22:16:14Z nyef: (I now know how to do P-A without requiring a specific flag register, but it requires a conditional branch. Win some, lose some.) 2014-04-20T22:39:56Z davazp quit (Remote host closed the connection) 2014-04-20T22:42:52Z slyrus quit (Ping timeout: 245 seconds) 2014-04-20T22:46:20Z leuler quit (Quit: ERC Version 5.1.2 $Revision: 1.796.2.6 $ (IRC client for Emacs)) 2014-04-20T22:58:36Z slyrus joined #sbcl 2014-04-20T23:00:40Z ehaliewicz joined #sbcl 2014-04-20T23:20:42Z slyrus quit (Ping timeout: 240 seconds) 2014-04-20T23:47:42Z zRecursive joined #sbcl 2014-04-20T23:53:02Z nyef: Hrm. Five non-descriptor-regs, only one of which really should be component-live... 2014-04-20T23:54:57Z ltbarcly joined #sbcl 2014-04-21T00:36:03Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-21T00:37:31Z krzysz00 joined #sbcl 2014-04-21T00:43:48Z michael_lee joined #sbcl 2014-04-21T01:58:18Z nyef quit (Quit: G'night all.) 2014-04-21T02:27:50Z ehaliewicz quit (Remote host closed the connection) 2014-04-21T02:31:30Z ltbarcly joined #sbcl 2014-04-21T02:38:53Z christoph_debian quit (Ping timeout: 252 seconds) 2014-04-21T02:51:59Z christoph_debian joined #sbcl 2014-04-21T03:07:51Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-21T03:20:18Z ltbarcly joined #sbcl 2014-04-21T03:33:08Z edgar-rft joined #sbcl 2014-04-21T03:37:54Z drmeister joined #sbcl 2014-04-21T03:39:32Z ltbarcly quit (Ping timeout: 245 seconds) 2014-04-21T03:41:42Z ltbarcly joined #sbcl 2014-04-21T03:56:47Z LiamH quit (Quit: Leaving.) 2014-04-21T04:20:29Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-21T04:52:04Z ltbarcly joined #sbcl 2014-04-21T04:54:40Z attila_lendvai joined #sbcl 2014-04-21T04:54:40Z attila_lendvai quit (Changing host) 2014-04-21T04:54:40Z attila_lendvai joined #sbcl 2014-04-21T04:55:18Z ltbarcly quit (Client Quit) 2014-04-21T05:08:35Z slyrus joined #sbcl 2014-04-21T05:12:30Z pranavrc joined #sbcl 2014-04-21T05:12:30Z pranavrc quit (Changing host) 2014-04-21T05:12:30Z pranavrc joined #sbcl 2014-04-21T05:25:06Z krzysz00 quit (Ping timeout: 240 seconds) 2014-04-21T05:32:21Z krzysz00 joined #sbcl 2014-04-21T05:40:04Z michael_lee quit (Quit: Ex-Chat) 2014-04-21T05:49:57Z angavrilov joined #sbcl 2014-04-21T06:00:31Z scymtym_ quit (Ping timeout: 250 seconds) 2014-04-21T06:04:27Z krzysz00 quit (Ping timeout: 276 seconds) 2014-04-21T06:27:14Z ltbarcly joined #sbcl 2014-04-21T06:39:06Z tmh_ quit (Ping timeout: 240 seconds) 2014-04-21T06:47:21Z tmh_ joined #sbcl 2014-04-21T07:32:33Z zRecursive quit (Remote host closed the connection) 2014-04-21T07:49:06Z oleo quit (Ping timeout: 276 seconds) 2014-04-21T07:55:29Z fiveop joined #sbcl 2014-04-21T07:57:32Z stassats joined #sbcl 2014-04-21T08:01:34Z oleo joined #sbcl 2014-04-21T08:04:38Z stassats` joined #sbcl 2014-04-21T08:07:38Z stassats quit (Ping timeout: 240 seconds) 2014-04-21T08:21:56Z |3b|: should sbcl WARN about (lambda (a) (typecase a ((or (array * (* * 3)) (array * (* * 4))) (case (array-rank a) (2 (aref a 1 2)))))) ? 2014-04-21T08:22:36Z |3b|: with a simpler type, it figures out the AREF is dead code and doesn't warn 2014-04-21T08:23:29Z |3b|: (simplified version of code in opticl) 2014-04-21T08:37:47Z |3b|: looks like the deftransform for array-rank could be a bit smarter and calculate the rank more directly instead of using array-type-dimensions-or-give-up, though that wouldn't help if the array types were different ranks 2014-04-21T08:41:13Z |3b|: though possibly it could still restrict the type of the value it returns enough to avoid that branch of the case? 2014-04-21T09:02:05Z fiveop: Krystof: It's 35 bytes from the first address in the map to the start of the functions assembler code. 2014-04-21T09:02:38Z fiveop: +' 2014-04-21T09:02:45Z stassats`: why isn't the map formatted that way then? 2014-04-21T09:20:13Z stassats`: fiveop: do you use slam.sh? 2014-04-21T09:20:19Z stassats`: i can't figure it to work 2014-04-21T09:25:08Z fiveop: no I don't 2014-04-21T09:25:36Z fiveop: I think it does target1 and host2 and I don't want to run them on the same machine 2014-04-21T09:26:18Z stassats`: i disabled running targe-1, but i figured the host sbcl can't really load a target cold core 2014-04-21T09:27:19Z stassats`: if that's what slam does 2014-04-21T09:29:18Z stassats`: it looks like find-package is broken 2014-04-21T09:29:20Z |3b|: seems like that works for union types, not sure if intersection types could be a problem or not 2014-04-21T09:29:48Z stassats`: which means hashing, which means lack of modarith, which means, bignum woes, there they come 2014-04-21T09:30:51Z fiveop: I'm not sure 2014-04-21T09:31:06Z kludge` quit (Ping timeout: 240 seconds) 2014-04-21T09:31:19Z stassats`: hm, no, speeding up make-host-2.sh should still be possible, compile changed files, then dump the core 2014-04-21T09:34:46Z stassats`: right, (gethash string *package-names*) fails 2014-04-21T09:36:17Z kludge` joined #sbcl 2014-04-21T09:37:02Z fiveop: How did you figure that out? 2014-04-21T09:37:31Z stassats`: ok, got slam.sh to work, 6 seconds compared with 1 minute 2014-04-21T09:37:48Z stassats`: and it takes me 5 seconds to sync the core back 2014-04-21T09:37:58Z stassats`: fiveop: show0 2014-04-21T09:42:09Z stassats`: 10 second turn-around time is much nicer 2014-04-21T10:04:20Z stassats`: could with-pinned-objects be a problem? 2014-04-21T10:05:56Z stassats`: with-hashtable-locks certainly is a problem 2014-04-21T10:11:24Z milosn quit (Remote host closed the connection) 2014-04-21T10:13:15Z stassats`: the problem is in call-with-recursive-system-lock 2014-04-21T10:13:53Z stassats`: and probably by extension, with without-interrupts 2014-04-21T10:14:53Z milosn joined #sbcl 2014-04-21T10:19:59Z stassats`: without-interrupts it is 2014-04-21T10:42:06Z stassats`: now it looks like an unwind-protect problem 2014-04-21T10:48:50Z stassats` quit (Ping timeout: 240 seconds) 2014-04-21T10:54:11Z nyef joined #sbcl 2014-04-21T10:54:24Z nyef: G'morning all. 2014-04-21T10:57:08Z nyef: ... 35 bytes from map file to code start doesn 2014-04-21T10:57:11Z nyef: 't seem right. 2014-04-21T10:57:41Z nyef: At all. Isn't the real number 23? 2014-04-21T11:00:06Z |3b| supposes if i were smart i'd have copied down error messages for a bug report before hacking my copy of sbcl to avoid the problem :p 2014-04-21T11:01:00Z krzysz00 joined #sbcl 2014-04-21T11:19:25Z cmack`` joined #sbcl 2014-04-21T11:21:01Z cmack` quit (Ping timeout: 252 seconds) 2014-04-21T11:25:17Z fiveop: nyef: 35 = 0x23 2014-04-21T11:25:52Z nyef: Decimal 23. 2014-04-21T11:26:01Z jdz joined #sbcl 2014-04-21T11:26:27Z nyef: #define SIMPLE_FUN_CODE_OFFSET 23 2014-04-21T11:26:29Z fiveop: I think I didn't include the compute-lra 2014-04-21T11:26:39Z fiveop: that's three instructions 2014-04-21T11:26:41Z fiveop: 12 bytes 2014-04-21T11:26:46Z fiveop: ups 2014-04-21T11:27:53Z fiveop quit (Quit: bye) 2014-04-21T11:29:25Z nyef: ... I hope he meant the compute-code. 2014-04-21T11:33:12Z nyef: Ahh, a fresh, hot cup of tea. Now I just need an atomic vector plotter and an improbability drive. (-: 2014-04-21T11:44:12Z |3b|: https://bugs.launchpad.net/sbcl/+bug/1310574 2014-04-21T11:44:25Z |3b|: and now it is time for sleep, or at least things requiring less thinking :) 2014-04-21T11:53:09Z nyef hopes that he defined the various multiply instructions correctly. 2014-04-21T12:06:34Z nyef: Okay, BIGNUM-MULT-AND-ADD-4-ARG with four UNSIGNED-REG inputs worked first try, and now I get a memory-fault at -4 (pc=-4) just after "/about to loop over *!INITIAL-SYMBOLS* to make packages". 2014-04-21T12:06:50Z nyef: That means I'm caught back up, doesn't it? 2014-04-21T12:25:22Z hlavaty joined #sbcl 2014-04-21T12:25:32Z pkhuong: seems so 2014-04-21T12:25:55Z pkhuong: it'd likely be a good idea to sync up the various branches just so everyone gets the same error (: 2014-04-21T12:27:03Z nyef: I'm running what I've just pushed to my public-repo plus a #!-arm on the definition of ERROR and an alternate trap generation scheme for the undefined-tramp. 2014-04-21T12:27:15Z nyef: And I've merged the four new commits from fiveop's repo. 2014-04-21T12:32:43Z pkhuong: pretty sure I prefer eor/mov cs to mov cc/mov cs ;) 2014-04-21T12:32:56Z pkhuong: mm.. old school conditional VOPs, nyef? 2014-04-21T12:33:33Z nyef: Which is the "new school"? 2014-04-21T12:33:54Z pkhuong: (:conditional fe) 2014-04-21T12:33:59Z pkhuong: * (:conditional :ge) 2014-04-21T12:34:54Z pkhuong: and a branch-if VOP that can interpret (:ge)... and eventually conditional move VOPs. 2014-04-21T12:35:16Z nyef: Ah, okay. I cribbed from the SPARC backend. 2014-04-21T12:35:19Z drmeister quit (Remote host closed the connection) 2014-04-21T12:35:29Z psilord quit (Quit: Leaving.) 2014-04-21T12:36:28Z nyef: Which is odd, doesn't SPARC also have conditional move instructions? 2014-04-21T12:36:30Z pkhuong: I only ported x86oids over, but I'd say that ARM is a perfect candidate for conditional move insertion (and other kinds of predication that the new form of conditionals enables) 2014-04-21T12:36:44Z nyef: ARM is, absolutely. 2014-04-21T12:37:22Z nyef: More items for the list, I guess. 2014-04-21T12:39:31Z pkhuong: New school dates from "Sun Jan 11 18:33:31 2009" so way after most real work on the sparc backend. 2014-04-21T12:39:56Z nyef: We might want to switch most of the backends to the new-style conditional VOPs, even if they don't have predicated MOVEs, to well and truly set the precedent. 2014-04-21T12:40:03Z pkhuong: See 5cf3c4259d529e180d75d4d140f344e600d2b06b, d95f1e6476aa63695e018a7769a1ae9e002fca36 2014-04-21T12:41:03Z pkhuong: I think the main impetus was to get generic predicates to emit sane code, instead of [static call] [cmp with NIL] jcc 2014-04-21T12:42:08Z nyef: Hmm. 2014-04-21T12:42:25Z nyef: Not a right-now thing, but a future-optimization thing, I think. 2014-04-21T12:43:03Z pkhuong: could be incremental too. 2014-04-21T12:43:20Z pkhuong: Once you have branch-if, it's easy to opt in for new conditionals 2014-04-21T12:43:34Z pkhuong: or even only the special cases that branch-if handles. 2014-04-21T12:45:14Z stassats` joined #sbcl 2014-04-21T12:45:36Z nyef: Right, right. But still a later-optimization thing. 2014-04-21T12:46:00Z drmeister joined #sbcl 2014-04-21T12:46:00Z stassats`: so, how likely is it that unwind-protect is broken on arm? 2014-04-21T12:46:16Z nyef: I wouldn't've expected it to be broken. 2014-04-21T12:48:42Z nyef: Admittedly, 8e54e3a255469090a366aa164c6f7d9a2649fc09 was from late-October 2012, but I'm fairly sure I had some simple test cases going at the time. 2014-04-21T12:49:29Z eudoxia joined #sbcl 2014-04-21T12:50:22Z stassats`: ok, simplified form down to unwind-protect does fail 2014-04-21T12:50:27Z nyef: And bda2968f9aa55d4135538a8a60947be0bd3d330c appears to have my original UWP test case... 2014-04-21T12:50:28Z stassats`: unwinding 2014-04-21T12:51:29Z stassats`: http://paste.lisp.org/display/142229 2014-04-21T12:52:07Z nyef: That's... Hunh. 2014-04-21T12:52:36Z nyef: All in the same file? 2014-04-21T12:53:06Z stassats`: i added it into cold-init, the call is from !cold-init 2014-04-21T12:53:47Z ltbarcly quit (Quit: Textual IRC Client: www.textualapp.com) 2014-04-21T12:54:31Z stassats`: the original error is a bit different, a memfault at #xfffffffc 2014-04-21T12:54:54Z nyef: Yeah, but that'd be contextual anyway if it were something uninitialized. 2014-04-21T12:57:51Z drmeister quit (Ping timeout: 276 seconds) 2014-04-21T12:59:07Z pkhuong: nyef: lisp stack grows up? 2014-04-21T12:59:30Z nyef: Yes, which is one of the reasons that gdb can easily break things horribly. 2014-04-21T13:00:06Z nyef: If I had another register, I'd spend it to get a separate lisp stack pointer. 2014-04-21T13:00:21Z stassats`: blimey, paste too large 2014-04-21T13:00:27Z pkhuong: and there's no biasing of TN offset, unlike x86? 2014-04-21T13:00:31Z stassats`: tried to annotate with a trace file 2014-04-21T13:00:47Z nyef: Shouldn't be a bias, IIRC. 2014-04-21T13:00:59Z nyef: Check LOAD-STACK or whatever it is in macros. 2014-04-21T13:01:55Z stassats`: http://common-lisp.net/~sboukarev/trace.txt 2014-04-21T13:06:06Z stassats`: i can't find an array operation in there 2014-04-21T13:06:55Z nyef: If we're taking an undirected control transfer, there wouldn't have to be. 2014-04-21T13:07:37Z stassats`: i should probably move that function into another file to get a clearer trace 2014-04-21T13:07:59Z pkhuong: nlx-entry-multiple is suspect. 2014-04-21T13:08:18Z pkhuong: ";; Reset the CSP. / (inst add csp-tn result num)" 2014-04-21T13:08:29Z nyef: pkhuong: Not applicable: This is the "clean" unwind case, not the NLX unwind case. 2014-04-21T13:08:53Z nyef: stassats`: Disable cold-init in boot order, add the :arm-bootstrap flag to the arm-boostrap file (next-to-last in the order or so). 2014-04-21T13:11:36Z nyef: I'm at this point myself, and am just waiting out a full host-1 / host-2 cycle (partly to get the after-xc core so that I can slam). 2014-04-21T13:12:01Z stassats`: do i need to run host-1 after modifying build-order? 2014-04-21T13:12:20Z nyef: I don't believe so, no, but you might need to run host-2. 2014-04-21T13:12:34Z stassats`: a full host-2? i'm slamming currently 2014-04-21T13:12:49Z nyef: ISTR having hacked slam to pick up on build-order, but I don't remember when, in which tree, or if I committed it upstream. 2014-04-21T13:13:23Z nyef: Ah, yeah, you should be able to slam. 2014-04-21T13:13:42Z nyef: That was 1.0.33.10. 2014-04-21T13:14:16Z pkhuong: OK. Off to make ad delivery quicker and more reliable. I should have a real arm box in january ;) 2014-04-21T13:14:37Z nyef: What constitutes a "real arm box"? 2014-04-21T13:15:20Z pkhuong: 1.2 GHz with debian out of the box. 2014-04-21T13:15:47Z nyef: So... still another bare-board machine? 2014-04-21T13:15:51Z stassats`: january? 2014-04-21T13:16:04Z stassats`: january is, like, next year 2014-04-21T13:16:31Z pkhuong: nyef: yeah. except with an fpga. 2014-04-21T13:16:44Z jsnell_: bunnie's laptop? 2014-04-21T13:16:53Z pkhuong: jsnell_: yes (: 2014-04-21T13:17:36Z pkhuong: stassats`: I'm not exactly rich in $$ and now poor in free time. 2014-04-21T13:18:18Z jsnell_: neat. though I have to say it looked a bit impractical with the screen on the wrong side of the lid :-) 2014-04-21T13:18:24Z stassats`: strange, i get a trace for a file i added before cold-init, but it complains about undefined function when i call the function defined in it from !cold-init 2014-04-21T13:18:49Z drmeister joined #sbcl 2014-04-21T13:19:16Z pkhuong: jsnell_: it does. more like a luggable than a laptop. 2014-04-21T13:19:20Z nyef: Okay, now I get "/UNWINDT-TEST" followed by a fault at -4. 2014-04-21T13:19:22Z stassats`: at least i have a trace file 2014-04-21T13:19:40Z nyef: And it's the ONLY thing in cold-init, so I can step it with GDB. (-: 2014-04-21T13:20:50Z stassats`: complete with a typo! 2014-04-21T13:21:24Z nyef: Yes, but it's your typo. d-: 2014-04-21T13:21:40Z stassats`: yep 2014-04-21T13:21:51Z pranavrc quit 2014-04-21T13:21:57Z stassats`: i need to get back to the original memory fault too 2014-04-21T13:21:57Z drmeister quit (Read error: Connection reset by peer) 2014-04-21T13:25:16Z prxq joined #sbcl 2014-04-21T13:25:22Z stassats`: slam.sh still doesn't fully pick up my build-order changes 2014-04-21T13:25:41Z stassats`: i'll just wait for make-host-2 then 2014-04-21T13:27:52Z stassats`: ok, -4 fault here too 2014-04-21T13:32:55Z attila_lendvai quit (Quit: Leaving.) 2014-04-21T13:33:59Z nyef: I think I may have found it. 2014-04-21T13:35:27Z nyef: Have a look for uses of compute-lra in compiler/arm/call, and see if you can not-see what I'm not seeing. 2014-04-21T13:36:14Z drmeister joined #sbcl 2014-04-21T13:38:30Z nyef: (Hint: It'd be the following line.) 2014-04-21T13:40:08Z stassats`: well, i'm not really familiar with LRA 2014-04-21T13:40:37Z nyef: In three out of four cases, it's immediately stashed to the stack. 2014-04-21T13:40:55Z nyef: Guess which case is being involved in UWP? 2014-04-21T13:41:12Z stassats`: the same thing happens on ppc 2014-04-21T13:41:47Z nyef: That'd be more frightening if PPC used the same return conventions. 2014-04-21T13:43:09Z psilord joined #sbcl 2014-04-21T13:43:15Z nyef: So, in a way, we both were right about this one: It's triggered by unwind-protect, and it's not unwind-protect itself. 2014-04-21T13:44:19Z nyef: CATS. CATS ARE NICE. SO ARE FIXED CALL VOPS. 2014-04-21T13:46:28Z nyef: And pushed. 2014-04-21T13:46:39Z stassats`: well, when i add a store-to-stack, it reaches primitive-halt successfully 2014-04-21T13:47:13Z nyef: Doing another host-2 to see where cold-init breaks now. (-: 2014-04-21T13:48:02Z stassats`: ok, you also have bignum vops from the other night, i'll pull now 2014-04-21T13:48:54Z nyef: Not quite the VOPs from yesterday, AIUI, but yeah. 2014-04-21T13:49:58Z stassats`: next what should fail is make-package 2014-04-21T13:50:08Z nyef: Reason being? 2014-04-21T13:50:36Z stassats`: it failed when i removed that unwind-protect from the cold-init 2014-04-21T13:50:42Z nyef: Ah. 2014-04-21T13:51:10Z stassats`: and there it is 2014-04-21T13:51:51Z stassats`: make-or-remake-package-hashtable => two-arg-= 2014-04-21T13:53:04Z stassats`: a type error, something is not a number 2014-04-21T13:56:11Z nyef: NIL is not a number. 2014-04-21T14:00:37Z stassats`: the default implementation of /hexstr is not very useful 2014-04-21T14:01:05Z stassats`: i made it to return all the values it receives and print the first one, easier to wrap things around 2014-04-21T14:04:01Z zacts joined #sbcl 2014-04-21T14:09:43Z stassats`: the error is in (ceiling 0 0.75), is FP already implemented? 2014-04-21T14:10:21Z nyef: It's mostly-implemented. 2014-04-21T14:10:45Z nyef: The complex stuff isn't implemented, and there's probably disabled code scattered about the source tree. 2014-04-21T14:13:12Z nyef: Oh, and every so often I have to fix somethingorother with the instruction definitions. 2014-04-21T14:15:14Z nyef: Umm... And we don't have the floating-point-modes stuff in yet. 2014-04-21T14:19:21Z stassats`: ./slam.sh completes faster than it takes to rsync the core back 2014-04-21T14:20:52Z stassats`: it looks like multiple-value problem again 2014-04-21T14:20:57Z nyef: Fun and games. 2014-04-21T14:21:39Z nyef: If you can narrow down a test case? 2014-04-21T14:21:59Z stassats`: in the process 2014-04-21T14:27:26Z stassats`: it looks like the second return value is NIL, instead of the remainder 2014-04-21T14:28:59Z nyef: From what, and is this a full call or open-coded? 2014-04-21T14:29:23Z stassats`: full call 2014-04-21T14:32:52Z stassats`: nyef: http://paste.lisp.org/display/142229#1 2014-04-21T14:36:30Z nyef: Hrm. So we're looking at RETURN-MULTIPLE or DEFAULT-UNKNOWN-VALUES? 2014-04-21T14:37:03Z stassats`: (defun tt () (values 1 10)) and (/hexstr (nth-value 1 (tt))) is enough 2014-04-21T14:39:11Z nyef: ... Ugh. Not RETURN-MULTIPLE, RETURN. 2014-04-21T14:41:37Z cmack`` is now known as cmack 2014-04-21T14:41:43Z stassats`: the return part seems okish 2014-04-21T14:42:11Z nyef: Yes, it does. 2014-04-21T14:46:27Z stassats`: i think something is wrong with checking the number of returned values 2014-04-21T14:46:37Z nyef: Which side? 2014-04-21T14:46:51Z stassats`: either 2014-04-21T14:48:23Z stassats`: return moves 8 into nargs, so, ok here 2014-04-21T14:48:39Z nyef: How many regs are subsequently overwritten? 2014-04-21T14:48:52Z nyef: Should just be R2. 2014-04-21T14:49:05Z stassats`: r1 is conditionally filled with NIL 2014-04-21T14:49:18Z nyef: On the return side? What? 2014-04-21T14:49:25Z stassats`: on the call side 2014-04-21T14:49:48Z nyef: Ah, okay. 2014-04-21T14:49:55Z stassats`: or it's the wrong call 2014-04-21T14:51:05Z nyef: Should be a MOVNE R1, reg_NULL, a MOVEQ to clear the stack, and then the COMPUTE-CODE? 2014-04-21T14:51:26Z stassats`: this part http://paste.lisp.org/display/142229#2 2014-04-21T14:51:45Z nyef: Yeah, that looks right. 2014-04-21T14:52:31Z nyef: What about the other side? 2014-04-21T14:52:32Z stassats`: which opreation results in NE or EQ flags? 2014-04-21T14:52:39Z nyef: MRS. 2014-04-21T14:52:44Z nyef: In lisp-return in macros. 2014-04-21T14:52:51Z stassats`: ah, the callee side 2014-04-21T14:52:59Z nyef: Err, sorry, MSR. 2014-04-21T14:53:27Z stassats`: MSR 8, 240 2014-04-21T14:53:33Z stassats`: no idea what that means 2014-04-21T14:53:42Z nyef: Right, that'd be for multiple-values. 2014-04-21T14:54:02Z stassats`: there's a SUB operation following, can it overwrite? 2014-04-21T14:54:16Z nyef: No, it'd have to be a SUBS to overwrite. 2014-04-21T14:54:21Z stassats`: or is it a delay slot or something? 2014-04-21T14:54:30Z stassats`: is MSR a return operation? 2014-04-21T14:54:35Z nyef: No delay slots on this architecture. 2014-04-21T14:54:41Z nyef: No, the SUB is the return operation. 2014-04-21T14:54:52Z stassats`: interesting... 2014-04-21T14:54:56Z nyef: MSR is move to system register, in this case the status flags. 2014-04-21T14:55:09Z nyef: The destination of the SUB is the program counter. 2014-04-21T14:55:51Z stassats`: yeah, i see that, interesting way 2014-04-21T14:56:28Z nyef: I'd far rather have a spare GPR, but this at least makes PC-relative addressing easy enough, and there are fun games that can be played with it. 2014-04-21T14:56:52Z stassats`: just wait until ARM-64 is ubiquitous 2014-04-21T14:57:04Z zacts quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2014-04-21T14:57:09Z nyef: ... There's an ARM-64? 2014-04-21T14:57:26Z stassats`: sure 2014-04-21T14:57:48Z nyef: Does it have an ARM-32 compatibility mode, at least? 2014-04-21T14:58:01Z stassats`: i think it does 2014-04-21T14:58:18Z nyef: Clearly, I'm going to need a multi-CPU version in a 1U. 2014-04-21T14:58:35Z stassats`: it doubles the number of registers 2014-04-21T15:01:23Z stassats`: so, is 240 the wrong status? 2014-04-21T15:01:32Z nyef: No, 240 should be correct. 2014-04-21T15:01:43Z stassats`: then, is NE the wrong check? 2014-04-21T15:02:35Z nyef: If you want to double-check, though, I'd step through with gdb and check register contents on entry to the RETURN operation, the status register after the MSR, and the behavior on the caller side after the return. 2014-04-21T15:10:40Z nyef: Worst comes to worst, and the MSR turns out to be the problem (unlikely), we can use TST NULL,NULL and TST NULL,PC to set or clear the Z flag. 2014-04-21T15:14:10Z reb: My Lisp binary is hung in a futex. I connect with gdb. How can I get some indication of what Lisp function my code is hung in? 2014-04-21T15:14:55Z nyef: reb: call lisp_backtrace(10) 2014-04-21T15:15:05Z reb: Thanks! 2014-04-21T15:15:32Z nyef: You'll want to do that from whatever thread is actually stuck. 2014-04-21T15:15:57Z nyef: Or just do it to all of the threads and sort through the mess. 2014-04-21T15:16:21Z stassats`: does #xF0 really sets all status registers? 2014-04-21T15:16:51Z nyef: It should. Feel free to verify by stepping through in gdb. 2014-04-21T15:17:03Z stassats`: i'm stepping through the manual currently 2014-04-21T15:17:30Z krzysz00 quit (Ping timeout: 240 seconds) 2014-04-21T15:18:11Z nyef: ... Maybe it needs to be #xf0000000? 2014-04-21T15:18:25Z stassats`: that's what i'm currently gathering 2014-04-21T15:19:38Z stassats`: i can just try it while i continue reading 2014-04-21T15:21:07Z nyef: I shall do so as well, as it seems very likely. 2014-04-21T15:21:15Z nyef: Does need to be built from host-1, though. 2014-04-21T15:21:23Z stassats`: well, my reading seems to confirm that #xf0000000 is the right constant 2014-04-21T15:22:21Z stassats`: nyef: i already made that mistake yesterday 2014-04-21T15:24:04Z segv- joined #sbcl 2014-04-21T15:26:06Z stassats`: i, for some reason, have an undefined HEXSTR error, but, the context shows that it's called with 10, not NIL 2014-04-21T15:26:11Z stassats`: so, #xf0000000 it is 2014-04-21T15:27:53Z stassats`: right, hexstr is defined in cold-init, which i wasn't loading 2014-04-21T15:30:42Z nyef: Infinite loop. 2014-04-21T15:30:49Z nyef: Still in !PACKAGE-COLD-INIT. 2014-04-21T15:30:51Z stassats`: ok, with that, it know hangs in !package-cold-init, i gather that it's due to another unimplemented vop 2014-04-21T15:31:03Z nyef: Almost certainly. 2014-04-21T15:31:22Z nyef: Shall I prep a commit for the fix to LISP-RETURN? 2014-04-21T15:32:13Z stassats`: well, it seems to work correctly 2014-04-21T15:32:45Z ltbarcly joined #sbcl 2014-04-21T15:33:54Z stassats`: gdb suggests that the loop is in TRUNCATE 2014-04-21T15:33:55Z nyef: And pushed. 2014-04-21T15:38:41Z scymtym_ joined #sbcl 2014-04-21T15:45:51Z ltbarcly quit (Ping timeout: 252 seconds) 2014-04-21T15:46:03Z stassats` keeps mixing up /hexstr0 and /show 2014-04-21T15:47:20Z stassats`: and again, argh 2014-04-21T15:47:24Z ltbarcly joined #sbcl 2014-04-21T15:48:59Z stassats`: ok, it's positive-primep this time 2014-04-21T15:53:51Z stassats`: (truncate 6524 20) is enough 2014-04-21T15:53:54Z stassats`: ;; what? 2014-04-21T15:55:50Z stassats`: no truncate vops whatsoever, that explains 2014-04-21T15:56:34Z nyef: Yeah, that'd do it. 2014-04-21T15:59:26Z stassats`: who's gonna write it? 2014-04-21T15:59:56Z nyef: I'll tackle it in a few hours if nobody has got to it by then. 2014-04-21T16:02:43Z stassats`: fixnum-truncate on ppc is an assembly routine 2014-04-21T16:06:51Z stassats`: wait what, arm doesn't have a division operator? 2014-04-21T16:07:38Z stassats`: then i'm not going to attempt it 2014-04-21T16:09:05Z stassats`: i can write positive-primep without division and move on to the next hurdle 2014-04-21T16:23:12Z stassats`: with a glorious positive-primep (who uses negative primes anyway?), i get stuck at %enter-new-nicknames 2014-04-21T16:31:47Z nyef: ... gethash, maybe? 2014-04-21T16:32:30Z stassats`: puthash 2014-04-21T16:34:03Z stassats`: with-hashtable-locks again 2014-04-21T16:34:23Z stassats`: though if i remove it, i get a control stack error 2014-04-21T16:34:52Z stassats`: this is endless 2014-04-21T16:44:12Z stassats`: got to create COMMON-LISP, and then immediately fault at 1 2014-04-21T16:45:18Z nyef: fault at 1? I take it that there's a different PC this time? 2014-04-21T16:46:31Z stassats`: i think it's still with-something-locks 2014-04-21T16:47:39Z stassats`: i guess i need to figure out what's wrong with locks 2014-04-21T16:48:03Z nyef: Did you have any more changes after the new positive-primep? 2014-04-21T16:48:13Z stassats`: nope 2014-04-21T16:48:18Z nyef: Okay, thanks. 2014-04-21T16:48:37Z stassats`: just purging with-hashtable-locks and the like 2014-04-21T16:48:45Z nyef: Ah. 2014-04-21T16:50:18Z attila_lendvai joined #sbcl 2014-04-21T16:50:19Z attila_lendvai quit (Changing host) 2014-04-21T16:50:19Z attila_lendvai joined #sbcl 2014-04-21T16:50:53Z stassats`: and gethash is alright with with-hashtable-locks, but not puthash 2014-04-21T16:51:22Z nyef: Might be something lower-level, like another VOP working with an uninitialized temporary or something. 2014-04-21T16:52:01Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-21T16:53:17Z nyef: So, want to hear something crazy? I can't find any evidence of division support in the alpha backend other than for floats and that one bignum operation. 2014-04-21T16:53:35Z nyef: Oh, unless it's an assembly-routine. 2014-04-21T16:53:36Z stassats`: not even in assembly routines? 2014-04-21T16:53:58Z nyef: Yeah, that's where it's hiding. 2014-04-21T16:55:57Z slyrus quit (Ping timeout: 240 seconds) 2014-04-21T17:03:00Z ltbarcly joined #sbcl 2014-04-21T17:09:01Z stassats`: cool, the problem is a package lock withing a package lock 2014-04-21T17:09:12Z stassats`: or, without-interrupts within without-interrupts 2014-04-21T17:34:42Z stassats`: ok, nested unwind-protect with special binding 2014-04-21T17:40:30Z stassats`: there it is: http://paste.lisp.org/display/142229#3 2014-04-21T17:42:04Z sdemarre joined #sbcl 2014-04-21T17:43:21Z stassats`: so, is the binding stack guard in the right places, and is unbind-to-here correct? 2014-04-21T17:45:11Z nyef_ joined #sbcl 2014-04-21T17:45:33Z nyef quit (Ping timeout: 250 seconds) 2014-04-21T17:45:57Z ltbarcly_ joined #sbcl 2014-04-21T17:46:05Z nyef_: Ugh. Binding stack guard? Lovely. 2014-04-21T17:46:36Z nyef_: I make no guarantees on that at this point. 2014-04-21T17:46:52Z stassats`: is it possible that unbind-to-here goes the wrong way? 2014-04-21T17:47:04Z sdemarre quit (Read error: Operation timed out) 2014-04-21T17:48:04Z nyef_: unbind-to-here moves down, unbind moves down, bind moves up... 2014-04-21T17:49:03Z ltbarcly quit (Ping timeout: 276 seconds) 2014-04-21T17:51:39Z kanru quit (Ping timeout: 276 seconds) 2014-04-21T17:56:05Z stassats`: no nesting needed, just uwp + special binding 2014-04-21T17:57:11Z nyef_: Does it still apply if the special binding is undone before doing unbind-to-here? 2014-04-21T17:58:23Z stassats`: http://paste.lisp.org/display/142229#4 2014-04-21T17:58:42Z ltbarcly joined #sbcl 2014-04-21T17:59:33Z nyef_: That's really odd. 2014-04-21T17:59:38Z ltbarcly_ quit (Read error: Operation timed out) 2014-04-21T18:00:15Z nyef_: And it's not just the binding that triggers it? 2014-04-21T18:00:23Z nyef_: It's the combination of hte binding and an unbind-to-here? 2014-04-21T18:01:44Z attila_lendvai quit (Quit: Leaving.) 2014-04-21T18:02:30Z stassats`: right, a bit strange, since by the time unbind-to-here runs it is already unbound 2014-04-21T18:02:51Z stassats`: maybe it just touches the binding stack in a way that makes unbind-to-here break 2014-04-21T18:03:15Z nyef_: Shouldn't break if the binding stack pointer is effectively unchanged. 2014-04-21T18:04:27Z nyef_: load-symbol-value and store-symbol-value look good. 2014-04-21T18:04:57Z stassats`: wrapping the whole unwind-protect in (let (*) ...) doesn't fail 2014-04-21T18:05:09Z nyef_: So it's binding within the UWP? 2014-04-21T18:05:52Z stassats`: putting (let (*) (/show0)) before the whole thing doesn't fail either 2014-04-21T18:06:25Z stassats`: neither is wrapping the cleanup form problematic 2014-04-21T18:07:00Z stassats`: nor is wrapping the call to the local FOO in (let (*)) 2014-04-21T18:07:32Z stassats`: only when the body of FOO is wrapped does it fail 2014-04-21T18:09:43Z nyef_ sighs. Trace-file? 2014-04-21T18:09:57Z stassats`: it even fails when (foo) is not a local function 2014-04-21T18:10:21Z stassats`: trace file again too large for paste.lisp 2014-04-21T18:11:32Z nyef_: Maybe I should just replicate locally, that way I can also step through with gdb, which forces me to actually look at what's going on at each instruction. 2014-04-21T18:12:08Z stassats`: http://common-lisp.net/~sboukarev/trace.txt 2014-04-21T18:12:12Z stassats`: (may need to refresh) 2014-04-21T18:22:59Z stassats`: i guess i've had enough of cold-init for the day 2014-04-21T18:23:09Z nyef_: Hrm. The binding and unbinding stuff looks reasonable. It might be that it isn't reasonable, or that there's something corrupting something, somewhere, or... I don't know. 2014-04-21T18:23:36Z nyef_: I'll take a deeper look tonight, with a test case this small it should be easy to step through. 2014-04-21T18:23:40Z nyef_: Thank you for your assistance. 2014-04-21T18:29:00Z stassats`: looks like arm64 has a division instruction 2014-04-21T18:29:56Z stassats`: and high 8 bits of addresses may be used for whatever 2014-04-21T18:31:48Z pkhuong: for whatever, or unused? 2014-04-21T18:32:34Z slyrus joined #sbcl 2014-04-21T18:32:55Z stassats`: ignored, can be used for tags 2014-04-21T18:33:46Z pkhuong: well... Hopefully I'll be around to see what happens 20 years from now. 2014-04-21T18:34:10Z stassats`: i think the easiest way to get an arm64 target is the current iphone 2014-04-21T18:34:47Z pkhuong: But it's not that easy to get a toolchain running on them. 2014-04-21T18:42:23Z eudoxia quit (Quit: leaving) 2014-04-21T18:53:13Z stassats`: unbind-to-here really shouldn't do anything 2014-04-21T19:09:11Z White_Flame quit (Ping timeout: 250 seconds) 2014-04-21T19:11:14Z krzysz00 joined #sbcl 2014-04-21T19:21:28Z stassats`: special binding inhibits TCO, maybe that's it 2014-04-21T19:23:14Z stassats`: yup, (defun bar () (declare (optimize debug)) (/show0 test2)) fails too (not at symbol binding, though) 2014-04-21T19:23:40Z nyef_: Tail-position, multiple values? 2014-04-21T19:23:56Z stassats`: tail position, unknown values 2014-04-21T19:24:07Z nyef_: Damnit. I thought I had these bits nailed. /-: 2014-04-21T19:25:51Z nyef_: Oh, tell me that I didn't invert the condition sense in RECEIVE-UNKNOWN-VALUES? 2014-04-21T19:26:40Z nyef_: I did, didn't I? 2014-04-21T19:27:13Z nyef_: See if switching :ne and :eq in receive-unknown-values helps at all. 2014-04-21T19:31:51Z stassats`: it does help, i'm greeted by the fault 1 from earlier today 2014-04-21T19:32:36Z stassats`: that was when i removed all the instances of without-interrupts wrapping, so, it seems to do the right thing 2014-04-21T19:33:09Z stassats`: and reading it confirms too 2014-04-21T19:35:11Z stassats`: another bug bites the dust 2014-04-21T19:37:01Z stassats`: i couldn't leave it hanging, or wouldn't have been able to fall asleep, and that fault@1 is fully for tomorrow 2014-04-21T19:38:03Z stassats`: i can paste positive-primep if you want to proceed without fixing up fixnum truncates 2014-04-21T19:39:12Z stassats`: http://paste.lisp.org/display/142229#5 2014-04-21T19:40:18Z oleo: (declare (sb-ext:muffle-conditions style-warning)) within a defun has no effect here..... only a declaim from toplevel has.... 2014-04-21T19:40:19Z oleo: bleh 2014-04-21T19:40:52Z oleo: my version is old tho, maybe it got fixed in the meantime.... 2014-04-21T19:42:08Z nyef_: Okay, so I need to swap those two conditions and to sort out TRUNCATE, and then I should be caught up to where you are? 2014-04-21T19:42:23Z stassats`: yep 2014-04-21T19:42:36Z nyef_: Sounds good. Thank you again. 2014-04-21T19:43:08Z stassats`: i just can't wait to get out of the bloody cold-init 2014-04-21T19:43:36Z nyef_: Wouldn't that be nice? The good news is that cold-init actually shakes down a goodly portion of the system. 2014-04-21T20:05:26Z stassats` quit (Ping timeout: 252 seconds) 2014-04-21T20:12:35Z krzysz00 quit (Ping timeout: 255 seconds) 2014-04-21T20:20:29Z prxq quit (Quit: Leaving) 2014-04-21T20:22:39Z krzysz00 joined #sbcl 2014-04-21T20:27:54Z angavrilov quit (Remote host closed the connection) 2014-04-21T20:33:15Z oleo quit (Ping timeout: 250 seconds) 2014-04-21T20:34:04Z oleo joined #sbcl 2014-04-21T20:40:15Z jdz quit (Quit: Leaving...) 2014-04-21T20:42:53Z drmeister quit (Ping timeout: 264 seconds) 2014-04-21T20:55:30Z drmeister joined #sbcl 2014-04-21T21:13:30Z reb quit (Ping timeout: 240 seconds) 2014-04-21T21:23:21Z drmeister quit (Remote host closed the connection) 2014-04-21T21:24:44Z nyef_: I'm fairly certain that there are two bugs in our ARM implementation of %BIGFLOOR. 2014-04-21T21:27:18Z pkhuong: BTW, interesting bug for linux/arm: https://bug985155.bugzilla.mozilla.org/attachment.cgi?id=8409733 2014-04-21T21:29:41Z nyef_: Iiinteresting. 2014-04-21T21:30:20Z nyef_: Also, a mix of completely bloody typical and "I wonder if that's what's causing my grief with SIGSEGV after SIGTRAP?" 2014-04-21T21:30:35Z nyef_: Thank you for the heads-up, I'll have to see what I can do in that direction tonight. 2014-04-21T21:43:27Z nyef_: The quotient from (%BIGFLOOR 0 7 3) is #xfffffffd. 2014-04-21T21:44:02Z nyef_: I am having trouble comprehending precisely HOW this is even close to wrong. 2014-04-21T21:46:10Z drmeister joined #sbcl 2014-04-21T21:47:24Z krzysz00 quit (Ping timeout: 252 seconds) 2014-04-21T21:56:55Z nyef_: Oops. And I have a better direction for the SIGSEGV after SIGTRAP damage. Ugh. 2014-04-21T22:06:17Z psilord quit (Quit: Leaving.) 2014-04-21T22:10:47Z drmeister quit (Read error: Connection reset by peer) 2014-04-21T22:11:17Z brown joined #sbcl 2014-04-21T22:11:40Z brown is now known as Guest75598 2014-04-21T22:15:20Z eudoxia joined #sbcl 2014-04-21T22:15:35Z nyef_: Hrm. Always using SA_ONSTACK is actually WORSE. Weird. 2014-04-21T22:58:41Z krzysz00 joined #sbcl 2014-04-21T23:00:08Z drmeister joined #sbcl 2014-04-21T23:01:12Z krzysz00 quit (Client Quit) 2014-04-21T23:01:25Z krzysz00 joined #sbcl 2014-04-21T23:11:38Z segv- quit (Ping timeout: 240 seconds) 2014-04-21T23:23:08Z drmeister quit (Remote host closed the connection) 2014-04-21T23:26:30Z drmeister joined #sbcl 2014-04-21T23:28:59Z drmeister quit (Read error: Connection reset by peer) 2014-04-21T23:29:34Z drmeister joined #sbcl 2014-04-21T23:30:16Z drmeiste_ joined #sbcl 2014-04-21T23:33:04Z drmeiste_ quit (Read error: Connection reset by peer) 2014-04-21T23:33:26Z drmeiste_ joined #sbcl 2014-04-21T23:33:38Z drmeister quit (Ping timeout: 240 seconds) 2014-04-21T23:39:23Z drmeiste_ quit (Ping timeout: 276 seconds) 2014-04-21T23:43:14Z nyef_: Heh. I'm just looking at the back of my copy of "Hacker's Delight" and noticing that the first review says "it deserves a spot on your shelf right next to Knuth", and that's EXACTLY where it sits on my shelf. (-: 2014-04-21T23:43:23Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-21T23:45:16Z psilord joined #sbcl 2014-04-21T23:46:03Z drmeister joined #sbcl 2014-04-21T23:51:50Z pkhuong: mine are all in two bins 2014-04-21T23:51:58Z drmeister quit (Ping timeout: 240 seconds) 2014-04-21T23:53:41Z drmeister joined #sbcl 2014-04-21T23:54:35Z eudoxia quit (Quit: leaving) 2014-04-21T23:55:10Z nyef_: Ah, that special realization that you have five non-descriptor-regs, need to keep NFP in reserve because it tends to get marked component-live, have already used the other four, and need to use OCFP for something else. /-: 2014-04-21T23:55:51Z nyef_: (Specifically, I need a conditional error trap.) 2014-04-21T23:57:16Z drmeiste_ joined #sbcl 2014-04-21T23:57:40Z pkhuong: I don't even see how bignum-floor can work :| 2014-04-21T23:58:29Z drmeister quit (Ping timeout: 264 seconds) 2014-04-21T23:58:37Z nyef_: It can't. The MVN screws it up. The condition code is plausibly also wrong, but for a simple test fixing (%BIGFLOOR 0 7 3) MVN at least gets it going. 2014-04-21T23:59:08Z pkhuong: how about dividing by more than 33? 2014-04-21T23:59:22Z nyef_: Why would there be a problem with that? 2014-04-21T23:59:23Z pkhuong: it seems to be missing a doubling/halving step somewhere. 2014-04-22T00:00:32Z nyef_: REM starts off with the high bits of the dividend, and the lower bits are shifted in via the ADC instruction at the end of the loop. 2014-04-22T00:01:16Z nyef_: QUO starts off with the low bits of the dividend, is shifted left by one bit each time through the loop, with the next bit of the quotient shifted in at the right. 2014-04-22T00:01:26Z drmeiste_ quit (Remote host closed the connection) 2014-04-22T00:01:49Z nyef_: The :GE condition on SUB is plausibly wrong, as it's a SIGNED greater-than-or-equal. 2014-04-22T00:01:57Z pkhuong: so if I divide 100000:0 by 2, what happens? 2014-04-22T00:02:16Z pkhuong: shouldn't the divisor get shifted to the until it's just below REM? 2014-04-22T00:02:26Z pkhuong: -to the 2014-04-22T00:03:16Z nyef_: Honestly? I don't know. This code is based on the SPARC version... Which uses :ltu as the conditional. Joy. 2014-04-22T00:05:57Z nyef_: It may be that the high bits never actually exceed the divisor. 2014-04-22T00:06:00Z nyef_: Or even equal it. 2014-04-22T00:06:20Z krzysz00 quit (Ping timeout: 252 seconds) 2014-04-22T00:06:27Z pkhuong: or that we always have v8/v9 builds 2014-04-22T00:06:59Z nyef_: %BIGFLOOR is used in two places in bignum.lisp. In the first case, it's easy to show that the high-bits are always less than the divisor. 2014-04-22T00:07:25Z nyef_: (How? It starts at zero, and is subsequently updated with the remainder, which is by definition less than the divisor.) 2014-04-22T00:08:27Z nyef_: I'm having trouble following the second use (in BIGNUM-TRUNCATE-GUESS), so I can't really say for sure there. 2014-04-22T00:12:21Z pkhuong: given how quick the thing is... it might almost sense to call out to an assembly routine that enters PA and has fun mangling all available registers. 2014-04-22T00:13:07Z nyef_: Eh, I figured out two angles on the problem, one of which I'm going with now, and one of which I'll back-patch in later because it cleans up some noise in the generic (non-arm-specific) code. 2014-04-22T00:13:47Z drmeister joined #sbcl 2014-04-22T00:19:36Z jaimef: is there any sort of benchmark that is used often to measure regressions? 2014-04-22T00:21:49Z zRecursive joined #sbcl 2014-04-22T00:26:18Z drmeister quit (Read error: Connection reset by peer) 2014-04-22T00:26:51Z drmeister joined #sbcl 2014-04-22T00:30:30Z Hydan` joined #sbcl 2014-04-22T00:32:02Z Hydan quit (Ping timeout: 245 seconds) 2014-04-22T00:34:01Z jsnell_: jaimef: cl-bench usually. there's some infrastructure around that, to build all sbcl versions, store benchmark results in db, and produce graphs: https://github.com/antifuchs/autobench 2014-04-22T00:36:44Z nyef_: Ooh. Does that also run the test suite, and allow for building --with-sb-options ? 2014-04-22T00:37:37Z nyef_: Looks like the website with the results is gone? 2014-04-22T00:39:29Z jsnell_: my guess is that the code hasn't been changed since the --with-foo options were added. but it could build with diffent features enabled 2014-04-22T00:40:40Z nyef_: Mmm. I'm going to have to bookmark this page. 2014-04-22T00:40:41Z jsnell_: and yes, I'm not aware of any currently running instance (not that I'm really aware of anythings sbcl-related these days). but it used to be incredibly useful 2014-04-22T00:50:53Z slyrus quit (Ping timeout: 252 seconds) 2014-04-22T00:51:48Z ivan`` quit (Changing host) 2014-04-22T00:51:48Z ivan`` joined #sbcl 2014-04-22T01:08:23Z meiji11 joined #sbcl 2014-04-22T01:52:39Z ltbarcly joined #sbcl 2014-04-22T02:04:49Z nyef_: Latest arm-port changes pushed. 2014-04-22T02:14:25Z kanru joined #sbcl 2014-04-22T02:18:19Z krzysz00 joined #sbcl 2014-04-22T02:23:15Z nyef_: ... the CDR slot of the CONS with pointer #x00000000 is 1, isn't it? 2014-04-22T02:23:52Z nyef_: Yeah, list-pointer-lowtag has to be 3, because of the NIL thing. 2014-04-22T02:26:04Z nyef_: SPEC gets blanked to zero at some point after taking its first element for APPLY #'MAKE-PACKAGE. 2014-04-22T02:28:22Z nyef_: And I think that it's high time that I stopped for the evening. 2014-04-22T02:35:15Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-22T02:40:06Z christoph_debian quit (Ping timeout: 276 seconds) 2014-04-22T02:52:31Z christoph_debian joined #sbcl 2014-04-22T03:09:05Z attila_lendvai joined #sbcl 2014-04-22T03:09:05Z attila_lendvai quit (Changing host) 2014-04-22T03:09:05Z attila_lendvai joined #sbcl 2014-04-22T03:11:40Z ehaliewicz joined #sbcl 2014-04-22T03:13:03Z ltbarcly joined #sbcl 2014-04-22T03:18:32Z nyef_ quit (Quit: G'night all.) 2014-04-22T03:26:36Z ltbarcly_ joined #sbcl 2014-04-22T03:30:47Z ltbarcly quit (Ping timeout: 276 seconds) 2014-04-22T03:50:32Z yacks joined #sbcl 2014-04-22T03:55:50Z tmh- joined #sbcl 2014-04-22T04:06:05Z tmh- quit (Remote host closed the connection) 2014-04-22T04:17:24Z tmh- joined #sbcl 2014-04-22T04:20:05Z scymtym_ quit (Ping timeout: 264 seconds) 2014-04-22T05:03:44Z pranavrc joined #sbcl 2014-04-22T05:03:45Z pranavrc quit (Changing host) 2014-04-22T05:03:45Z pranavrc joined #sbcl 2014-04-22T05:06:56Z oleo quit (Quit: Leaving) 2014-04-22T05:07:14Z tmh- quit (Read error: Connection reset by peer) 2014-04-22T05:17:20Z drmeister quit (Remote host closed the connection) 2014-04-22T05:25:00Z ltbarcly_ quit (Quit: Computer has gone to sleep.) 2014-04-22T05:36:20Z ltbarcly joined #sbcl 2014-04-22T05:39:27Z ltbarcly quit (Client Quit) 2014-04-22T05:53:33Z pkhuong quit (Ping timeout: 250 seconds) 2014-04-22T05:54:08Z pkhuong joined #sbcl 2014-04-22T05:56:57Z tmh- joined #sbcl 2014-04-22T06:04:20Z krzysz00 quit (Ping timeout: 255 seconds) 2014-04-22T06:13:17Z angavrilov joined #sbcl 2014-04-22T06:17:49Z redline6561 quit (Ping timeout: 250 seconds) 2014-04-22T06:19:33Z irsol quit (Ping timeout: 250 seconds) 2014-04-22T06:19:54Z redline6561 joined #sbcl 2014-04-22T06:20:26Z irsol joined #sbcl 2014-04-22T06:23:27Z ASau quit (Ping timeout: 250 seconds) 2014-04-22T06:32:29Z meiji11 quit (Remote host closed the connection) 2014-04-22T06:55:33Z Hydan` quit (Ping timeout: 276 seconds) 2014-04-22T06:55:33Z slyrus joined #sbcl 2014-04-22T07:06:11Z drmeister joined #sbcl 2014-04-22T07:07:05Z DGASAU joined #sbcl 2014-04-22T07:11:05Z drmeister quit (Ping timeout: 264 seconds) 2014-04-22T07:21:33Z prxq joined #sbcl 2014-04-22T07:29:48Z momo-reina joined #sbcl 2014-04-22T07:55:56Z pkhuong quit (Ping timeout: 255 seconds) 2014-04-22T08:00:21Z attila_lendvai quit (Quit: Leaving.) 2014-04-22T08:00:42Z prxq quit (Quit: Leaving) 2014-04-22T08:07:46Z momo-reina quit (Remote host closed the connection) 2014-04-22T08:09:00Z zRecursive quit (Remote host closed the connection) 2014-04-22T08:14:23Z pkhuong joined #sbcl 2014-04-22T08:27:55Z stassats joined #sbcl 2014-04-22T08:54:59Z drmeister joined #sbcl 2014-04-22T08:59:26Z drmeister quit (Ping timeout: 252 seconds) 2014-04-22T09:29:51Z jdz joined #sbcl 2014-04-22T09:31:37Z kludge` quit (Ping timeout: 245 seconds) 2014-04-22T09:37:31Z kludge` joined #sbcl 2014-04-22T09:52:38Z stassats quit (Ping timeout: 252 seconds) 2014-04-22T10:04:33Z nyef joined #sbcl 2014-04-22T10:04:51Z nyef: G'morning all. 2014-04-22T10:06:24Z Krystof: hiya 2014-04-22T10:36:45Z DGASAU` joined #sbcl 2014-04-22T10:37:18Z DGASAU quit (Ping timeout: 240 seconds) 2014-04-22T10:38:52Z DGASAU` is now known as DGASAU 2014-04-22T10:43:44Z drmeister joined #sbcl 2014-04-22T10:48:02Z drmeister quit (Ping timeout: 240 seconds) 2014-04-22T10:53:22Z nyef: Ugh. In http://paste.lisp.org/display/142247 the disassembly for CALL-VARIABLE... Shouldn't CFP be getting saved anywhere before it gets overwritten? /-: 2014-04-22T10:56:39Z nyef: CALL-VARIABLE doesn't use ALLOCATE-FULL-CALL-FRAME, does it? 2014-04-22T11:01:12Z krzysz00 joined #sbcl 2014-04-22T11:04:10Z tmh- quit (Remote host closed the connection) 2014-04-22T11:13:58Z jdz quit (Remote host closed the connection) 2014-04-22T11:15:12Z nyef: If I'm seeing a lot of random stuff in the SHOW output, including a goodly run of "/globally declaring ~A special", that means that it's processing toplevel forms, doesn't it? 2014-04-22T11:16:34Z nyef: Also seeing "/entering early-defstructs.lisp". 2014-04-22T11:16:55Z Krystof: yes! 2014-04-22T11:17:06Z nyef: Sweet. 2014-04-22T11:19:23Z nyef: Still a good chunk more to do before first GC, but I'm liking this so far. 2014-04-22T11:33:57Z michael_lee joined #sbcl 2014-04-22T11:36:36Z jdz joined #sbcl 2014-04-22T11:38:38Z yacks quit (Ping timeout: 252 seconds) 2014-04-22T11:43:49Z michael_lee quit (Quit: Ex-Chat) 2014-04-22T11:56:06Z stassats joined #sbcl 2014-04-22T11:57:28Z Krystof: how far does it get? 2014-04-22T11:58:58Z ehaliewicz quit (Ping timeout: 240 seconds) 2014-04-22T12:00:32Z nyef: early-defstructs. 2014-04-22T12:01:40Z nyef: something about an odd-fixnum not being a simple-string. 2014-04-22T12:03:41Z stassats: and i'm up to speed, /entering early-defstructs.lisp too 2014-04-22T12:09:26Z stassats: (symbolicate 'name "-") is broken 2014-04-22T12:09:41Z nyef: That was fast! 2014-04-22T12:10:18Z stassats: now, is it passing a fixnum to find-symbol* or find-symbol* mucks it up by itself 2014-04-22T12:19:21Z drmeister joined #sbcl 2014-04-22T12:24:58Z eudoxia joined #sbcl 2014-04-22T12:31:14Z stassats: can the type check for simple-string turn a string into a fixnum? 2014-04-22T12:34:13Z stassats: it is still a string before INTERN calls INTERN*, INTERN* already gets a fixnum 2014-04-22T12:34:15Z drmeister quit (Remote host closed the connection) 2014-04-22T12:37:51Z nyef: What VOP is involved? 2014-04-22T12:38:20Z stassats: i removed the declaration, it still fails 2014-04-22T12:38:29Z nyef: INTERN is a small enough function to check the trace file on. 2014-04-22T12:39:49Z stassats: what about &key processing? 2014-04-22T12:40:18Z nyef: It's an &OPTIONAL, not an &KEY. 2014-04-22T12:40:43Z nyef: Oh! You mean on the INTERN* side? 2014-04-22T12:40:44Z stassats: in intern* 2014-04-22T12:41:13Z nyef: slap a breakpoint on intern* using gdb and the trace-file, see what the actual parameters are at that point? 2014-04-22T12:42:12Z stassats: i don't want to figure out gdb incantations 2014-04-22T12:42:27Z stassats: i removed the &key parameter, it doesn't get fixnumied anymore 2014-04-22T12:42:44Z nyef: Ouch. 2014-04-22T12:42:57Z nyef: Okay, so we're looking at... probably &MORE-arg handling. 2014-04-22T12:43:13Z stassats: let's try &rest 2014-04-22T12:43:29Z krzysz00 quit (Ping timeout: 264 seconds) 2014-04-22T12:43:47Z nyef: Might only happen if we're using more-arg-context and not if we're using listify-rest-args. 2014-04-22T12:43:51Z stassats: &rest is enough 2014-04-22T12:44:23Z nyef: Okay, that's fun. 2014-04-22T12:45:13Z nyef: It's probably copy-more-arg, then. 2014-04-22T12:46:11Z drmeister joined #sbcl 2014-04-22T12:46:31Z nyef: All of the register args are live, and there should be two more-args on the stack. 2014-04-22T12:47:28Z stassats: there was a copy-more-arg problem on x86, where it overwrote the wrong registers 2014-04-22T12:47:40Z stassats: the wrong stack space 2014-04-22T12:47:47Z stassats: it is still present on everything except x86oids 2014-04-22T12:48:17Z nyef: Because it copied "upwards" rather than "downwards"? 2014-04-22T12:48:17Z stassats: but it required more parameters, between 8 and 20, iirc 2014-04-22T12:48:37Z drmeister quit (Read error: Connection reset by peer) 2014-04-22T12:48:49Z stassats: the problem was that sometimes it need to copy one direction, sometimes another 2014-04-22T12:49:09Z nyef: Really? Ouch. 2014-04-22T12:49:18Z stassats: see 20102d9bd3c62691cc2c27725ff7ffdcab54ab2b 2014-04-22T12:50:09Z stassats: x86-64 started to fail with 8 arguments, ppc currently fails with 17 2014-04-22T12:51:39Z pranavrc quit 2014-04-22T12:52:00Z nyef: Still not getting it, but not exactly trying too hard yet. 2014-04-22T12:52:35Z nyef: Our control stack frames start at 8 slots, with a hard minimum of 3. 2014-04-22T12:52:53Z stassats: (string a &rest args) works, trying (string a b &rest args) now 2014-04-22T12:53:41Z stassats: can't seem to reproduce what intern* demonstrates 2014-04-22T12:59:36Z nyef: Lovely. 2014-04-22T13:00:56Z nyef: Okay, I'm prepping to get a trace-file and to step through the affected code. 2014-04-22T13:01:13Z nyef: This may take a while, however, because I have actual paying work to do today. 2014-04-22T13:03:29Z stassats: i don't quite get the early-defstructs-args.lisp-expr thing 2014-04-22T13:03:51Z stassats: why not just put the thing in early-defstructs.lisp 2014-04-22T13:06:09Z pkhuong: yeah, same bug on ARM. 2014-04-22T13:06:22Z pkhuong: (re copy-more-arg) 2014-04-22T13:06:53Z stassats: but i don't expect it to materialize on just three arguments 2014-04-22T13:08:20Z pkhuong: same. checking it just the same in case we have unexpected stack grows up issues. 2014-04-22T13:10:09Z stassats: my hand constructed test cases doesn't demonstrate the problem directly, but it's enough to cause CORRUPTING WARNING bonanza 2014-04-22T13:10:15Z stassats: later, probably 2014-04-22T13:12:22Z pkhuong: mm.. NULL tn should really get more explanations. I assume that it's because arm sucks at constants and because NIL happens to be the first object in static space. 2014-04-22T13:12:48Z nyef: It's also traditional for partitioned-register-set ports. 2014-04-22T13:17:22Z pkhuong: you're having way too much fun with predicated execution ;) 2014-04-22T13:17:26Z nyef: http://paste.lisp.org/display/142248 2014-04-22T13:17:32Z pkhuong: the 64 bit port will be "fun" 2014-04-22T13:17:59Z nyef: I still have gdb open at this point, if anyone wants me to take a poke at anything, but calling runtime functions WILL corrupt the stack. 2014-04-22T13:19:21Z pkhuong: sub r5, lr, r7, lsr #6 <- where does that come from? 2014-04-22T13:19:29Z nyef: Annotated with a small chunk of trace-file. 2014-04-22T13:19:43Z nyef: That's from COMPUTE-CODE. 2014-04-22T13:20:38Z stassats: it doesn't look there like anything touching R0 2014-04-22T13:20:40Z nyef: The two high bits of SIMPLE-FUN-WIDETAG or whatever it's called are zero. The header data field contains the number of words from the start of the code-object to the start of the header. 2014-04-22T13:20:55Z pkhuong: . hack. h/t 2014-04-22T13:21:39Z stassats: nyef: more dissasembly? 2014-04-22T13:22:30Z nyef: stassats: Maybe it's getting corrupted at the call to FIND-SYMBOL* ? 2014-04-22T13:23:10Z stassats: no, i added (/hexstr name) just after declare 2014-04-22T13:23:42Z stassats: it may get corrupted during a call to HEXSTR 2014-04-22T13:24:31Z nyef: Oddly enough, that disassembly runs to the end of SIMPLE-STRING-P. 2014-04-22T13:25:55Z nyef: It's then stashed to CS4, then there's the key-arg processing... 2014-04-22T13:26:06Z pkhuong: time for paying work, but I will say that a predicated pre-incrementing store with offset is no RISC. 2014-04-22T13:26:16Z nyef: Heh. 2014-04-22T13:26:27Z nyef: Yeah, I need to get some paying work done, myself. 2014-04-22T13:26:42Z pkhuong: I'm sure you could have added a lsr in there ;) 2014-04-22T13:27:21Z nyef: For a store? No, the memory-operand is encoded in the same part of the instruction word as a shifter-operand. 2014-04-22T13:27:38Z stassats: even when i copy the whole intern and intern*, i don't get the same problem 2014-04-22T13:32:55Z pkhuong: btw, stassats, the only thing preventing me from merging jump tables is our lack of jump-jump elimination... I don't know if you see how do that easily in IR2. 2014-04-22T13:34:21Z stassats: i don't 2014-04-22T13:34:28Z nyef: Hrm. It's definitely corrupted by the time we call FIND-SYMBOL*. 2014-04-22T13:35:47Z nyef: More the point, the first invocation of FIND-SYMBOL*. 2014-04-22T13:42:20Z nyef: That's not a random fixnum, that's a stack frame address. 2014-04-22T13:42:34Z nyef: ... oh. 2014-04-22T13:43:36Z nyef: Ugh, don't have time for this right now. 2014-04-22T14:03:17Z oleo joined #sbcl 2014-04-22T14:05:41Z segv- joined #sbcl 2014-04-22T14:20:11Z stassats: huh, length is corrupted as well 2014-04-22T14:20:46Z stassats: in INTERN, length is 11, inside intern* it's 171 2014-04-22T14:20:57Z nyef: Lovely. 2014-04-22T14:21:39Z stassats: 17, not 11 2014-04-22T14:21:55Z stassats: your gdb dump shows 0x44, which is 17 2014-04-22T14:22:32Z oleo: on which files do you use gdb ? 2014-04-22T14:23:00Z drmeister joined #sbcl 2014-04-22T14:23:08Z stassats: all the three arguments are saved onto the stack, then they are moved up, then they are fetched to call find-symbol* 2014-04-22T14:23:14Z nyef: I open gdb on src/runtime/sbcl, then set args to specify a core file. 2014-04-22T14:23:17Z stassats: or is it down, but doesn't matter 2014-04-22T14:23:40Z stassats: maybe something else is overwriting the stack 2014-04-22T14:24:26Z Guest75598: Any SBCL threading gurus around? 2014-04-22T14:24:29Z Guest75598 is now known as reb 2014-04-22T14:24:41Z stassats: and adding /hexstr breaks LDB 2014-04-22T14:24:44Z nyef: reb: I don't know about "guru", but what's up? 2014-04-22T14:25:57Z reb: I'm running a test program that creates a bunch of threads. It hangs when trying to exit. From GDB I can tell there are two threads when it's hung. 2014-04-22T14:26:58Z yacks joined #sbcl 2014-04-22T14:27:03Z reb: One stack trace has %exit -> %exit-other-threads -> join-thread -> ... %%wait-for-mutex 2014-04-22T14:27:36Z reb: The other has make-thread -> call-with-system-mutex -> ... %%wait-for-mutex. 2014-04-22T14:28:06Z reb: It looks like both threads may be waiting on the system mutex. 2014-04-22T14:28:16Z nyef: Sounds like a race condition of some sort, but not in any part of the thead system I'm overly familiar with. 2014-04-22T14:28:49Z stassats: i now have the proper #x44, maybe /hexstr caused that corruption 2014-04-22T14:32:10Z nyef: Okay, I see the stack slots getting scribbled on. 2014-04-22T14:33:59Z Krystof: reb: which one of those is the main thread? 2014-04-22T14:35:01Z nyef: Oh, hell. We conditionally DON'T set the stack pointer correctly in COPY-MORE-ARG. 2014-04-22T14:35:32Z nyef: We may not set the stack pointer correctly at all, for that matter. 2014-04-22T14:37:58Z drmeister quit (Remote host closed the connection) 2014-04-22T14:38:27Z ASau joined #sbcl 2014-04-22T14:44:48Z Guest18924 joined #sbcl 2014-04-22T14:48:42Z Guest18924 is now known as _hmt 2014-04-22T15:10:25Z nyef: Three lines of "/about to do a TEST-CASE in pred.lisp", then a complex-single-float turns out not to be a REAL. 2014-04-22T15:11:29Z stassats: did you resolve copy-more-arg? 2014-04-22T15:11:40Z nyef: Yeah, just putting the commit together now. 2014-04-22T15:11:47Z stassats: good 2014-04-22T15:14:27Z reb: Krystof: I think the exiting thread is the main one. 2014-04-22T15:15:38Z reb: Krystof: I think I know what's happening. I'm starting and shutting down Swank a bunch of times. Swank has several background threads. It looks like one is trying to start up while my main thread is trying to exit. 2014-04-22T15:16:17Z nyef: Pushed. 2014-04-22T15:17:14Z reb: Krystof: There's some timeout code in the exit logic, so the program does eventually quit ... it just takes a minute to get unstuck. 2014-04-22T15:17:47Z nyef: I... might be seeing this lockup every so often in production at work. 2014-04-22T15:17:48Z stassats: what about recklessly? 2014-04-22T15:18:10Z stassats: or is it called abort now 2014-04-22T15:18:15Z stassats: (sb-ext:exit :abort t) 2014-04-22T15:19:09Z stassats: nyef: i now get CORRUPTION WARNING ad infinitum 2014-04-22T15:20:34Z nyef: stassats: --lose-on-corruption. 2014-04-22T15:20:45Z stassats: even with --lose-on-corruption 2014-04-22T15:21:14Z nyef: That's not right... 2014-04-22T15:21:39Z stassats: after about to do a test-case 2014-04-22T15:22:20Z oleo: don't they grab a mutex ? 2014-04-22T15:23:02Z nyef: See if http://paste.lisp.org/display/142250 helps at all? 2014-04-22T15:25:42Z nyef: (Yes, I AM planning on tidying this up and committing it upstream. It provides just that extra bit of information when something goes wRong sometimes... Like the ability to backtrace from a SIGSEGV, examine context registers, and so on.) 2014-04-22T15:27:10Z oleo: internals page says some signals are neve safe..... 2014-04-22T15:30:28Z stassats: nyef: no help 2014-04-22T15:31:15Z nyef: Hrm. 2014-04-22T15:32:08Z nyef: Only other deltas in my working tree are the :trace-file flag for target-package and a #!-arm for ERROR. 2014-04-22T15:32:44Z nyef: (Because I want the undefined-function trap at that point so that I can figure out what's going on, because I'd kept ending up in ERROR too early in cold-init.) 2014-04-22T15:32:51Z stassats: i had :trace-file for target-package too 2014-04-22T15:36:49Z stassats: and it's an internal error, not a call to ERROR, so i don't see how disabling it will help 2014-04-22T15:38:39Z stassats: but it does 2014-04-22T15:41:39Z stassats: so, realpart of a complex is wrong 2014-04-22T15:42:12Z nyef: ... Yeah, that's not surprising. What's surprising is that it compiled something that tried it. 2014-04-22T15:43:08Z nyef: At some point I'm going to have to sort out the horrid mess which is signal handling. /-: 2014-04-22T15:46:57Z stassats: wait, i missed #-arm all along 2014-04-22T15:48:27Z stassats: ok, let's disable the test too 2014-04-22T15:49:10Z stassats: maybe new temporary feature, :arm-complex-vops 2014-04-22T15:54:14Z nyef: Maybe I'll just implement the complex VOPs this evening. 2014-04-22T15:54:40Z nyef: After all, that worked for TRUNCATE, and it shouldn't be nearly as difficult to figure out. 2014-04-22T15:55:40Z stassats: and CORRUPTION WARNINGs is back 2014-04-22T15:56:50Z nyef: In interrupt.c, look for SA_ONSTACK. Change it so that it always applies, instead of applying only to SIGSEGV. The symptoms should change a bit, but it's still bleeding useless. /-: 2014-04-22T15:58:49Z stassats: i now get up to lexenv.lisp 2014-04-22T15:59:10Z nyef: From disabling the complex stuff? 2014-04-22T15:59:23Z stassats: yes 2014-04-22T16:00:06Z nyef: About halfway through toplevel files, then? 2014-04-22T16:00:15Z nyef: Up from a quarter. 2014-04-22T16:00:40Z stassats: fails at parse-lambda-list-like-thing 2014-04-22T16:02:12Z stassats: actually, it's fndb, not lexenv 2014-04-22T16:04:49Z nyef: More damage with &MORE args? 2014-04-22T16:05:43Z stassats: and it's the first defknown 2014-04-22T16:06:42Z stassats: the call to specifier-type 2014-04-22T16:17:59Z stassats: if i call parse-lambda-list-like-thing with the same argument, but from the top-level of fndb, it doesn't fail 2014-04-22T16:18:44Z stassats: it returns 13 values, may be something due to that 2014-04-22T16:23:11Z stassats: ok, when i call it myself with a multiple-value-bind, it does crash 2014-04-22T16:23:21Z nyef: Thirteen values? 2014-04-22T16:23:40Z nyef: Umm... RETURN, RETURN-MULTIPLE, or something else? 2014-04-22T16:24:51Z nyef: Hrm. Could be on the receive end as well, and that'd be the nasty case for default-unknown-values. 2014-04-22T16:25:17Z stassats: backtrace shows that it is still in parse-lambda-list-like-thing 2014-04-22T16:25:18Z nyef: Well, the mostly-nasty case, as the actual nasty case would involve a mismatch in the number of values. 2014-04-22T16:27:30Z nyef: There's nothing superficially wrong with the RETURN VOP... 2014-04-22T16:30:08Z stassats: can reproduce with just stuffing (values 13-times) 2014-04-22T16:32:26Z stassats: binding only 3 values doesn't crash 2014-04-22T16:32:49Z stassats: binding 4 does 2014-04-22T16:32:59Z stassats: and that's goes through the stack 2014-04-22T16:33:04Z nyef: But it fails on the caller side, not the callee side? 2014-04-22T16:33:14Z stassats: according to the backtrace 2014-04-22T16:33:27Z nyef: Err... that's backwards, isn't it? Callee side, not the caller side? 2014-04-22T16:33:38Z stassats: according to the backtrace, it fails on the callee side 2014-04-22T16:33:55Z nyef: But we're in return processing, so the backtrace might lie. 2014-04-22T16:34:38Z stassats: ok, returning just 4 values and binding all 4 of them fails too 2014-04-22T16:34:41Z nyef: The CODE register is still pointing to the callee. 2014-04-22T16:34:57Z nyef: What about returning 4 but only binding 3? 2014-04-22T16:35:02Z nyef: And vice-versa? 2014-04-22T16:35:08Z stassats: returning 13 and binding 3 doesn't crash 2014-04-22T16:35:21Z stassats: con says r0-r3 are the values i returned 2014-04-22T16:35:24Z nyef: So it's caller-side, default-unknown-values. 2014-04-22T16:35:49Z nyef: And the CODE register is unlikely to have been recomputed, which might explain the backtrace. 2014-04-22T16:35:58Z nyef: Unless it doesn't. 2014-04-22T16:37:14Z stassats: h, i return 3, which is #xc as a fixnum, and the memory fault is at #xC, coincidence? 2014-04-22T16:37:20Z stassats: the third value is 3 2014-04-22T16:38:09Z stassats: it was 2014-04-22T16:38:21Z stassats: when changed to 33, the fault is at C 2014-04-22T16:39:08Z stassats: returnin 5 values, the fault is at 10 2014-04-22T16:39:24Z nyef: What is OCFP? 2014-04-22T16:39:36Z nyef: When the fault occurs, that is. 2014-04-22T16:39:40Z stassats: 6 => 14 2014-04-22T16:40:10Z stassats: https://oeis.org/search?q=12%2C16%2C20 2014-04-22T16:40:42Z stassats: "Multiples of 4. " sounds about right 2014-04-22T16:41:03Z nyef: So, NARGS? 2014-04-22T16:41:43Z stassats: http://paste.lisp.org/display/142251 2014-04-22T16:41:55Z stassats: nargs is 18 2014-04-22T16:42:10Z nyef: Hrm. 2014-04-22T16:42:21Z nyef: Having a context on a memory fault is nice, isn't it? 2014-04-22T16:42:25Z drmeister joined #sbcl 2014-04-22T16:43:33Z nyef: What's the word in memory at 0x4fdc3dcc? 2014-04-22T16:44:11Z stassats: i have a differrent pc now, but anyway 2014-04-22T16:44:14Z nyef: Actually, what are the words from 4fdc3dc8 through ...d4? 2014-04-22T16:44:40Z nyef: The word before PC when it faults to two words after. 2014-04-22T16:45:50Z stassats: http://paste.lisp.org/display/142251#1 2014-04-22T16:46:43Z stassats: why does context print PC twice? 2014-04-22T16:47:57Z nyef: Because the first time it's printing the GPRs. 2014-04-22T16:48:12Z stassats: i see 2014-04-22T16:49:10Z stassats: my fourth arg seems to be hiding at CFP+44 2014-04-22T16:49:48Z nyef: VALUES-ON-STACK (TEMP) is packed to OCFP. MOVE-TEMP is packed to LEXENV... 2014-04-22T16:50:12Z stassats: what does O mean? 2014-04-22T16:50:19Z stassats: minion: what does OCFP stand for? 2014-04-22T16:50:19Z minion: Omphalectomy Cystostomy Fitly Poimenics 2014-04-22T16:50:26Z nyef: Old Current Frame Pointer. 2014-04-22T16:50:56Z stassats: and NFP is New Fame Pointer? 2014-04-22T16:51:06Z nyef: Number-stack Frame Pointer. 2014-04-22T16:51:14Z stassats: ok 2014-04-22T16:51:35Z kanru quit (Ping timeout: 276 seconds) 2014-04-22T16:52:07Z nyef: I see the bug, but fixing it might get a little tricky. 2014-04-22T16:53:49Z nyef: I need to add an ignored temporary on ocfp-offset to define-full-call. 2014-04-22T16:54:29Z nyef: I don't know if I can, due to register pressure, or if I need it for unknown returns. 2014-04-22T16:55:13Z nyef: Okay, looks like I don't need it for unknown. 2014-04-22T16:56:14Z stassats thinks how to move on to the next thing 2014-04-22T16:58:22Z nyef: Add ,@(when (eq return :fixed) '(ocfp-temp)) to the :ignore clause in define-full-call, and ,@(:temporary (:sc non-descriptor-reg :offset ocfp-offset :from :eval) ocfp-temp) in parallel to the :ignore clause and see if it helps? 2014-04-22T16:58:30Z nyef: Err... 2014-04-22T16:58:41Z nyef: Damnit, that last ,@ should've included a WHEN clause. 2014-04-22T16:58:51Z stassats: i switched to returning a list for parse-lambda-list-like-thing 2014-04-22T16:59:04Z nyef: Ah, okay. Leave me the tricky bits. (-: 2014-04-22T17:02:42Z stassats: parse-args-types returns many values too 2014-04-22T17:04:16Z stassats: what should happen when i run out of heap? 2014-04-22T17:05:08Z stassats: the next error is unknown type specifier, double-float-significand 2014-04-22T17:06:34Z nyef: Hitting the GC trigger page should result in a SIGSEGV, which should be handled transparently by setting some flag somewhere and resuming, unless GC is enabled in which case there's a P-A dance and we die screaming because our SIGTRAP stuff is still broken. 2014-04-22T17:07:55Z stassats: that's due to integer-decode-double-float 2014-04-22T17:08:18Z nyef: In src/code/deftypes-for-target, kill the #!-arm conditional. 2014-04-22T17:08:56Z stassats: just did 2014-04-22T17:11:48Z stassats: bignum-gcd-order-and-substract is next 2014-04-22T17:12:07Z stassats: returns four values 2014-04-22T17:12:13Z stassats: gah, i'll wait for your fix then 2014-04-22T17:12:19Z stassats: make it five 2014-04-22T17:13:11Z jdz quit (Remote host closed the connection) 2014-04-22T17:18:09Z stassats: Memory fault at 66fff000, what does that mean? 2014-04-22T17:18:15Z stassats: top of stack is LIST 2014-04-22T17:19:01Z nyef: On the one hand, that's suspiciously aligned. On the other, we don't have VOPs defined for LIST or LIST*. 2014-04-22T17:19:09Z stassats: i mean, is that maybe GC? 2014-04-22T17:19:20Z nyef: Yeah, just a sec. 2014-04-22T17:19:43Z stassats: list uses copying &more into a list vop 2014-04-22T17:20:08Z stassats: aka listify-rest-args 2014-04-22T17:20:40Z nyef: Look up current_auto_gc_trigger in sbcl.nm and see if you can dump that word of memory. 2014-04-22T17:22:43Z stassats: 0x308fc : 0x00000000 2014-04-22T17:22:57Z nyef: Hrm. 2014-04-22T17:23:15Z stassats: p current_auto_gc_trigger works too 2014-04-22T17:23:23Z nyef: Neat. 2014-04-22T17:23:28Z nyef: Oh, you're in gdb? 2014-04-22T17:23:38Z stassats: right 2014-04-22T17:23:45Z stassats: ldb says invalid Lisp-level address: 0x308fc 2014-04-22T17:23:55Z nyef: Oh dear. That's actually dynamic-space-0-end. 2014-04-22T17:24:20Z nyef: Are you pre- or post- first GC? 2014-04-22T17:26:41Z stassats: no idea 2014-04-22T17:27:04Z nyef: Is it still processing toplevels? 2014-04-22T17:27:10Z stassats: i figure my replacement of VALUES with LIST may have to do with that 2014-04-22T17:27:16Z stassats: it's still in fndb.lisp 2014-04-22T17:27:20Z nyef: Ouch. 2014-04-22T17:27:49Z nyef: Okay, I think you might be done for now, and I have quite a bit of stuff to clear up to try and get to where you are and see what's what. 2014-04-22T17:28:12Z stassats: i can increase dynamic-space-size! 2014-04-22T17:28:28Z nyef: ... True. 2014-04-22T17:29:07Z nyef: Things will really get bad at GC time if you do that, but that's going to be a pain anyway. 2014-04-22T17:29:39Z stassats: what's the default? 2014-04-22T17:30:41Z nyef: It's in compiler/arm/parms.lisp. 2014-04-22T17:32:25Z stassats: 383MB, and i can't specify more at --dynamic-space-size 2014-04-22T17:32:40Z stassats: mmap: wanted 524288000 bytes at 0x67000000, actually mapped at 0x6e400000 2014-04-22T17:32:58Z slyrus quit (Ping timeout: 240 seconds) 2014-04-22T17:33:49Z reb: stassats: Thanks for the suggestion to use (sb-ext:exit :abort t). My code's running inside a test framework, so perhaps I should try to make it clean up everything. Hmmm. 2014-04-22T17:39:23Z davazp joined #sbcl 2014-04-22T17:42:53Z stassats: why are there two dynamic spaces? 2014-04-22T17:43:53Z nyef: Because it's cheneygc. 2014-04-22T17:53:26Z stassats: consing bignums doesn't help either (no modarith) 2014-04-22T17:54:17Z stassats: sbcl now takes up 633 MB, i don't think i can go much further 2014-04-22T17:55:28Z Krystof: the good news is that after first GC works, second GC is usually much easier 2014-04-22T17:55:36Z Krystof: and after second GC works, third GC works almost always! 2014-04-22T17:56:21Z jsnell_: the first gc is no problem. it's just that the rest of the system doesn't work once the GC has mangled up everything ;-) 2014-04-22T17:56:41Z attila_lendvai joined #sbcl 2014-04-22T17:56:42Z attila_lendvai quit (Changing host) 2014-04-22T17:56:42Z attila_lendvai joined #sbcl 2014-04-22T18:03:29Z Krystof: ah, happy days 2014-04-22T18:04:09Z stassats: is cheneygc really that platform dependent? 2014-04-22T18:04:25Z Krystof: life was simpler when all I had to worry about was the random scribbling all over memory 2014-04-22T18:04:52Z Krystof: stassats: handling signals is hilariously platform dependent 2014-04-22T18:05:43Z nyef: You want hilarious? We're going to have to patch a trampoline with four NOPs that jumps to the real signal handler in place to cover for some kernel bug. 2014-04-22T18:06:09Z nyef: Plus our control stack moves opposite to C's, and is purely boxed data, but has to share the same stack pointer. 2014-04-22T18:06:25Z Krystof: the first part is not as hilarious as the sparc linux signal handling problem 2014-04-22T18:06:33Z Krystof: the second part, I concede, is totally hilarious 2014-04-22T18:07:51Z Krystof: "the &void_context+37 hack was removed" 2014-04-22T18:07:53Z Krystof is sad 2014-04-22T18:09:09Z davazp quit (Read error: Connection reset by peer) 2014-04-22T18:14:40Z stassats: i may just skip fndb 2014-04-22T18:18:39Z stassats: and first thing i get is "CAR" is not a known function 2014-04-22T18:22:56Z jdz joined #sbcl 2014-04-22T18:25:07Z stassats: disabled type parsing by %defknown 2014-04-22T18:25:14Z stassats: back to more 3+ values failures 2014-04-22T18:25:44Z stassats: ok, i'm now os-cold-init-or-reinit 2014-04-22T18:30:55Z attila_lendvai quit (Quit: Leaving.) 2014-04-22T18:39:57Z stassats: i'll really wait for mvb fixes 2014-04-22T18:45:12Z slyrus joined #sbcl 2014-04-22T18:55:44Z fiveop joined #sbcl 2014-04-22T18:57:48Z eudoxia quit (Quit: leaving) 2014-04-22T19:01:14Z nyef: Hello fiveop. 2014-04-22T19:02:04Z edgar-rft quit (Quit: computation corrupted because experience expired) 2014-04-22T19:07:58Z fiveop: Hi 2014-04-22T19:09:15Z ltbarcly joined #sbcl 2014-04-22T19:09:56Z fiveop: nyef: does (mov hi 0) have any advantage over (eor hi hi hi)? 2014-04-22T19:10:12Z stassats: i think with fixed values cold-init should complete without any GC 2014-04-22T19:10:19Z stassats: more so with modarith 2014-04-22T19:11:11Z nyef: fiveop: I have no idea, other than making the constant obvious. We could, I suppose, define a ZEROIZE macro or similar. 2014-04-22T19:20:55Z Krystof: stassats: modarith is usually nice and easy to implement 2014-04-22T19:20:59Z Krystof: single-instruction vops 2014-04-22T19:21:50Z stassats: yeah, i added for +, - and logandc*, seems to not break anything 2014-04-22T19:22:37Z fiveop: nyef: did you push what you talked about earlier today? 2014-04-22T19:23:29Z nyef: With the exception of the egregious runtime damage for getting a context (and thus backtrace) for sigsegv, everything I have is pushed. I have quite the list to work on, however. 2014-04-22T19:25:41Z fiveop: If I have files A and B in build-order (in that order) and define a constant in file A, will that be available at readtime of file A (assuming packages are correct etc.)? 2014-04-22T19:26:01Z fiveop: the last A should be a B 2014-04-22T19:27:11Z stassats: i surely expect it to be 2014-04-22T19:28:21Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-22T19:30:57Z leuler joined #sbcl 2014-04-22T19:35:39Z stassats quit (Ping timeout: 252 seconds) 2014-04-22T19:36:19Z ltbarcly joined #sbcl 2014-04-22T19:36:47Z fiveop quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2014-04-22T19:37:15Z fiveop joined #sbcl 2014-04-22T19:51:46Z DGASAU quit (Remote host closed the connection) 2014-04-22T19:54:04Z DGASAU joined #sbcl 2014-04-22T20:03:01Z prxq joined #sbcl 2014-04-22T20:23:23Z angavrilov quit (Remote host closed the connection) 2014-04-22T20:33:10Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-22T20:39:05Z drmeister quit (Quit: Leaving...) 2014-04-22T20:39:49Z fiveop: When I run nyef's arm head, I get a bunch of "/NIL" lines before the application is killed. Is this what is currently expected? Do we have an idea where that's happening? (use-package ?) 2014-04-22T20:40:19Z ltbarcly joined #sbcl 2014-04-22T20:40:33Z nyef: That sounds like it's dying a touch early. 2014-04-22T20:40:39Z DGASAU quit (Remote host closed the connection) 2014-04-22T20:42:55Z fiveop: as in: it should run longer? 2014-04-22T20:43:14Z fiveop: I don't see "about to SETQ *IN-PACKAGE-INIT*" 2014-04-22T20:43:58Z nyef: package-init is before all of the /NILs. 2014-04-22T20:44:19Z DGASAU joined #sbcl 2014-04-22T20:44:59Z fiveop: I see about to do *!DEFERRED-USE-PACKAGES!*, then the /NILs then the kill 2014-04-22T20:46:26Z nyef: I get a bunch of /NILs, the setq for *i-p-i*, and then it moves on to globaldb. 2014-04-22T20:46:46Z nyef: Oh! #!-arm out the definition of ERROR, see if that does anything. 2014-04-22T20:47:53Z fiveop: as in (defun error in cold-error? 2014-04-22T20:48:07Z nyef: That'd be the one. 2014-04-22T20:48:37Z nyef: Among other things, it means we get a trap to LDB if it gets called "too early" in cold-init, which we're still trying to get through. 2014-04-22T20:48:50Z oleo quit (Ping timeout: 276 seconds) 2014-04-22T20:58:29Z nyef quit (Ping timeout: 264 seconds) 2014-04-22T21:01:14Z oleo joined #sbcl 2014-04-22T21:09:38Z fiveop: #!-arm out the definition, rebuild, same behaviour. 2014-04-22T21:10:04Z fiveop: The first time I run make-target-2 the last few NILs take ages, then it gets killed 2014-04-22T21:10:29Z fiveop: any further invocations (after a rebuild) go very quickly through the NILs to the kill 2014-04-22T21:11:56Z nyef joined #sbcl 2014-04-22T21:13:08Z fiveop: could you have a quick look at the logs 2014-04-22T21:15:47Z nyef: I'm at a loss, I'm afraid. 2014-04-22T21:16:53Z fiveop: can you tell me where the code is that outputs the /NILs. I haven't found that yet, and I can continue looking at it from there 2014-04-22T21:19:54Z prxq quit (Quit: Leaving) 2014-04-22T21:23:02Z nyef: Looks like somewhere under USE-PACKAGE ? 2014-04-22T21:24:05Z fiveop: I'll look further 2014-04-22T21:40:02Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-22T21:49:23Z ltbarcly joined #sbcl 2014-04-22T21:51:59Z Bicyclidine joined #sbcl 2014-04-22T21:54:59Z fiveop: with-single-package-locked-error 2014-04-22T21:58:09Z fiveop: all kinds of kernel threads go nuts, when it processes the 'slow NILs' 2014-04-22T21:58:47Z fiveop: kswapd0, ksoftirqd 2014-04-22T22:00:55Z fiveop: I get killed, because the kernel goes oom and chooses sbcl to get killed 2014-04-22T22:05:59Z fiveop: SBCL uses about 25% of the available memory at the end. But that's almost all the memory in use on the device in total. 2014-04-22T22:12:03Z jdz quit (Remote host closed the connection) 2014-04-22T22:13:38Z fiveop: the kernel logs made me think, it might have something to do with NFS, so I copied the runtime and core onto the ssd 2014-04-22T22:14:33Z fiveop: it runs a lot faster now, but it still gets killed :/ (though the stacktrace with nfs functions vanished from the kernel log ... ) 2014-04-22T22:21:06Z fiveop: more on thursday 2014-04-22T22:21:11Z fiveop quit (Quit: bye) 2014-04-22T22:39:05Z leuler quit (Quit: ERC Version 5.1.2 $Revision: 1.796.2.6 $ (IRC client for Emacs)) 2014-04-22T22:43:45Z slyrus quit (Ping timeout: 252 seconds) 2014-04-22T22:51:33Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-22T22:54:40Z eudoxia joined #sbcl 2014-04-22T22:59:18Z ltbarcly joined #sbcl 2014-04-22T22:59:28Z ASau: Just for your information. 2014-04-22T22:59:51Z ASau: ASDF-Install was removed, but I still see doc/sbcl-asdf-install.1 man page source. 2014-04-22T23:17:04Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-22T23:23:17Z ltbarcly joined #sbcl 2014-04-22T23:28:31Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-22T23:47:34Z nyef: I... really couldn't be bothered hacking on SBCL/ARM tonight. 2014-04-22T23:47:52Z zRecursive joined #sbcl 2014-04-23T00:02:20Z ferada quit (Remote host closed the connection) 2014-04-23T00:10:06Z ltbarcly joined #sbcl 2014-04-23T00:12:06Z ltbarcly quit (Client Quit) 2014-04-23T00:40:42Z nicdev` is now known as nicdev 2014-04-23T00:54:37Z _hmt quit (Remote host closed the connection) 2014-04-23T00:55:20Z eudoxia quit (Quit: leaving) 2014-04-23T00:56:25Z Bicyclidine quit (Ping timeout: 252 seconds) 2014-04-23T01:02:09Z ltbarcly joined #sbcl 2014-04-23T01:08:45Z nyef: Whee... Let's build two cross compilers AT ONCE, on the SAME system. And one of them GCC. /-: 2014-04-23T01:20:49Z pkhuong: crossbuilding gcc, or just building a crossgcc? (: 2014-04-23T01:23:28Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-23T01:24:50Z nyef: Building a crossgcc. 2014-04-23T01:25:02Z nyef: I need a mips64 cross-compile environment. 2014-04-23T01:26:43Z nyef: (When faced with the choice of "port SBCL/MIPS to IRIX", "port SBCL/MIPS to OpenBSD/MIPS64", or "port the Linux kernel to a new MIPS64 system", well... ) 2014-04-23T01:33:49Z nyef: ... I'm sure ths would have approved of this plan. 2014-04-23T01:45:51Z nyef: minion: Memo for stassats: This patch at least builds, and might fix the ARM M-V-B thing: http://paste.lisp.org/display/142254 (I see no difference here, because I still need to fix up complex floats). 2014-04-23T01:45:51Z minion: Remembered. I'll tell stassats when he/she/it next speaks. 2014-04-23T02:15:28Z kanru joined #sbcl 2014-04-23T02:19:20Z ltbarcly joined #sbcl 2014-04-23T02:33:45Z ASau: If you want to build cross-compiler for a (known) NetBSD target, it is trivial. :) 2014-04-23T02:38:56Z nyef: ... Unfortunately, it's not actually supported by NetBSD. 2014-04-23T02:39:11Z christoph_debian quit (Ping timeout: 276 seconds) 2014-04-23T02:39:50Z nyef: The machine family is tier 2, but there's no mention of the actual hardware model family, and there's a note about another model family not being supported because it requires a 64-bit toolchain. 2014-04-23T02:39:55Z zRecursive: Is netbsd still alive ? i fell only freebsd active now 2014-04-23T02:40:51Z zRecursive: openbsd even cannot pay for its electricity :) 2014-04-23T02:42:03Z zRecursive: usage 2014-04-23T02:42:18Z ASau: Basically everything is "tier 2" unless it is really mainstream platform. 2014-04-23T02:42:26Z nyef: I get the feeling that if I were to divide SBCL backends up into tiers, it'd come out as tier 1: x86, x86-64, PPC; tier 2: SPARC; tier 3: MIPS, ALPHA, HPPA. 2014-04-23T02:43:47Z nyef: Tier 1 platforms have threads, tier 2 platforms at least have gencgc, and tier 3 platforms get no love. 2014-04-23T02:52:04Z christoph_debian joined #sbcl 2014-04-23T02:56:14Z pkhuong: ppc is tier1? 2014-04-23T02:56:50Z pkhuong: In terms of os, it's something like linux, fbsd/darwin/solaris, windows, the rest. 2014-04-23T03:00:34Z nyef: Yeah, I'll agree with that OS ranking. 2014-04-23T03:01:07Z joshe: :| 2014-04-23T03:01:08Z nyef: PPC is our go-to platform for sanity-checking threading stuff and is our proxy for non-x86oid systems in general. 2014-04-23T03:01:58Z pkhuong: arm will be fun. 2014-04-23T03:01:59Z nyef: Okay, maybe OpenBSD before Windows. I don't know that anybody here actually runs Windows on any kind of a regular basis. 2014-04-23T03:02:54Z joshe: however I hear someone is actively working on threading support for windows 2014-04-23T03:02:56Z nyef: ARM will probably be a special corner of hell until late this year or early next year. 2014-04-23T03:03:09Z nyef: Itanium might be interesting. 2014-04-23T03:04:16Z pkhuong: joshe: define actively working on? 2014-04-23T03:04:52Z joshe: well, more actively than I am for openbsd 2014-04-23T03:05:38Z pkhuong: unlike open's, it works. 2014-04-23T03:05:47Z nyef: joshe: You want to do an OpenBSD/MIPS port of SBCL, right? Right? 2014-04-23T03:06:12Z joshe: although apparently I last did so on Wed Dec 11 18:25:00 2013 -0800 2014-04-23T03:07:06Z joshe: nyef: my port list would be ordered hppa, alpha, sparc, sparc64 2014-04-23T03:07:28Z nyef: Do you actually have hppa or alpha machines? 2014-04-23T03:07:41Z joshe: I have an hppa which no longer powers on 2014-04-23T03:07:52Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-23T03:07:58Z nyef: Ah, that's a bit of a pity. 2014-04-23T03:08:14Z joshe: and I could get access to alpha machines running openbsd fairly easily 2014-04-23T03:09:40Z nyef: The issue, of course, being that the backends have probably bitrotted, and it'd be easier to get them running again on Linux first before porting, rather than dealing with two sets of unknowns at once. 2014-04-23T03:11:02Z pkhuong: write a C backend, declare simultaneous non-victory ;) 2014-04-23T03:11:29Z nyef: Heh. 2014-04-23T03:12:35Z joshe: at one point I had ecl fairly close to bootstrapping sbcl, but it seems to have regressed significantly the last time I tried 2014-04-23T03:13:42Z pkhuong: plus, with safepoints, we might even get away with stuff like a shadow stack. 2014-04-23T03:16:08Z nyef: Okay, with a bit of luck I might finally be done with the test cases in pred.lisp. 2014-04-23T03:18:39Z tmh- joined #sbcl 2014-04-23T03:25:14Z tmh- quit (Remote host closed the connection) 2014-04-23T03:29:23Z nyef: Hrm. SIGILL, some kind of FPU instruction in the middle of a run of integer code. 2014-04-23T03:41:08Z Bicyclidine joined #sbcl 2014-04-23T03:59:04Z nyef: You know what's absolutely frightening? Seeing a component end with a CALL-NAMED VOP. Watching the instruction stream as it carefully emits an LRA header, restores reg_CODE, loads the NFP from the stack, and then fall off the end into the next component. 2014-04-23T04:00:39Z edgar-rft joined #sbcl 2014-04-23T04:10:39Z nyef quit (Quit: G'night all.) 2014-04-23T04:41:56Z attila_lendvai joined #sbcl 2014-04-23T04:41:57Z attila_lendvai quit (Changing host) 2014-04-23T04:41:57Z attila_lendvai joined #sbcl 2014-04-23T05:06:39Z oleo quit (Quit: Leaving) 2014-04-23T05:28:40Z pranavrc joined #sbcl 2014-04-23T05:28:41Z pranavrc quit (Changing host) 2014-04-23T05:28:41Z pranavrc joined #sbcl 2014-04-23T05:34:02Z Bicyclidine quit (Ping timeout: 240 seconds) 2014-04-23T05:39:22Z Bicyclidine joined #sbcl 2014-04-23T06:08:14Z angavrilov joined #sbcl 2014-04-23T06:08:39Z slyrus joined #sbcl 2014-04-23T06:12:53Z segv- quit (Ping timeout: 264 seconds) 2014-04-23T06:28:29Z slyrus quit (Ping timeout: 264 seconds) 2014-04-23T06:31:06Z Bicyclidine quit (Quit: stupid computer. brb) 2014-04-23T06:36:24Z Bike_ joined #sbcl 2014-04-23T06:36:59Z Bike_ is now known as Bicyclidine 2014-04-23T06:39:50Z slyrus joined #sbcl 2014-04-23T06:53:37Z slyrus quit (Ping timeout: 252 seconds) 2014-04-23T07:29:58Z Bicyclidine quit (Ping timeout: 240 seconds) 2014-04-23T07:31:44Z prxq joined #sbcl 2014-04-23T07:40:18Z ASau quit (Ping timeout: 240 seconds) 2014-04-23T07:42:34Z White_Flame joined #sbcl 2014-04-23T07:48:18Z zRecursive left #sbcl 2014-04-23T07:56:54Z prxq quit (Remote host closed the connection) 2014-04-23T07:57:44Z ferada joined #sbcl 2014-04-23T08:06:41Z prxq joined #sbcl 2014-04-23T08:06:42Z prxq_ joined #sbcl 2014-04-23T08:06:45Z prxq quit (Read error: Connection reset by peer) 2014-04-23T08:06:59Z prxq_ is now known as prxq 2014-04-23T08:11:38Z yacks quit (Ping timeout: 240 seconds) 2014-04-23T08:20:43Z jdz joined #sbcl 2014-04-23T08:41:49Z yacks joined #sbcl 2014-04-23T08:50:53Z erikvarga joined #sbcl 2014-04-23T09:17:59Z erikvarga left #sbcl 2014-04-23T09:31:20Z kludge` quit (Ping timeout: 255 seconds) 2014-04-23T09:32:22Z nyef joined #sbcl 2014-04-23T09:32:43Z nyef: G'morning all. 2014-04-23T09:36:21Z kludge` joined #sbcl 2014-04-23T09:36:40Z Krystof: morning 2014-04-23T09:38:27Z nyef: I'm at "/INDEX-IN-COLD-TOPLEVELS=.." 0x00008000, last filename seen is SYS:SRC;COMPILER;LEXENV.LISP, and it seems to land in an infinite loop somewhere under BIGNUM-GCD. 2014-04-23T09:39:16Z stassats joined #sbcl 2014-04-23T09:41:07Z nyef: G'morning stassats. Feel free to ignore the message that I left on minion. 2014-04-23T09:41:15Z stassats: minion: forget my memos 2014-04-23T09:41:16Z minion: OK, I threw it out. 2014-04-23T09:41:26Z stassats: i think the standard shouldn't have included upgraded-complex-part-type, since complexes are immutable 2014-04-23T09:41:33Z nyef: Hunh. Didn't know minion could do that. 2014-04-23T09:42:07Z stassats: the only reason would be to simplify the typechecking, maybe in the 80s it was expensive to test both realpart and imgpart 2014-04-23T09:42:57Z stassats: anyhow, only lispworks properly implements COMPLEX type specifiers 2014-04-23T09:43:58Z nyef: Surely we can't let them get away with being the only one? 2014-04-23T09:44:50Z stassats: i got past BIGNUM-GCD, with my s/values/list/ changes 2014-04-23T09:46:21Z nyef: Hrm. So there's possibly more damage in that area. Lovely. 2014-04-23T09:46:50Z stassats: what i'm sure of, the behaviour of complex types shouldn't be changed during the freeze 2014-04-23T09:47:21Z stassats: so, the transform should just get the previous behaviour back and any clhs induced bastardization to be done later 2014-04-23T09:49:10Z stassats: nyef: did you fix stack passed values issues? 2014-04-23T09:49:52Z nyef: I might have. If I had to in order to get this far, then it's fixed. Otherwise, it isn't. 2014-04-23T09:50:20Z nyef: I guess I could revert that part of my working tree and see if it breaks anything. 2014-04-23T09:51:03Z stassats: bignum-gcd-order-and-substract is the place where it failed for me due to values 2014-04-23T09:52:51Z nyef: Okay, so I possibly haven't fixed it, or there's more than the one touchpoint that I found that needs fixing. 2014-04-23T09:53:42Z stassats: i'll try your patch anyways 2014-04-23T09:54:15Z nyef: Okay, and I'll start putting together commits for the various float changes I made last night. 2014-04-23T09:56:45Z nyef: ... starting with the instruction encoding fix. 2014-04-23T10:01:13Z stassats: your patch had unbalanced parenthesis 2014-04-23T10:01:23Z nyef: ... How odd. 2014-04-23T10:03:05Z nyef: Oh, ouch. Not just unbalanced parens. 2014-04-23T10:03:26Z stassats: comma not inside backqoute 2014-04-23T10:03:37Z nyef: The missing bit is "ocfp-temp)))" 2014-04-23T10:03:40Z stassats: or was that due to parenthesis 2014-04-23T10:03:45Z nyef: On the last changed line of the patch. 2014-04-23T10:05:02Z stassats: ok, past make-host-1 2014-04-23T10:15:28Z stassats: ok, i get hung up in index-in-cold-toplevels 2014-04-23T10:16:02Z nyef: I sent a sigsegv to the process so that I could get a backtrace. 2014-04-23T10:17:40Z stassats: ah, i disabled parsing types to get through FNDB, perhaps why i didn't stumble there 2014-04-23T10:18:07Z nyef: So it's possibly something else entirely. 2014-04-23T10:18:24Z nyef: I'm almost done catching up commits for float fixes. 2014-04-23T10:21:13Z stassats: logior? 2014-04-23T10:21:26Z nyef: Trying a build without the call changes in order to see if it breaks earlier. 2014-04-23T10:21:45Z stassats: no, there's a define-binop 2014-04-23T10:23:31Z stassats: i just don't want to comb through a trace fil! 2014-04-23T10:23:32Z stassats: e 2014-04-23T10:24:15Z stassats: especially since it loses all the names 2014-04-23T10:27:33Z stassats: there's a LOOP in bignum-binary-gcd, and it loops 2014-04-23T10:28:42Z nyef: Hrm. Sounds like a good time to use /HEXSTR, especially given that the values should be unboxed at that point... 2014-04-23T10:29:20Z stassats: (zerop (%bignum-ref r 0)) is always NIL 2014-04-23T10:32:13Z stassats: is sub-w/borrow correct? 2014-04-23T10:32:32Z nyef: Okay, killing the call changes leads to a fault at 30 interrupted in PARSE-LAMBDA-LIST-LIKE-THING. 2014-04-23T10:32:53Z nyef: I make no representation that any bignum VOPs are entirely correct. 2014-04-23T10:33:12Z stassats: parse-l.. is the multiple values thing 2014-04-23T10:33:22Z stassats: i think your patch is not incorrect 2014-04-23T10:33:26Z nyef: Yeah, that's what I'm remembering too. 2014-04-23T10:33:31Z nyef: And that's the conclusion that I just came to. 2014-04-23T10:33:51Z nyef: Going to put a commit together for it, push, and then look into bignum stuff again. 2014-04-23T10:34:45Z stassats: on x86, sub-w/borrow the carry flag is reversed from the carry-in 2014-04-23T10:35:14Z nyef: Hardware carry flag behavior varies by arch. 2014-04-23T10:35:23Z nyef: It might-or-might-not be an issue. 2014-04-23T10:35:33Z nyef: ... Small test case, verified against x86? 2014-04-23T10:36:31Z stassats: haven't finished thinking hard yet 2014-04-23T10:36:37Z nyef: Fair enough. 2014-04-23T10:37:04Z nyef: Although, this would be more the write-down-the-problem stage. (-: 2014-04-23T10:38:19Z nyef: Commits pushed. 2014-04-23T10:41:09Z stassats: thinking hard is hard 2014-04-23T10:44:33Z stassats: it seems to be wrong indeed 2014-04-23T10:45:03Z stassats: compared to x86 2014-04-23T10:45:29Z nyef: What do you have for results as compared with your test oracle? 2014-04-23T10:56:59Z stassats: mixed up by 1 2014-04-23T10:57:07Z stassats: will try the right thing later 2014-04-23T10:58:05Z nyef: So, the carry output is 0 when it should be 1 and vice versa? 2014-04-23T11:04:54Z stassats: it's names it's a "borrow" 2014-04-23T11:07:50Z nyef: My intuition is not currently tuned for binary subtraction, with or without a carry or borrow input or output. 2014-04-23T11:08:16Z nyef: Especially since the semantics tend to vary on a per-CPU basis, and I deal with several different CPU types. 2014-04-23T11:16:08Z nyef: ... Still seems to fail at the same point. 2014-04-23T11:26:44Z prxq quit (Quit: Leaving) 2014-04-23T11:34:30Z stassats: at least something got fixed 2014-04-23T11:38:13Z stassats: or not, i modified (inst rsbs temp c 0) to (inst rsbs temp c 1), it's still wrong 2014-04-23T11:38:45Z stassats: i probably have no idea what rsbs does 2014-04-23T11:39:16Z attila_lendvai quit (Quit: Leaving.) 2014-04-23T11:39:55Z nyef: Reverse subtract. 2014-04-23T11:40:13Z nyef: It's like subs, only the argument order is reversed. 2014-04-23T11:40:31Z nyef: But you can't get a carry output from it (there's no rsbc or rsbcs). 2014-04-23T11:40:40Z nyef: Or is it borrow at that point? Whichever. 2014-04-23T11:41:27Z stassats: i'm confused alright 2014-04-23T11:42:04Z stassats: it should 0 c then? 2014-04-23T11:42:38Z nyef: Yeah, it's temp = 0 - c. 2014-04-23T11:42:47Z stassats: except that it can't be an immediate 2014-04-23T11:43:15Z stassats: wait, i'm confused again 2014-04-23T11:43:49Z nyef: Reverse-subtract is for the case where you want an immediate on the left side. 2014-04-23T11:44:11Z stassats: i don't know what i want anymore 2014-04-23T11:45:48Z afleck quit (Read error: Operation timed out) 2014-04-23T11:45:56Z afleck joined #sbcl 2014-04-23T11:46:33Z jdz quit (Quit: Leaving...) 2014-04-23T11:48:21Z nyef: I think I have an angle for making sure that both add-w/carry and sub-w/borrow are correct. 2014-04-23T11:50:03Z nyef: Two angles, really. 2014-04-23T11:50:49Z stassats: so, what is needed is carry=NOT C 2014-04-23T11:51:00Z stassats: or the first bit of C 2014-04-23T11:52:34Z stassats: what about CMP C, 1? 2014-04-23T11:54:17Z nyef: Umm... positive-fixnum, so it'd have to be 4, but that might work. 2014-04-23T11:57:39Z stassats: with CMP C, 1, i get up to case-failure in type-intersection2 2014-04-23T11:58:24Z nyef: I'm clearly not thinking straight, still. 2014-04-23T12:00:23Z nyef: C is a fixnum. 1 means "no borrow from right", BORROW is an unsigned, 1 means "no borrow from left". Both are actually BITs. 2014-04-23T12:01:12Z stassats: 4 - 1 => C, no borrow, 0 - 1, borrow? 2014-04-23T12:01:23Z nyef: Ignore the bit about 4 for now. 2014-04-23T12:01:38Z nyef: I'm still semi-confused. 2014-04-23T12:01:45Z stassats: well, it works 2014-04-23T12:02:15Z stassats: both in the wild in and my tests are equal to x86 output 2014-04-23T12:03:46Z stassats: it may break with 32-bit fixnums, though 2014-04-23T12:05:00Z eudoxia joined #sbcl 2014-04-23T12:08:55Z stassats: and case-failure due to round-numeric-bound having #+-sb-xc-host due to floats 2014-04-23T12:09:38Z nyef: Oh, there is a reverse subtract with carry. Neat. 2014-04-23T12:11:22Z stassats: ok, i'm back to !pathname-cold-init 2014-04-23T12:11:43Z stassats: for some reason it think logical pathname has a funky character 2014-04-23T12:13:27Z stassats: "illegal character for logical pathname:~% ~S", that's in R3, the character itself is somewhere on the stack 2014-04-23T12:14:09Z stassats: and it's #\: 2014-04-23T12:14:13Z stassats: weird... 2014-04-23T12:17:01Z stassats: it's parsing "SYS:SRC;**....." 2014-04-23T12:22:54Z stassats: broken character VOPs? 2014-04-23T12:23:28Z stassats: the error is in (unless (member ch '(#\; #\: #\.)) (error "illegal char")) 2014-04-23T12:28:36Z nyef: Broken character VOPs wouldn't be entirely unsurprising. 2014-04-23T12:29:55Z stassats: so, do you accept CMP C, 1 as a solution? 2014-04-23T12:30:08Z nyef: Build still going, but probably. 2014-04-23T12:31:23Z nyef: We don't have character comparison VOPs. 2014-04-23T12:31:50Z stassats: it should be using EQ there 2014-04-23T12:32:01Z nyef: Okay, fair point. 2014-04-23T12:32:28Z stassats: but, looking and char-code/code-char, it performs shifting, but wouldn't the shifting be done by the move vops? 2014-04-23T12:33:04Z stassats: at least that's what happens on x86 2014-04-23T12:33:08Z stassats: though ppc shifts too 2014-04-23T12:33:41Z stassats: oh, x86 returns unsigneds 2014-04-23T12:35:05Z nyef: We should probably return unsigneds as well, saves us double-shifting in some circumstances, plus lets us elide the code entirely in LOCATION= cases. 2014-04-23T12:35:19Z nyef: Well, maybe not LOCATION= if it accounts for SC, but you know what I mean. 2014-04-23T12:35:27Z nyef: Still, that's an optimization for later. 2014-04-23T12:35:48Z stassats: ok, so, (member #\: '(#\; #\: #\.)) returns NIL 2014-04-23T12:35:55Z stassats: let's isolate that 2014-04-23T12:37:39Z stassats: maybe it's not boxed properly, since it's coming from a string, unboxed 2014-04-23T12:39:27Z stassats: (as in, that member is opencoded) 2014-04-23T12:40:13Z nyef: Okay, I'm at an ecase failure in round-numeric-bound, which IIRC is one of the two egregious bits of float-disablement that actually affect other backends. 2014-04-23T12:40:42Z stassats: i just removed the conditionals 2014-04-23T12:41:15Z stassats: a (defun test (char) (member char '(#\; #\: #\.))) works, (defun test (string) (member (schar string 0) '(#\; #\: #\.))) doesn't 2014-04-23T12:45:21Z nyef: So, possibly array access VOPs? 2014-04-23T12:46:00Z stassats: http://paste.lisp.org/display/142258 2014-04-23T12:46:14Z stassats: three MOV #, 66 looks suspcious 2014-04-23T12:46:49Z stassats: 66 being character-widetag 2014-04-23T12:46:55Z nyef: Yes, that's a little bit odd. 2014-04-23T12:49:17Z stassats: (eq (char string 0) #\:) fails, obviously 2014-04-23T12:49:28Z nyef: Have a look at compiler/arm/move, the LOAD-IMMEDIATE move-fun, character case, the definition of ENCODED-CHARACTER. 2014-04-23T12:50:04Z nyef: Shouldn't it be (DPB CODEPOINT (BYTE 24 8) CHARACTER-WIDETAG)? 2014-04-23T12:51:45Z stassats: that sounds more correct 2014-04-23T12:56:32Z stassats: i'm at float-cold-init-or-reinit 2014-04-23T12:56:38Z nyef: Okay, I'm at pathname-cold-init. 2014-04-23T12:56:48Z nyef: Let me guess... floating-point-mods? 2014-04-23T12:56:53Z nyef: Err... -modes? 2014-04-23T12:56:59Z stassats: dpb above fixes pathname-cold-init 2014-04-23T12:57:20Z stassats: you guessed right 2014-04-23T12:59:02Z nyef: I should be there soon, I have some commits to put together, and some paying work to deal with, but the mode control should at least be straightforward (ha-ha). 2014-04-23T12:59:23Z stassats: i'll skip over it 2014-04-23T13:00:52Z stassats: now at !pprint-cold-init 2014-04-23T13:01:15Z stassats: that's cold-init almost done 2014-04-23T13:01:51Z nyef: ... First use of compiler, isn't it? 2014-04-23T13:01:57Z stassats: it is 2014-04-23T13:02:05Z stassats: vector-t-p is failing out 2014-04-23T13:02:36Z stassats: the compilation is not that far, just at make-lisp-source-info 2014-04-23T13:12:15Z nyef: Okay, I'm at set-floating-point-modes. 2014-04-23T13:19:24Z segv- joined #sbcl 2014-04-23T13:29:47Z pranavrc quit 2014-04-23T13:42:40Z oleo joined #sbcl 2014-04-23T13:49:16Z jdz joined #sbcl 2014-04-23T14:05:56Z yacks quit (Ping timeout: 252 seconds) 2014-04-23T14:11:49Z stassats: back to vector-t-p, looks like it full-calls typep 2014-04-23T14:12:56Z nyef: In pprint-cold-init, doing the first set-pprint-dispatch, does a few "//string-ouch"es, an "/interning ~A", then fault at 0 from somewhere in the runtime. 2014-04-23T14:13:28Z stassats: that's vector-t-p 2014-04-23T14:14:37Z nyef: PC disagrees between the memory-fault report and the context on the stack... 2014-04-23T14:15:00Z stassats: just noticed that 2014-04-23T14:16:27Z nyef: The fault address is given as somewhere between _start and alloc_cons, the PC in the context is in undefined_tramp. 2014-04-23T14:17:22Z nyef: We're probably close enough to the start of the lisp control stack here that the SIGTRAP handler is DOA. 2014-04-23T14:17:34Z stassats: may be it's calling ERROR? 2014-04-23T14:17:56Z slyrus joined #sbcl 2014-04-23T14:17:58Z stassats: shouldn't, because it gets a simple-vector as an argument 2014-04-23T14:18:18Z nyef: No, LEXENV in the context is an FDEFN for SIMPLE-ARRAY-COMPLEX-DOUBLE-FLOAT-P. 2014-04-23T14:18:21Z stassats: but it has two full calls to something, apparently typep, which should have got opencoded 2014-04-23T14:18:41Z nyef: I bet there's a bunch of disabled float stuff in compiler/generic. 2014-04-23T14:19:09Z stassats: *saetp* has single and double arrays disabled 2014-04-23T14:19:11Z nyef: Yeah, four cases in generic/vm-array seem likely. 2014-04-23T14:19:34Z nyef: And complex-single and complex-double, yes. 2014-04-23T14:19:38Z nyef: That'd probably do it. 2014-04-23T14:21:04Z nyef: Building now... 2014-04-23T14:21:20Z stassats: data-vector-set/simple-array-single-flaot is not the name of a defined VOP 2014-04-23T14:21:28Z nyef: Ah, damn. 2014-04-23T14:22:03Z stassats: "should be easy" 2014-04-23T14:24:15Z stassats: but what does it have to do with vector-t-p? 2014-04-23T14:25:01Z nyef: I have no idea, but I'm thinking that this needs addressing anyway and might fix the bit that I'm hitting. 2014-04-23T14:25:35Z yacks joined #sbcl 2014-04-23T14:27:47Z nyef: Okay, I need to deal with other stuff for a bit, then I'll come back to this. 2014-04-23T14:31:35Z stassats: i wrote my own vector-t-p which is just checks for simple-vector, and it fails too 2014-04-23T14:32:33Z stassats: it's implemented as a single tail-call, probably that's when typep calls SIMPLE-ARRAY-COMPLEX-DOUBLE-FLOAT-P on it 2014-04-23T14:33:57Z nyef: This is actually the last part of cold-init. 2014-04-23T14:34:24Z nyef: From here on out, things get nastier with signal handling. 2014-04-23T14:39:52Z attila_lendvai joined #sbcl 2014-04-23T14:39:52Z attila_lendvai quit (Changing host) 2014-04-23T14:39:52Z attila_lendvai joined #sbcl 2014-04-23T14:40:55Z stassats: i replaced vector-t-p definition with simple-vector-p, now i fail at 0 in make-core-component 2014-04-23T14:41:44Z nyef: fiveop set up allocate-code-object, you might desk-check it. 2014-04-23T14:42:03Z nyef: At least you're not running out of heap anymore, right? (-: 2014-04-23T14:42:23Z stassats: not yet 2014-04-23T14:59:53Z stassats: sanctify-for-execution missing? 2014-04-23T15:00:23Z nyef: Almost certainly, yes. 2014-04-23T15:00:31Z nyef: We might need to look up how to do that. 2014-04-23T15:00:45Z stassats: i added a stub 2014-04-23T15:00:59Z stassats: (that's what's being called by make-core-component) 2014-04-23T15:01:13Z nyef: Ahh. 2014-04-23T15:03:38Z erikvarga joined #sbcl 2014-04-23T15:04:11Z drmeister joined #sbcl 2014-04-23T15:08:06Z stassats: next failure make-compiled-debug-fun 2014-04-23T15:09:49Z nyef: Next time I decide to write a new backend from scratch, please remind me to pick an architecture with more than 15 registers. 2014-04-23T15:10:47Z Krystof: how many does arm64 have? 2014-04-23T15:11:05Z nyef: Don't know, but I heard that it was more. 2014-04-23T15:11:10Z stassats: 31? 2014-04-23T15:11:27Z Krystof: if it's more, that's fine 2014-04-23T15:11:46Z stassats: did you expect it to be less? 2014-04-23T15:12:01Z nyef: Whatever it is, it's someone else's turn. 2014-04-23T15:15:13Z stassats: doing first GC 2014-04-23T15:15:23Z stassats: memory fault at 76fff000 2014-04-23T15:15:48Z Krystof: \o/ 2014-04-23T15:16:01Z nyef: I think that's on me. 2014-04-23T15:16:12Z stassats: and something's wrong array type checks 2014-04-23T15:16:18Z stassats: i had to get rid of them to proceed 2014-04-23T15:16:26Z stassats: it calls an undefined somewhere 2014-04-23T15:16:42Z nyef: Yeah, a general cleanup of disabled whatevers is also on me. 2014-04-23T15:16:53Z nyef: (Or on fiveop, if he gets to it first.) 2014-04-23T15:17:35Z nyef: An interesting question, btw, is if you can do a native x86oid build off of the arm port tree. 2014-04-23T15:17:50Z nyef: (That is, did we break anything?) 2014-04-23T15:18:28Z stassats: i think i diverged from your state by disabling set-floating-point-modes and vector typechecks, i'm yet to figure out what's wrong with typechecks 2014-04-23T15:18:57Z nyef: I'm still working on float array vops. 2014-04-23T15:19:29Z nyef: Have a grep for arm conditionals in src/compiler and src/code, see if anything is obvious. 2014-04-23T15:19:53Z stassats: tried already 2014-04-23T15:20:03Z stassats: will try anew 2014-04-23T15:20:22Z nyef: Might be that sorting the float array vops fixes the type checks. 2014-04-23T15:20:38Z stassats: oh, and i added a stab for sanctify-for-execution 2014-04-23T15:20:52Z stassats: or a stub 2014-04-23T15:22:22Z stassats: EQL has some float #-arms 2014-04-23T15:22:27Z stassats: %eql 2014-04-23T15:23:19Z nyef: There are going to be a number of them. I'll probably do the SAP VOPs and then do a search-and-destroy on the conditionals. 2014-04-23T15:24:31Z stassats: what if i disable the first gc 2014-04-23T15:26:32Z stassats: a REPL, can you believe it? 2014-04-23T15:26:50Z stassats: * (+ 10 #C(20 10)) => #C(30 10) 2014-04-23T15:26:58Z nyef: Oh, wow. 2014-04-23T15:27:09Z nyef: The classic REPL test is, of course (+ 3 4) => 9 2014-04-23T15:28:13Z Krystof: (* 2 3) 2014-04-23T15:28:31Z Krystof: http://www.advogato.org/person/crhodes/diary.html?start=12 2014-04-23T15:28:56Z nyef: Heh. 2014-04-23T15:30:16Z Krystof: also http://www.advogato.org/person/crhodes/diary/56.html 2014-04-23T15:30:22Z Krystof: and http://www.advogato.org/person/crhodes/diary/57.html 2014-04-23T15:30:58Z Krystof: but way cool 2014-04-23T15:31:36Z stassats: (1+ most-positive-fixnum) => INFO: Control stack guard page unprotected 2014-04-23T15:31:46Z nyef: Heh. 2014-04-23T15:31:56Z Krystof: the simplest tests are the best 2014-04-23T15:32:12Z nyef: Yeah, things are generally still screwy, but this is awesome, and I want to have a REPL too! 2014-04-23T15:32:28Z stassats: i want to have DISASSEMBLE 2014-04-23T15:32:56Z nyef: Hrm. It got further this time. 2014-04-23T15:33:13Z nyef: Yeah, sanctify-for-execution. 2014-04-23T15:33:41Z stassats: the repl works without sanctify-for-execution, so, is it needed? 2014-04-23T15:34:12Z Krystof: can you actually compile code? 2014-04-23T15:34:14Z nyef: It depends on the underlying memory cache model. 2014-04-23T15:34:29Z Krystof: (funcall (compile nil '(lambda (x y) (* x y))) 2 3) 2014-04-23T15:34:45Z nyef: I want to implement DISASSEMBLE, but I'm also fairly sure I'm nearing burnout on this little project and need some downtime from it. 2014-04-23T15:35:12Z stassats: (funcall (compile nil '(lambda (x) (+ 30 x))) 20) .. /show things => 50 2014-04-23T15:35:12Z nyef: I definitely call dibs on DISASSEMBLE, though! 2014-04-23T15:35:57Z stassats: i can test my vector typep in the repl! 2014-04-23T15:36:22Z Krystof: I remember when I got enough of swankr working to continue developing the thing in itself 2014-04-23T15:36:24Z Krystof: it's a nice moment 2014-04-23T15:36:40Z stassats: (funcall (compile nil '(lambda (x) (typep x '(simple-array t (*))))) #()) fails 2014-04-23T15:37:03Z nyef: Does getting to sanctify-for-execution imply working vector-typep things? 2014-04-23T15:37:28Z stassats: no, vector-typep fails before 2014-04-23T15:37:46Z stassats: make that a yes then? 2014-04-23T15:37:59Z nyef: Yeah, I was wondering for a bit there. (-: 2014-04-23T15:38:55Z nyef: I think this is actually a pretty decent working arrangement, with you running out front, finding and localizing the issues, then using some kludgy workaround, while I follow along and actually fix the problems that you've tracked down. 2014-04-23T15:39:52Z stassats: it's certainly in a better state than it was on sunday 2014-04-23T15:42:35Z nyef: We need sanctify, the I and D caches aren't necessarily coherent, depending on implementation. 2014-04-23T15:46:20Z stassats: i need to benchmarks something 2014-04-23T15:51:32Z stassats: i can't seem to be able to loop 2014-04-23T15:52:44Z stassats: i guess it's still those vector typep stuff 2014-04-23T15:52:52Z stassats: (tagbody tag (print 'hello) (go tag)) works 2014-04-23T15:56:14Z ltbarcly joined #sbcl 2014-04-23T15:56:28Z nyef: Okay, stubbed sanctify for now... with the un-stubbed lisp side available in the file so that I have the hook to do it properly this evening. 2014-04-23T15:56:56Z stassats: i'll wait for whatever you did do get vector-typep working before proceeding 2014-04-23T15:58:17Z nyef: That'll probably be this evening, I really do need to set this aside for a bit to get some other stuff done. 2014-04-23T15:58:54Z nyef: fiveop should be getting a bit of a surprise tomorrow. 2014-04-23T16:16:15Z slyrus_ joined #sbcl 2014-04-23T16:17:32Z slyrus quit (Ping timeout: 252 seconds) 2014-04-23T16:17:33Z slyrus_ is now known as slyrus 2014-04-23T16:20:22Z nyef: I have a REPL (disabled the whole GC thing for now). 2014-04-23T16:20:35Z nyef: stassats: What LOOP form were you trying that didn't work? 2014-04-23T16:20:47Z stassats: (dotimes (i 10) (print 'hello)) 2014-04-23T16:21:08Z nyef: WFM. 2014-04-23T16:21:33Z stassats: it's the vector thing then 2014-04-23T16:21:41Z stassats: push your changes! 2014-04-23T16:21:44Z nyef: Yeah, about what we both figured. 2014-04-23T16:22:15Z nyef: I've got paying work to deal with, I just didn't want to be this --> <-- close to a REPL and not get there before taking a break. 2014-04-23T16:27:56Z drmeister quit (Ping timeout: 276 seconds) 2014-04-23T16:35:22Z Krystof: nyef: amusingly those arrows had a linebreak between them in my client :) 2014-04-23T16:35:31Z nyef: Heh. 2014-04-23T16:35:41Z stassats: mine too 2014-04-23T16:36:08Z slyrus quit (Ping timeout: 255 seconds) 2014-04-23T16:39:14Z davazp joined #sbcl 2014-04-23T16:45:16Z erikvarga left #sbcl 2014-04-23T16:45:28Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-23T16:55:53Z davazp quit (Ping timeout: 276 seconds) 2014-04-23T17:08:21Z ltbarcly joined #sbcl 2014-04-23T17:08:47Z jdz quit (Remote host closed the connection) 2014-04-23T17:09:02Z scymtym_ joined #sbcl 2014-04-23T17:39:10Z ltbarcly: so I see a lot of references to the compiler giving notes on optimization, but I don't get any? 2014-04-23T17:39:22Z ltbarcly: how do you get the compiler to output notes on what it wasn't able to optimize? 2014-04-23T17:58:10Z stassats: (optimize speed) 2014-04-23T18:14:05Z jdz joined #sbcl 2014-04-23T18:14:55Z Shinmera joined #sbcl 2014-04-23T18:15:10Z jdz quit (Remote host closed the connection) 2014-04-23T18:15:45Z jdz joined #sbcl 2014-04-23T18:19:35Z sdemarre joined #sbcl 2014-04-23T18:25:37Z eudoxia quit (Quit: leaving) 2014-04-23T18:30:32Z jdz_ joined #sbcl 2014-04-23T18:32:28Z Shinmera: I'm trying to compile SBCL on Windows using 32x MinGW/MSYS. It compiles fine for a while and then shortly after "beginning GENESIS" barfs up: http://pastebin.com/8D5aVYaf Am I missing something obvious that I should do beforehand? 2014-04-23T18:34:13Z jdz quit (Ping timeout: 252 seconds) 2014-04-23T18:34:44Z Intensity joined #sbcl 2014-04-23T18:38:15Z oleo: adjust your #include stuff or your specify your includedir ? 2014-04-23T18:38:27Z oleo: where are you calling that from ? 2014-04-23T18:38:36Z oleo: mingws shell ? 2014-04-23T18:39:22Z Shinmera: Msys shell and then simply ./make.sh 2014-04-23T18:41:11Z Shinmera: Maybe I've set MSYS up badly, I'll try around a bit 2014-04-23T18:41:42Z oleo: yes try that... 2014-04-23T18:42:51Z oleo: otherwise try to give an -I/usr/include relative to its root....i think that's the problme maybe.... 2014-04-23T18:43:01Z jdz joined #sbcl 2014-04-23T18:43:14Z Shinmera: alright 2014-04-23T18:46:11Z jdz_ quit (Ping timeout: 255 seconds) 2014-04-23T18:56:59Z Shinmera: I installed MSYS through a setup instead of through MinGW's package downloader and it passes beyond the last point of failure now. 2014-04-23T18:57:54Z Shinmera: Sadly it still can't compile completely and now shows a backtrace: http://pastebin.com/mCX92GAp 2014-04-23T18:58:11Z stassats: that paste is useless 2014-04-23T18:58:19Z Shinmera: I assumed as much 2014-04-23T18:58:20Z stassats: the error was before that 2014-04-23T18:59:20Z Shinmera: Before that it shows a bunch of style warnings and mentions that it finished writing unix.lisp-obj-tmp 2014-04-23T18:59:37Z stassats: before that 2014-04-23T19:02:30Z dsp_ joined #sbcl 2014-04-23T19:04:44Z Shinmera: Am I looking for a C compiler warning/error or a Lisp one? 2014-04-23T19:04:55Z stassats: lisp, of course 2014-04-23T19:05:26Z Shinmera: Farther up I'm seeing more style warnings about undefined functions and unable to optimize notes 2014-04-23T19:06:10Z slyrus joined #sbcl 2014-04-23T19:07:58Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-23T19:08:40Z stassats: style warnings are of no interest 2014-04-23T19:11:32Z stassats: Shinmera: did you try latest windows binaries from sbcl.org? 2014-04-23T19:11:48Z Shinmera: The binary with the installer runs fine 2014-04-23T19:11:57Z stassats: 1.1.17? 2014-04-23T19:11:59Z Shinmera: Yes 2014-04-23T19:12:07Z stassats: cool, i made those 2014-04-23T19:12:22Z stassats: haven't had any feedback yet 2014-04-23T19:12:53Z Shinmera: I can ignore anything above the line where it says 'compiling file "c:/sbcl-1.1.17/src/code/unix.lisp"' though, I hope? 2014-04-23T19:13:05Z stassats: yes 2014-04-23T19:13:26Z stassats: you can paste the output starting from there until Unhandled SIMPLE-ERROR in thread # 5040 2014-04-23T20:03:21Z pkhuong: stassats: can it do ackermann? 2014-04-23T20:04:52Z afleck left #sbcl 2014-04-23T20:05:02Z stassats: (ackermann 3 10) => 8189 2014-04-23T20:05:09Z stassats: TIME is broken somewhere at FORMAT 2014-04-23T20:06:51Z stassats: probably due to sb! packages 2014-04-23T20:07:44Z stassats: i haven't tried make-tartet-2 yet! 2014-04-23T20:08:34Z stassats: ok, that hanged the phone 2014-04-23T20:15:28Z Shinmera: Hmm 2014-04-23T20:16:04Z Shinmera: Do I have to activate some compiler options to make my binary include SB-POSIX? It runs fine otherwise, but quicklisp requires it and it's apparently missing. 2014-04-23T20:16:35Z Shinmera: The installed binary works fine, so I'm guessing I didn't include something properly. 2014-04-23T20:17:11Z stassats: 1.1.17 has a broken installer maker 2014-04-23T20:17:47Z stassats: 6380ff7405e3dca2d04bbd1e9f2ac3c95efd6c12 has it fixed 2014-04-23T20:17:51Z Shinmera: Do you mean install.sh is broken or? 2014-04-23T20:18:18Z stassats: maybe install.sh is broken too, but make-windows-installer is certainly broken 2014-04-23T20:18:35Z Shinmera: I didn't make an installer package 2014-04-23T20:19:43Z stassats: (truncate (expt 2 32) 10) => 21074691228794313799883120286105 6 2014-04-23T20:20:20Z Shinmera: I just did make.sh --prefix=/c/sbcl/ && install.sh, which generates a nicely working setup, except for the missing sb-posix. 2014-04-23T20:20:31Z stassats: only sb-posix? 2014-04-23T20:20:43Z stassats: its test may have failed, you need to touch some file 2014-04-23T20:20:45Z Shinmera: I'm not sure if anything else is missing, but that's what I hit 2014-04-23T20:20:56Z stassats: try loading sb-bsd-posix? 2014-04-23T20:21:00Z stassats: sockets 2014-04-23T20:21:57Z Shinmera: Hmm, I'm seeing a warning! note at the end of the make, saying that some self-tests failed for sb-gmp and sb-posix. 2014-04-23T20:22:21Z stassats: touch ./obj/asdf-cache/sb-posix/test-passed.test-report 2014-04-23T20:23:16Z edgar-rft quit (Quit: lifetime disappeared by mental meltdown) 2014-04-23T20:24:27Z Shinmera: Ah, that did the trick. Thanks a lot! 2014-04-23T20:24:55Z nyef: Your next challenge: Figure out which sb-posix test failed, and why. 2014-04-23T20:25:00Z ltbarcly joined #sbcl 2014-04-23T20:25:01Z stassats: (integer-length (truncate (expt 2 32) 10)) => 105 2014-04-23T20:25:14Z Shinmera: nyef: I might do that if I find the time for it tomorrow. 2014-04-23T20:25:14Z stassats: nyef: i've seen it fail sometime 2014-04-23T20:25:26Z stassats: something inconsequential 2014-04-23T20:25:42Z stassats: but, (integer-length (truncate (expt 2 33) 10)) => 30 2014-04-23T20:25:53Z nyef: Well, it's windows, it's by definition inconsequential. (-: 2014-04-23T20:28:28Z stassats: (list (sb!bignum::%bignum-length (expt 2 32)) (sb!bignum::%bignum-length (expt 2 33))) => (3 2) 2014-04-23T20:30:21Z stassats: (sb!bignum::%bignum-length (* most-positive-fixnum most-positive-fixnum most-positive-fixnum)) => 4, instead of 3 2014-04-23T20:30:39Z stassats: having a REPL is so much better 2014-04-23T20:30:57Z nyef: %bignum-length is get-header-data, FWIW. 2014-04-23T20:31:26Z nyef: But yeah, having a REPL is massively more convenient than having to hack the build order. 2014-04-23T20:32:58Z stassats: i don't think get-header-data is wrong, but the computation is making the length wrong 2014-04-23T20:33:05Z nyef: Yeah, that's fair. 2014-04-23T20:34:24Z stassats: is undefined_tramp broken? 2014-04-23T20:34:32Z stassats: (not like the debugger is not broken) 2014-04-23T20:34:46Z nyef: SIGTRAP in general is broken. 2014-04-23T20:35:16Z nyef: Fairly high on my list to fix, actually, because it would mean a much more robust REPL. 2014-04-23T20:41:14Z nyef: Ooh. (make-array 0 :element-type 'nil) behaves as expected. 2014-04-23T20:43:47Z nyef: Pushed sanctify implementation. 2014-04-23T20:48:27Z cmack quit (Remote host closed the connection) 2014-04-23T21:09:14Z nyef: Getting internal-error traps to work is going to be such an INTERESTING dance. Or, as Krystof put it, "hilarious". 2014-04-23T21:12:48Z Shinmera: Now I finally got my programming environment on an usb stick running on both linux and windows. Thanks again for the troubleshooting! 2014-04-23T21:13:37Z oleo quit (Ping timeout: 240 seconds) 2014-04-23T21:16:14Z attila_lendvai quit (Quit: Leaving.) 2014-04-23T21:25:05Z stassats: i've been dancing around it, and it has been (%bignum-length (%allocate-bignum 2)) => 3 all along 2014-04-23T21:26:12Z nyef: Let me know when you have anything localized enough for me to take a look at. 2014-04-23T21:27:15Z Shinmera quit (Quit: ZzZZ) 2014-04-23T21:27:24Z oleo joined #sbcl 2014-04-23T21:30:51Z stassats: first suspect: var-alloc 2014-04-23T21:40:26Z stassats: it includes the header in the size? 2014-04-23T21:41:10Z nyef: That's not right... 2014-04-23T21:41:27Z nyef: I'll check it in a couple minutes. 2014-04-23T21:43:44Z nyef: Hrm. I don't understand right now what that's doing. 2014-04-23T21:44:05Z yacks quit (Ping timeout: 264 seconds) 2014-04-23T21:44:25Z yacks joined #sbcl 2014-04-23T21:44:25Z stassats: words is the fixed amount an object takes, extra is the variable 2014-04-23T21:44:52Z nyef: I have too much state in my head right now to seriously dig into it. 2014-04-23T21:44:53Z stassats: so, 1 word for the bignum header and whatever the amount of digits extra 2014-04-23T21:45:26Z nyef: I note that PPC seems to have some sort of bias of words going on, I didn't check any of hte other backends. 2014-04-23T21:49:44Z stassats: looks like it improperly performs unfixnumization 2014-04-23T21:52:31Z stassats: or both 2014-04-23T21:53:27Z nyef: Okay, looks like I do need to pay attention now. 2014-04-23T21:53:55Z nyef: var-alloc and unfixnumization, you said? 2014-04-23T21:54:43Z stassats: it either unfixnumizes incorrectly, or adds the header size unneededly 2014-04-23T21:54:54Z stassats: i'm decoding the x86 var-alloc 2014-04-23T21:55:39Z nyef: It's worse than that. 2014-04-23T21:55:55Z stassats: i don't get x86's version 2014-04-23T21:57:59Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-23T21:58:00Z nyef: In var-alloc, change (* words n-word-bytes) to use (1- words). In the following add bytes bytes n-word-bytes, use (* 2 n-word-bytes). 2014-04-23T21:58:49Z stassats: what does that do? i can't think straight 2014-04-23T21:59:45Z stassats: ah, doesn't muck up the header with the actual size? 2014-04-23T21:59:53Z nyef: I can't think straight either, but basically the "normal" path (PPC and MIPS) set the header word to be too *LARGE* by two words, then take them out when applying the type tag. 2014-04-23T22:00:24Z nyef: We can't do that because of our constraints on immediate bytes. 2014-04-23T22:00:34Z stassats: those kinds of optimizations are the worst 2014-04-23T22:00:47Z nyef: Only when they're not explained. 2014-04-23T22:00:58Z stassats: they usually aren't, like here 2014-04-23T22:02:12Z stassats: (%bignum-length (%allocate-bignum 1)) => 1 2014-04-23T22:02:32Z stassats: * (expt 2 32) => 4294967296, (expt 2 33) => 8589934592 2014-04-23T22:03:09Z nyef: So, it's working better now? 2014-04-23T22:03:24Z stassats: ostensibly 2014-04-23T22:03:38Z nyef: Okay, I might have more good news once my build is done. 2014-04-23T22:04:04Z stassats: and i figured this one completely from the REPL 2014-04-23T22:04:52Z stassats: is alien not yet implemented? 2014-04-23T22:05:00Z stassats: sleep works, get-timezone does not 2014-04-23T22:05:15Z nyef: It's at least semi-implemented. 2014-04-23T22:05:30Z nyef: Streams require alien, after all. 2014-04-23T22:06:24Z stassats: i'm not sure whether to run make-tartet-2.sh again or not 2014-04-23T22:06:34Z nyef: Not yet. 2014-04-23T22:06:39Z stassats: it didn't end well last time 2014-04-23T22:06:42Z stassats: had to reboot 2014-04-23T22:06:46Z nyef: It won't end well this time, either. 2014-04-23T22:07:00Z stassats: i need to rename the packages 2014-04-23T22:07:15Z nyef: Don't know about having to reboot, but it still won't be pretty. 2014-04-23T22:08:03Z stassats: package renaming falters at %sp-string-compare 2014-04-23T22:08:22Z stassats: (and this whole arm porting session is an exercise in coming up with different words for "fail") 2014-04-23T22:08:30Z nyef: Heh. 2014-04-23T22:11:35Z stassats: screw pacakge remaning, i'll write my own TIME 2014-04-23T22:11:56Z nyef: http://paste.lisp.org/display/142264 2014-04-23T22:12:04Z drmeister joined #sbcl 2014-04-23T22:12:17Z stassats: a backtrace? 2014-04-23T22:12:35Z nyef: LDB backtrace is useless here. 2014-04-23T22:14:19Z nyef: Oh, hey. I bet we don't have CHAR=. 2014-04-23T22:15:01Z stassats: isn't it eql? 2014-04-23T22:15:31Z nyef: It might be if we don't have char=. 2014-04-23T22:15:55Z slyrus_ joined #sbcl 2014-04-23T22:16:02Z stassats: (char= #\: #\:) => T 2014-04-23T22:16:20Z slyrus quit (Ping timeout: 255 seconds) 2014-04-23T22:16:37Z slyrus_ is now known as slyrus 2014-04-23T22:18:01Z jdz quit (Remote host closed the connection) 2014-04-23T22:21:56Z stassats: i get random failures 2014-04-23T22:22:26Z nyef: Do you mean... intermittent? Or failures when calling RANDOM? 2014-04-23T22:22:34Z stassats: intermittent 2014-04-23T22:22:42Z cmack joined #sbcl 2014-04-23T22:22:45Z nyef: Lovely. 2014-04-23T22:23:09Z nyef: And they're not from running out of heap? 2014-04-23T22:23:27Z stassats: i restart, do the same thing, it succeeds 2014-04-23T22:23:37Z nyef: Hrm. That's not good. 2014-04-23T22:23:53Z Krystof: how sure are you that you have sanitize-for-execution right? 2014-04-23T22:24:01Z Krystof: what kinds of things are you doing? 2014-04-23T22:24:02Z nyef: Ah, good point. 2014-04-23T22:24:05Z stassats: i don't have it at all 2014-04-23T22:24:15Z stassats: but anyway, i just calculated a factorial of 50000 2014-04-23T22:25:12Z nyef: Manually-triggered GC is in business. I'll start prepping the commits. 2014-04-23T22:25:57Z stassats: i'll recompile with sanctify for execution 2014-04-23T22:25:59Z nyef: (Did a run of three GCs and they all returned to the REPL.) 2014-04-23T22:26:09Z stassats: i heard 3 is all you need 2014-04-23T22:26:24Z nyef: No, that was love, not three. 2014-04-23T22:27:08Z stassats: love as in 0? 2014-04-23T22:29:47Z stassats: and calculating the factorial took 7 seconds 2014-04-23T22:30:20Z stassats: just about 50 times slower than on the laptop 2014-04-23T22:31:13Z stassats: i think sanctify-for-execution it was, i now reliably can LOAD things 2014-04-23T22:31:45Z nyef: Okay, I'm going to go to dinner, then get back to sorting out commits from my working tree. 2014-04-23T22:32:19Z nyef: I think that my next big thing is going to be sorting out the interrupt handling mess. 2014-04-23T22:32:25Z nyef: ... Unless it isn't. 2014-04-23T22:35:31Z stassats: some random failures still present, but rarer 2014-04-23T22:37:44Z stassats: (fib 50) => 6277101738309684039858724727853387073209631205623600462197 2014-04-23T22:37:47Z stassats: that's not right 2014-04-23T22:37:51Z stassats: should be 12586269025 2014-04-23T22:38:00Z Krystof: it's nearly right 2014-04-23T22:38:21Z Krystof: * (format t "~X" 6277101738309684039858724727853387073209631205623600462197) 2014-04-23T22:38:21Z Krystof: 10000000200000004000000070000000C00000012EE333975 2014-04-23T22:39:20Z Krystof: * (format t "~D" #x2EE333975) 2014-04-23T22:39:20Z Krystof: 12586269045 2014-04-23T22:40:20Z segv- quit (Ping timeout: 252 seconds) 2014-04-23T22:45:17Z stassats: (ackermann 4 2) seems correct 2014-04-23T22:49:15Z stassats: (1+ 536870912) => 4831838210 2014-04-23T22:49:19Z stassats: that's for tomorrow 2014-04-23T22:49:36Z stassats: how many bugs can be there on bignums? 2014-04-23T22:57:19Z stassats quit (Read error: Operation timed out) 2014-04-23T22:57:56Z drmeister quit (Ping timeout: 276 seconds) 2014-04-23T23:10:42Z drmeister joined #sbcl 2014-04-23T23:24:22Z ltbarcly joined #sbcl 2014-04-23T23:26:11Z pkhuong quit (Ping timeout: 252 seconds) 2014-04-23T23:26:18Z pkhuong joined #sbcl 2014-04-23T23:32:16Z eudoxia joined #sbcl 2014-04-23T23:47:38Z ivan`` quit (Ping timeout: 258 seconds) 2014-04-23T23:48:10Z ivan`` joined #sbcl 2014-04-24T00:13:17Z slyrus quit (Ping timeout: 252 seconds) 2014-04-24T00:16:45Z zRecursive joined #sbcl 2014-04-24T00:28:40Z eudoxia quit (Read error: Connection reset by peer) 2014-04-24T00:38:11Z cmack` joined #sbcl 2014-04-24T00:39:18Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-24T00:39:47Z cmack quit (Ping timeout: 252 seconds) 2014-04-24T01:31:09Z ltbarcly joined #sbcl 2014-04-24T02:07:24Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-24T02:30:34Z nyef quit (Quit: G'night all.) 2014-04-24T02:38:41Z christoph_debian quit (Ping timeout: 264 seconds) 2014-04-24T02:46:00Z ltbarcly joined #sbcl 2014-04-24T02:52:55Z christoph_debian joined #sbcl 2014-04-24T02:55:14Z ehaliewicz joined #sbcl 2014-04-24T03:08:56Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-24T03:33:34Z ehaliewicz left #sbcl 2014-04-24T03:37:47Z attila_lendvai joined #sbcl 2014-04-24T03:37:48Z attila_lendvai quit (Changing host) 2014-04-24T03:37:48Z attila_lendvai joined #sbcl 2014-04-24T03:53:23Z cmack`` joined #sbcl 2014-04-24T03:54:48Z ooben joined #sbcl 2014-04-24T03:54:56Z cmack` quit (Ping timeout: 252 seconds) 2014-04-24T03:56:29Z ooben: I have a variable named "tmp" that is a list of basic-chars. How do I find how much space is used by "tmp"? And before anyone asks, my reason is "curiosity" 2014-04-24T03:56:56Z ooben: Sorry, it's a list of basic-string 2014-04-24T03:58:52Z zRecursive is now known as zRecursive{away} 2014-04-24T03:59:41Z scymtym_ quit (Ping timeout: 264 seconds) 2014-04-24T04:05:02Z kanru quit (Remote host closed the connection) 2014-04-24T04:05:22Z |3b|: first you define "space used by" :p 2014-04-24T04:05:38Z kanru joined #sbcl 2014-04-24T04:06:57Z |3b|: if you mean BASE-STRING, i think that needs 1 octet per character + a few words of header per string, + a few words per cons in the list 2014-04-24T04:07:50Z ooben: there's no sbcl function to use? i figured it would have to be in the internals somewhere. really i'd like to know space usage for any type t 2014-04-24T04:07:51Z |3b|: also, i think if it is taking up a large fraction of the heap, it should show up in ROOM 2014-04-24T04:08:52Z |3b|: 'size' isn't very well defined in CL, since most things just store references 2014-04-24T04:09:07Z |3b|: you could say your list of strings just takes up the size of N conses 2014-04-24T04:09:18Z |3b|: and the size of the strings isn't actually part of that 2014-04-24T04:10:34Z |3b|: or you could say TMP just takes up 1 word, since it doesn't actually contain the list, just a reference to the 1 cons at the head of the list 2014-04-24T04:12:14Z Bike: goes for C too, really, it's just that you don't ask about the size of recursive objects as much 2014-04-24T04:13:29Z ooben: i hate answers that invalidate the original question. but such is lisp haha 2014-04-24T04:14:35Z |3b|: best way to answer that sort of thing in sbcl is just allocate a bunch of things within TIME and see how many bytes consed it reports :) 2014-04-24T04:14:43Z |3b|: or at least easiest 2014-04-24T04:15:02Z |3b|: 'best' probably involves either not asking that question or knowing sbcl internals :p 2014-04-24T04:16:02Z ooben: are most people here devs? i was thinking about messing around with sbcl internals. i missed the Google SoC deadline but i'd still like to contribute something to a project i love 2014-04-24T04:16:27Z ooben: i'm thinking either sbcl or freebsd 2014-04-24T04:16:28Z |3b|: just keep in mind TIME reports allocation with a minimum granularity of a few k, so try for closer to a few MB... also, make sure it doesn't completely eliminate the allocation if the result is unused 2014-04-24T04:17:01Z |3b|: not sure about 'most here are devs', but most of the devs are here 2014-04-24T04:17:16Z ooben: how about this: are you a dev? 2014-04-24T04:17:17Z Bike: yeah, there's also interested bystanders like myself. 2014-04-24T04:17:32Z |3b| isn't really a dev, i just break things and try to track down bugs 2014-04-24T04:17:41Z ooben: i would quote your name but i'm still trying to figure out what character it is. no l or | 2014-04-24T04:17:55Z Bike: it's a pipe. | 2014-04-24T04:18:17Z ooben: oh, i guess it just looked different when bolded. well thanks for your help, |3b| 2014-04-24T04:18:40Z |3b|: yeah, default CL multiple escape character 2014-04-24T04:20:17Z |3b|: the devs (and us users too) would probably be happy to have more help though :) 2014-04-24T04:22:01Z drmeister quit (Remote host closed the connection) 2014-04-24T04:22:04Z ooben: i'd probably want to do something with threading a non-uniform memory access on Linux 2014-04-24T04:22:10Z ooben: *threading and 2014-04-24T04:22:56Z ooben: can multiple threads allocate memory at the same time or is there a global lock? what about multi-threaded GC? 2014-04-24T04:23:16Z White_Flame: One of the biggest issues with tracking space used is if symbols are involved. If all you're doing is following references that an object contains, then you're going to read the symbol-package, then everything in it, etc, beyond just the flyweighted aspect of symbol objects 2014-04-24T04:23:33Z White_Flame: each thread has a linear allocation buffer for "small" allocations 2014-04-24T04:23:43Z White_Flame: which does not need any locking (or even a subroutine call) to allocate from 2014-04-24T04:24:57Z White_Flame: the GC is still stop-the-world, afaik 2014-04-24T04:25:07Z |3b|: yeah 2014-04-24T04:25:23Z Bike: what parts of the clos are single-threaded, again? (not too related i'm just wondering now) 2014-04-24T04:27:28Z |3b|: not sure i'd call clos stuff 'single threaded' (at least stuff that doesn't need to call the compiler), i'd assume most of it is just small locks around modifications 2014-04-24T04:28:15Z Bike: i thought there were some odious locks around make-instance or something, i don't know 2014-04-24T04:28:25Z Bike: (i guess make-instance /is/ where the compiler comes in...) 2014-04-24T04:28:57Z |3b| was thinking more about modifying generic functions, or redefining classes 2014-04-24T04:29:42Z |3b|: make-instance is just calling a GF, no reason for any locks, aside from maybe a read lock to make sure that GF isn't being modified right then 2014-04-24T04:29:51Z Bike: doesn't it have ctors? 2014-04-24T04:30:13Z |3b| doesn't know much of anything about PCL or sbcl's version of it though, so could be completely wrong 2014-04-24T04:30:18Z |3b|: "ctor"? 2014-04-24T04:30:41Z Bike: precompiled constructors? i don't know much about it either, but i think it's one of the places pcl calls the compiler 2014-04-24T04:31:04Z |3b|: hmm, could be 2014-04-24T04:31:27Z |3b|: either way it should be fairly rare, like first time on a particular class or something 2014-04-24T04:42:25Z ooben quit (Quit: thanks, all!) 2014-04-24T05:04:04Z zRecursive{away}: clhs ~{} 2014-04-24T05:04:04Z specbot: Couldn't find anything for ~{}. 2014-04-24T05:04:08Z zRecursive{away}: clhs ~{ 2014-04-24T05:04:09Z specbot: http://www.lispworks.com/reference/HyperSpec/Body/22_cgd.htm 2014-04-24T05:04:39Z edgar-rft joined #sbcl 2014-04-24T05:07:39Z oleo quit (Quit: Leaving) 2014-04-24T05:12:05Z cmack``` joined #sbcl 2014-04-24T05:13:32Z slyrus joined #sbcl 2014-04-24T05:14:05Z cmack`` quit (Ping timeout: 264 seconds) 2014-04-24T05:16:49Z sdemarre joined #sbcl 2014-04-24T05:41:24Z ehaliewicz joined #sbcl 2014-04-24T05:45:26Z sdemarre quit (Ping timeout: 255 seconds) 2014-04-24T06:10:53Z drmeister joined #sbcl 2014-04-24T06:15:13Z drmeister quit (Ping timeout: 240 seconds) 2014-04-24T06:55:29Z hlavaty quit (Ping timeout: 264 seconds) 2014-04-24T07:27:15Z pranavrc joined #sbcl 2014-04-24T07:27:15Z pranavrc quit (Changing host) 2014-04-24T07:27:15Z pranavrc joined #sbcl 2014-04-24T07:41:37Z zRecursive{away} quit (Remote host closed the connection) 2014-04-24T07:59:31Z ASau quit (Ping timeout: 252 seconds) 2014-04-24T07:59:39Z drmeister joined #sbcl 2014-04-24T08:04:35Z drmeister quit (Ping timeout: 276 seconds) 2014-04-24T08:26:52Z jdz joined #sbcl 2014-04-24T08:49:48Z ehaliewicz quit (Remote host closed the connection) 2014-04-24T08:50:01Z ehaliewicz joined #sbcl 2014-04-24T09:05:12Z ehaliewicz quit (Remote host closed the connection) 2014-04-24T09:06:14Z jdz quit (Quit: Leaving...) 2014-04-24T09:24:55Z stassats joined #sbcl 2014-04-24T09:30:38Z kludge` quit (Ping timeout: 240 seconds) 2014-04-24T09:36:17Z kludge` joined #sbcl 2014-04-24T09:39:16Z cmack```` joined #sbcl 2014-04-24T09:41:06Z cmack``` quit (Ping timeout: 252 seconds) 2014-04-24T09:48:26Z drmeister joined #sbcl 2014-04-24T09:50:50Z nyef joined #sbcl 2014-04-24T09:51:08Z nyef: G'morning all. 2014-04-24T09:53:09Z drmeister quit (Ping timeout: 265 seconds) 2014-04-24T09:54:15Z Krystof: morning! 2014-04-24T10:33:05Z nyef: ... Oh. The reason that SIGSEGV doesn't get bounced to lisp cleanly is because the handler tends to run on the altstack, isn't it? 2014-04-24T10:51:18Z nyef: It's been about five years, is it time to do another massive round of changes to how we deal with interrupt handling? 2014-04-24T10:56:17Z stassats: straight to safepoints! 2014-04-24T10:59:53Z nyef: Oh, ouch. The ARM port forked before safepoints landed. 2014-04-24T11:02:13Z Krystof: getting to a repl without too many kludges is a reasonable point for a forward-port 2014-04-24T11:02:51Z Krystof: but I'm not totally in love with safepoints 2014-04-24T11:02:56Z nyef: Mmm. How many kludges do we have? 2014-04-24T11:07:18Z stassats: add-with-carry is broken, setting the carry flag incorrectly again 2014-04-24T11:08:05Z nyef: At least that was predictable. 2014-04-24T11:12:23Z nyef: Before we do a forward-port, let's get float printing, bignum-to-float, float raw slots, and possibly float sap functions working. 2014-04-24T11:13:04Z nyef: Oh, and float randoms. 2014-04-24T11:13:14Z stassats: i have frequent failures at do-core-fixups 2014-04-24T11:13:33Z nyef: Oh? 2014-04-24T11:13:49Z stassats: no idea, just see it in the backtrace 2014-04-24T11:14:02Z stassats: nondetermenistically 2014-04-24T11:14:09Z nyef: Lovely. 2014-04-24T11:14:53Z stassats: well, redefining a VOP fails reliably, just in different places 2014-04-24T11:15:50Z nyef: Is sb!vm::fixup-code-object defined as a function? 2014-04-24T11:16:23Z nyef: ... I bet it isn't. 2014-04-24T11:16:28Z stassats: unfbound 2014-04-24T11:16:37Z nyef: That'd do it. 2014-04-24T11:17:13Z nyef: Well, part of it, at least. It'd be a deterministic failure, really. 2014-04-24T11:17:39Z stassats: that was during LOAD, indeed 2014-04-24T11:19:37Z cmack```` quit (Ping timeout: 240 seconds) 2014-04-24T11:20:08Z nyef: So, my short-list is to flip the C input sense on add-w/carry, define fixup-code-object, and start clearing out the more egregious build-system damage. 2014-04-24T11:20:29Z stassats: i can deal with add-w/carry 2014-04-24T11:20:43Z stassats: unless you already know the solution 2014-04-24T11:21:18Z stassats: (trying to make a setup for testing assembly code without recompiling the whole thing) 2014-04-24T11:21:25Z nyef: Please do, then. And can you see if you can get float printing to work? It should just be a matter of removing reader conditionals. 2014-04-24T11:23:09Z stassats: is gas the only option? can't say i like its syntax 2014-04-24T11:25:34Z stassats: well, using vi and gas, fun 2014-04-24T11:27:36Z nyef: http://paste.lisp.org/display/142267 2014-04-24T11:28:06Z nyef: You might be able to redefine VOPs with that in place, for all I know. 2014-04-24T11:28:38Z attila_lendvai quit (Quit: Leaving.) 2014-04-24T11:30:52Z stassats: using nano is even worse 2014-04-24T11:31:19Z drmeister joined #sbcl 2014-04-24T11:32:44Z flip214: stassats: how about vim? or export the filesystem via NFS, and only compile on the ARM? 2014-04-24T11:32:53Z nyef: Tramp? 2014-04-24T11:33:23Z stassats: i have a castrated vim in form of vi, and the full vim will take up 28MB 2014-04-24T11:33:31Z stassats: and i created too small a partition 2014-04-24T11:33:42Z stassats: otherwise i would have just installed emacs 2014-04-24T11:33:52Z nyef: So... use emacs tramp? 2014-04-24T11:34:06Z stassats: yeah, the last option 2014-04-24T11:34:11Z flip214: well, resizing partitions should be possible on-the-fly nowadays ... 2014-04-24T11:34:18Z attila_lendvai joined #sbcl 2014-04-24T11:34:18Z attila_lendvai quit (Changing host) 2014-04-24T11:34:18Z attila_lendvai joined #sbcl 2014-04-24T11:34:32Z flip214: or you could rm -rf /usr/share/doc and get ~2GB or something like that ;) 2014-04-24T11:34:49Z stassats: 2GB? my whole partition is 512MB 2014-04-24T11:35:21Z attila_lendvai quit (Client Quit) 2014-04-24T11:36:36Z stassats: the installation was confusing, i didn't get whether i would just get 2GB or also some desktop GUI it was suggesting those 2GB would be good for 2014-04-24T11:37:52Z stassats: and i remembered what to type to tramp on the first try 2014-04-24T12:01:55Z stassats: nyef: successfully redefined add-with-carry with that fixup-code-object 2014-04-24T12:02:23Z nyef: Okay, I'll put a commit together for fixup-code-object then. 2014-04-24T12:11:22Z DGASAU quit (Remote host closed the connection) 2014-04-24T12:11:40Z stassats: so, add-w/carry needs cmp c 1 too 2014-04-24T12:11:58Z nyef: About what I figured. Any other changes, like switching the test sense on the output? 2014-04-24T12:12:23Z stassats: output is fine 2014-04-24T12:12:47Z stassats: and redefining vops at the repl is fun 2014-04-24T12:13:08Z stassats: but i still sometimes get mysterious failures 2014-04-24T12:13:49Z stassats: simple-eval-in-lexenv is near the top of the stack, then an unknown frame 2014-04-24T12:14:11Z nyef: Is this backtrace still from ldb? 2014-04-24T12:14:23Z stassats: yes 2014-04-24T12:14:44Z stassats: memory fault at #xFFFFFFF4 2014-04-24T12:15:55Z nyef: Go into interrupt.c, find the use of SA_ONSTACK, and arrange for it to always be set. Stuff will still break, but you might not see LDB so often. 2014-04-24T12:17:37Z stassats: isn't it always set? 2014-04-24T12:17:58Z nyef: For SIG_MEMORY_FAULT only. 2014-04-24T12:18:21Z nyef: Unfortunately, SIGTRAP thus leads to stomping on the stack. 2014-04-24T12:18:28Z stassats: ah, that's what you meant 2014-04-24T12:21:48Z stassats: strangely, i now get a %fixnum-to-digit-with-correct-sign overflowing the stack 2014-04-24T12:24:23Z DGASAU joined #sbcl 2014-04-24T12:24:39Z stassats: the add-w/carry fix may have nudged it into a different code path 2014-04-24T12:25:49Z nyef: Just pushed my branch again. 2014-04-24T12:25:57Z nyef: Does not include the add-w/carry fix yet. 2014-04-24T12:26:56Z stassats: ok, so, add-w/carry does break something 2014-04-24T12:30:22Z DGASAU quit (Ping timeout: 252 seconds) 2014-04-24T12:34:41Z stassats: since %fixnum-to-digit-with-correct-sign isn't a VOP yet, it forces bignum allocation on a full call, so, perhaps due to that 2014-04-24T12:34:52Z stassats: but why an endless recursion 2014-04-24T12:36:22Z DGASAU joined #sbcl 2014-04-24T12:41:36Z eudoxia joined #sbcl 2014-04-24T12:48:38Z stassats: implemented %fixnum-to-digit-with-correct-sign as a VOP, cold-init passes again 2014-04-24T12:49:20Z DGASAU quit (Ping timeout: 255 seconds) 2014-04-24T12:49:28Z stassats: http://paste.lisp.org/display/142267#1 2014-04-24T12:50:07Z stassats: (1+ 536870912) => 536870913 2014-04-24T12:50:27Z nyef: Okay, Thank you. 2014-04-24T12:50:31Z Krystof: progress :) 2014-04-24T12:51:13Z stassats: (fib 50) => 12586269025 instead of 6277101738309684039858724727853387073209631205623600462197 2014-04-24T12:51:42Z stassats: nyef: and that includes cmp c, 1 for add-w/carry 2014-04-24T12:52:02Z nyef: I figured I'd need both. 2014-04-24T12:55:00Z psilord quit (Quit: Leaving.) 2014-04-24T12:55:09Z DGASAU joined #sbcl 2014-04-24T13:00:37Z stassats: (defun s (n) (declare ((integer 1 100000) n)) (make-array (ceiling n 2))) is broken 2014-04-24T13:00:43Z stassats: (s 100) => #() 2014-04-24T13:02:45Z stassats: and i can rename packages now, and get TIME to break later than usual 2014-04-24T13:04:08Z nyef: I bet you can (setf sb!c::*compiler-trace-output* *standard-output*) to get trace results now. 2014-04-24T13:04:18Z nyef: If you weren't doing that already, that is. 2014-04-24T13:04:58Z stassats: Undefined function: ; TRACE 2014-04-24T13:05:01Z davazp joined #sbcl 2014-04-24T13:05:17Z stassats: must be in warm 2014-04-24T13:06:20Z stassats: (+ 1.5 30) => 31.5 2014-04-24T13:06:28Z stassats: shed the conditionals in print 2014-04-24T13:07:31Z nyef: All the conditionals in print, and just those? 2014-04-24T13:07:38Z nyef: What about bignum-to-float? 2014-04-24T13:09:17Z stassats: (float (1+ most-positive-fixnum )) => 5.368709e8 2014-04-24T13:09:23Z stassats: that's not right, is it? 2014-04-24T13:09:38Z nyef: what's m-p-f? 2014-04-24T13:09:46Z stassats: it is right, compared with the wrong sbcl 2014-04-24T13:10:24Z stassats: (float (1+ most-positive-fixnum) 1d0) => 5.36870912d8 2014-04-24T13:11:58Z nyef: So I can kill that conditional as well? 2014-04-24T13:12:44Z stassats: yep 2014-04-24T13:13:11Z stassats: (eql 1d0 1d0) => T 2014-04-24T13:13:17Z stassats: that's in %eql in pred.lisp 2014-04-24T13:13:23Z stassats: complexes and ratios still don't work 2014-04-24T13:14:04Z nyef: Hrm. I wonder why not? 2014-04-24T13:14:55Z stassats: recursion in known function definition 2014-04-24T13:15:03Z nyef: Ah, right. 2014-04-24T13:15:14Z stassats: numerator/realpart 2014-04-24T13:15:25Z nyef: There's something stupid about this, IIRC, but I forget what. 2014-04-24T13:16:33Z ltbarcly joined #sbcl 2014-04-24T13:23:51Z stassats: and TIME now fully works, provided the packages are renamed 2014-04-24T13:24:10Z drmeister quit (Remote host closed the connection) 2014-04-24T13:25:00Z nyef: Oh, right, %EQL gets inlined. 2014-04-24T13:29:11Z stassats: ./make-target-2.sh goes some length 2014-04-24T13:29:31Z davazp quit (Remote host closed the connection) 2014-04-24T13:29:40Z stassats: probably hits a GC 2014-04-24T13:29:50Z nyef: Or hits an internal-error trap. 2014-04-24T13:31:08Z stassats: it now seems surprisingly sturdy 2014-04-24T13:31:49Z nyef: Yeah, there's three big things on my list to fix, and a few small things, and we'll be ready for a forward-port. 2014-04-24T13:32:24Z nyef: And those big things are fixing internal-error traps, fixing the disassembler, and fixing interrupt handling. 2014-04-24T13:33:06Z stassats: i'll ponder on ceiling / make-array badness 2014-04-24T13:34:20Z stassats: and forward porting doesn't sound like fun 2014-04-24T13:35:01Z nyef: We'll probably have to discuss forward-porting strategy at some point, maybe this weekend. 2014-04-24T13:35:09Z LiamH joined #sbcl 2014-04-24T13:37:16Z nyef: Should we be cleaning up the lose-on-corruption interrupt context changes and putting them in the main system? 2014-04-24T13:40:43Z drmeister joined #sbcl 2014-04-24T13:52:23Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-24T13:59:45Z oleo joined #sbcl 2014-04-24T14:00:21Z pranavrc quit 2014-04-24T14:00:50Z drmeister quit (Read error: Connection reset by peer) 2014-04-24T14:01:14Z oleo quit (Killed (idoru (Spam is off topic on freenode.))) 2014-04-24T14:02:20Z oleo joined #sbcl 2014-04-24T14:05:05Z segv- joined #sbcl 2014-04-24T14:11:13Z erikvarga joined #sbcl 2014-04-24T14:21:22Z psilord joined #sbcl 2014-04-24T14:33:51Z charlie joined #sbcl 2014-04-24T14:33:59Z charlie is now known as cmack 2014-04-24T14:34:21Z drmeister joined #sbcl 2014-04-24T14:39:13Z LiamH quit (Quit: Leaving.) 2014-04-24T14:46:23Z erikvarga quit (Remote host closed the connection) 2014-04-24T14:47:16Z stassats: i can't manage to find a way to find an address of a function to pass it to gdb 2014-04-24T14:47:23Z stassats: (to get to a disassembler) 2014-04-24T14:48:45Z nyef: (+ 23 (sb!kernel:get-lisp-obj-address #'+)) 2014-04-24T14:49:12Z stassats: too late, i just figured that out 2014-04-24T14:49:31Z stassats: took a look at a trace of (funcall (the function x)) 2014-04-24T14:49:48Z nyef: Heh. 2014-04-24T14:50:15Z stassats: i just have a feeling that some instructions are misencoded 2014-04-24T14:51:02Z nyef: It's been known, but the worst cases should already have been dealt with. 2014-04-24T14:52:11Z drmeister quit (Remote host closed the connection) 2014-04-24T14:54:03Z stassats: MVN NARGS, 0 shows up as mvn r12, #-1073741824 ; 0xc0000000 2014-04-24T14:54:29Z nyef: Lovely. 2014-04-24T14:57:54Z stassats: it comes from an immediate move, but when i do (inst mvn nargs 0), it shows up alright 2014-04-24T14:58:18Z drmeister joined #sbcl 2014-04-24T14:58:51Z stassats: composite-immediate-instruction 2014-04-24T15:01:40Z nyef: ... Lovely. 2014-04-24T15:02:30Z stassats: -1 is the immediate, it correctly computes 0 for MVN, but something goes wrong anyway 2014-04-24T15:04:43Z stassats: the trace is showing MVN NARGS, 0 2014-04-24T15:07:45Z stassats: no, it computes something bogus 2014-04-24T15:08:03Z stassats: maybe the computation functions are just bonkers 2014-04-24T15:09:57Z stassats: right 2014-04-24T15:21:07Z stassats: (typep -1 '(unsigned-byte 32)) => T doesn't help 2014-04-24T15:23:19Z Krystof: it's sometimes amazing how much works with such brokenness 2014-04-24T15:24:45Z stassats: and somebody spelled lose as loose 2014-04-24T15:25:12Z nyef: ... That might or might not have been me. 2014-04-24T15:45:55Z attila_lendvai joined #sbcl 2014-04-24T15:45:56Z attila_lendvai quit (Changing host) 2014-04-24T15:45:56Z attila_lendvai joined #sbcl 2014-04-24T15:49:38Z stassats: ok, got (typep -1 '(unsigned-byte 32)) 2014-04-24T15:50:33Z stassats: nyef: http://paste.lisp.org/display/142267#2 2014-04-24T15:51:30Z stassats: wait, it's still broken 2014-04-24T15:51:34Z stassats: on negative bignums 2014-04-24T15:52:43Z drmeister quit (Quit: Leaving...) 2014-04-24T15:54:36Z drmeister joined #sbcl 2014-04-24T15:55:42Z stassats: i see what it was going for 2014-04-24T15:57:50Z stassats: nyef: correct version: http://paste.lisp.org/display/142267#3 2014-04-24T15:59:17Z nyef: I'll give it a try once I've finished with what I'm currently doing. 2014-04-24T16:01:08Z stassats: that fixes my original CEILING (actually ASH) problem 2014-04-24T16:01:27Z stassats: i can now get back to generating prime numbers 2014-04-24T16:02:15Z nyef: Oh, to be able to get through target-2, huh? 2014-04-24T16:02:24Z stassats: (extract-primes (sieve 100)) => #(2 3 5 7 11 13 17 19 23 29 31 37 41 43 47 53 59 61 67 71 73 79 83 89 97) 2014-04-24T16:02:39Z stassats: no, i was running random tests, factorial, fibonacci, number sieve 2014-04-24T16:02:53Z stassats: fibonacci and number sieve managed to find a bug each 2014-04-24T16:03:21Z hlavaty joined #sbcl 2014-04-24T16:05:51Z stassats: target-2 still fails compiling COMPUTE-LIMIT in pcl/cache.lisp 2014-04-24T16:09:44Z nyef: ... why is backtrace attempting to call sub-access-debug-var-slot on an immediate scn? 2014-04-24T16:10:24Z nyef: Oh well, what I have is an improvement, so I should commit it. 2014-04-24T16:11:06Z stassats: there was a change to the format of sc encoding, so if you copied a tramp from non-x86, you may have not updated it yet 2014-04-24T16:11:41Z slyrus quit (Ping timeout: 264 seconds) 2014-04-24T16:11:59Z stassats: no, the tree is before that 2014-04-24T16:16:32Z nyef: Oh, it's not from backtrace. 2014-04-24T16:17:03Z nyef: Okay, that's for me to keep digging into. 2014-04-24T16:17:31Z nyef: So, type-vop fixes? 2014-04-24T16:20:17Z stassats: i guess i need to set up a github repo 2014-04-24T16:20:24Z nyef: Heh. 2014-04-24T16:20:38Z nyef: I should probably move to github as well. 2014-04-24T16:20:50Z LiamH joined #sbcl 2014-04-24T16:21:14Z nyef: Is it just the change to that one fixnum test and the spelling correction? 2014-04-24T16:22:04Z stassats: yes, the last annotation 2014-04-24T16:23:06Z stassats: somebody thought that ANDS will both test for a fixnum and load the value intact into TEMP, which is not the case 2014-04-24T16:23:35Z drmeister quit (Remote host closed the connection) 2014-04-24T16:24:05Z eudoxia_ joined #sbcl 2014-04-24T16:24:15Z stassats: it changes to moving the value into TEMP and using TST, via %test-fixnum, because later TEMP also contains a bignum digit, which performs the same <0 comparison 2014-04-24T16:24:33Z nyef: Yeah, okay. That might have been my mistake. 2014-04-24T16:25:19Z stassats: the code has fiveop's name 2014-04-24T16:25:37Z nyef: Fair enough, then. 2014-04-24T16:25:44Z eudoxia quit (Ping timeout: 276 seconds) 2014-04-24T16:28:09Z drmeister joined #sbcl 2014-04-24T16:29:55Z eudoxia_ quit (Quit: leaving) 2014-04-24T16:30:03Z eudoxia joined #sbcl 2014-04-24T16:38:40Z stassats: i now want slime 2014-04-24T16:44:47Z nyef: Trying make-target-2. 2014-04-24T16:45:57Z nyef: Ooh. It survives a few GCs. 2014-04-24T16:46:11Z stassats: do you get past cache.lisp? 2014-04-24T16:46:33Z nyef: Don't know, I still have show on, so there's too much output to see. 2014-04-24T16:47:38Z nyef: Just saw something about wrapper.fasl. 2014-04-24T16:47:43Z nyef: Hrm. 2014-04-24T16:47:45Z nyef: It stopped. 2014-04-24T16:48:42Z nyef: ... and I don't know where. 2014-04-24T16:48:59Z nyef: I have a repl in the debugger, though. 2014-04-24T16:49:11Z stassats: cache is after wrapper 2014-04-24T16:49:31Z stassats: fails to me at compute-limit 2014-04-24T16:49:56Z nyef: It's at cache. 2014-04-24T16:50:02Z nyef: Found it in the backtrace. 2014-04-24T16:50:03Z stassats: compute limit has SQRT 2014-04-24T16:50:15Z nyef: Ah, lack of irrat functions? 2014-04-24T16:50:45Z stassats: right, when i tried enabled def-math-rtn, i got some TN conflict 2014-04-24T16:51:13Z nyef: Yup. The float and double-float alien stuff is hosed. 2014-04-24T16:51:55Z nyef: fiveop seems to have a hard time with the concept of single-floats and double-floats sharing a register set. 2014-04-24T16:52:31Z stassats: compiler probably tries to derive a type for (sqrt (sqrt x)) 2014-04-24T16:57:18Z nyef quit (Ping timeout: 240 seconds) 2014-04-24T17:09:26Z nyef joined #sbcl 2014-04-24T17:17:28Z tmh_ quit (Changing host) 2014-04-24T17:17:28Z tmh_ joined #sbcl 2014-04-24T17:19:00Z ltbarcly joined #sbcl 2014-04-24T17:31:26Z eudoxia quit (Quit: leaving) 2014-04-24T17:32:43Z nyef: Okay, new commits pushed. 2014-04-24T17:35:28Z tsuru joined #sbcl 2014-04-24T17:39:13Z attila_lendvai quit (Quit: Leaving.) 2014-04-24T17:39:33Z attila_lendvai joined #sbcl 2014-04-24T17:39:33Z attila_lendvai quit (Changing host) 2014-04-24T17:39:33Z attila_lendvai joined #sbcl 2014-04-24T17:53:52Z DGASAU quit (Remote host closed the connection) 2014-04-24T18:20:49Z nyef quit (Ping timeout: 240 seconds) 2014-04-24T18:27:12Z jdz joined #sbcl 2014-04-24T18:31:29Z ASau joined #sbcl 2014-04-24T18:33:11Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-24T18:34:48Z nyef joined #sbcl 2014-04-24T18:44:25Z ASau quit (Ping timeout: 240 seconds) 2014-04-24T18:56:09Z scymtym_ joined #sbcl 2014-04-24T19:07:06Z slyrus joined #sbcl 2014-04-24T19:07:32Z Blkt_ joined #sbcl 2014-04-24T19:07:45Z Munksgaard quit (Ping timeout: 246 seconds) 2014-04-24T19:07:46Z Blkt quit (Ping timeout: 246 seconds) 2014-04-24T19:09:49Z fe[nl]ix quit (Ping timeout: 246 seconds) 2014-04-24T19:11:27Z fe[nl]ix joined #sbcl 2014-04-24T19:12:44Z Munksgaard joined #sbcl 2014-04-24T19:23:36Z stassats: (sqrt 4) => 4.0 is a tad wrong 2014-04-24T19:24:01Z stassats: alien functions with two double-float arguments have a TN conflict 2014-04-24T19:24:06Z stassats: one argument is fine 2014-04-24T19:28:50Z nyef: Oh. In alien-type-method (double-float :arg-tn), the call to my-make-wired-tn, double the value of REGISTER. 2014-04-24T19:30:25Z erikvarga joined #sbcl 2014-04-24T19:30:36Z nyef: I really need to go over that code with the AAPCS documentation and several examples in hand. 2014-04-24T19:46:48Z edgar-rft quit (Quit: activity interrupted because no existence happened) 2014-04-24T20:04:49Z attila_lendvai quit (Quit: Leaving.) 2014-04-24T20:06:42Z nyef: I think I see what fiveop did here, and why. He's treating the double-reg space as though it's indexed as individual regs, rather than as paired single-regs, which explains at least part of the complexity. 2014-04-24T20:06:59Z ltbarcly joined #sbcl 2014-04-24T20:08:39Z fiveop joined #sbcl 2014-04-24T20:08:55Z fiveop: hi 2014-04-24T20:09:28Z fiveop: looks like you made a lot of progress on arm 2014-04-24T20:09:55Z Krystof: the most visible progress is always from "doesn't survive cold-init even remotely" to "woo, a repl" 2014-04-24T20:10:22Z stassats: the next one is "woo, a repl remotely" 2014-04-24T20:10:31Z stassats: i.e. slime 2014-04-24T20:10:34Z fiveop: :) 2014-04-24T20:11:05Z scymtym_ quit (Ping timeout: 264 seconds) 2014-04-24T20:13:29Z tmh_ quit (Ping timeout: 252 seconds) 2014-04-24T20:14:52Z fiveop: I hope a clean rebuild after pull will solve the oom problem, I had the last time. 2014-04-24T20:15:21Z stassats: i can run gc and room 2014-04-24T20:16:37Z fiveop: The last time I tried, there was no repl :) 2014-04-24T20:17:16Z tmh_ joined #sbcl 2014-04-24T20:17:26Z stassats: you missed all the fun 2014-04-24T20:19:06Z fiveop: Reading the git log, I probably did :/ 2014-04-24T20:19:23Z fiveop: Monday was fun, when I could 'participate' 2014-04-24T20:19:37Z yacks quit (Ping timeout: 240 seconds) 2014-04-24T20:19:51Z yacks joined #sbcl 2014-04-24T20:22:05Z stassats: (expt 2.0 3) => 2.0 is no good either 2014-04-24T20:22:49Z stassats: basically, all the libc functions return what's passed 2014-04-24T20:24:42Z fiveop: I wrote the stuff for floats in compiler/arm/c-call. So, your probably should have a look at that first ;) 2014-04-24T20:25:02Z fiveop: I still have the oom problem. 2014-04-24T20:25:24Z stassats: are you using the latest nyef branch? 2014-04-24T20:25:38Z stassats: and with everything synced? 2014-04-24T20:25:49Z fiveop: Yes 2014-04-24T20:26:11Z fiveop: When I first got it, nyefs branch already got farther than me. 2014-04-24T20:26:31Z stassats: ah, oom you mean, linux oom? 2014-04-24T20:26:34Z fiveop: yes 2014-04-24T20:26:42Z stassats: download more RAM 2014-04-24T20:26:43Z fiveop: as in oom killer is invoked :) 2014-04-24T20:27:05Z fiveop: The ram is nowhere near full, if I read the kernel logs correctly. I have no idea what happens there. 2014-04-24T20:27:46Z stassats: overcommit? 2014-04-24T20:28:47Z fiveop: yes 2014-04-24T20:29:06Z fiveop: I was thinking about trying something I found googling for it somewhere :) (setting some overcommit mode to 2 or something) 2014-04-24T20:29:27Z fiveop: (have to find it again, before being able to talk about it coherently) 2014-04-24T20:29:35Z stassats: /proc/sys/vm/overcommit_memory ? 2014-04-24T20:30:23Z stassats: you need to set it to 1, not 2 2014-04-24T20:33:00Z stassats: (sb!disassem::disassemble-code-component #'list) 2014-04-24T20:33:00Z stassats: 2014-04-24T20:33:09Z stassats: ; 34C: 1DE04FE2 BYTE #X1D, #XE0, #X4F, #XE2 2014-04-24T20:33:15Z stassats: almost working 2014-04-24T20:40:13Z tmh_ quit (Changing host) 2014-04-24T20:40:13Z tmh_ joined #sbcl 2014-04-24T20:44:45Z fiveop: mmap: Cannot allocate memory 2014-04-24T20:44:45Z fiveop: ensure_space: failed to validate 402649088 bytes at 0x4f000000 2014-04-24T20:44:45Z fiveop: 2014-04-24T20:44:50Z fiveop: after the overcommit change 2014-04-24T20:45:00Z stassats: to 1? 2014-04-24T20:45:08Z fiveop: that would be 2 2014-04-24T20:45:13Z stassats: you need 1 2014-04-24T20:45:46Z fiveop: let's see how far I get 2014-04-24T20:46:32Z fiveop: still, sbcl really slows down in "/about to do *!DEFERRED-USE-PACKAGES*" after the first couple of "/NIL" 2014-04-24T20:46:36Z fiveop: and gets killed ; 2014-04-24T20:46:58Z stassats: how much memory does it use at that point? 2014-04-24T20:49:13Z fiveop: It goes quickly to about 20% (of ~480MB) and then gets very slow (and various kernel threads go nuts in cpu usage) 2014-04-24T20:49:43Z fiveop: It gets killed at 24% 2014-04-24T20:49:48Z stassats: are you using the same setup as nyef? 2014-04-24T20:50:21Z fiveop: Well he can only have the weaker raspberry pi, if he hasn't the same. 2014-04-24T20:50:31Z fiveop: I think we run different flavours of linux, though. 2014-04-24T20:50:32Z stassats: and OS? 2014-04-24T20:50:47Z fiveop: I have arch running, I think he said something about debian. 2014-04-24T20:50:52Z stassats: i have twice as much memory and no problems 2014-04-24T20:51:09Z fiveop: I looked at ulimit -a as well, but there are no limits 2014-04-24T20:52:22Z fiveop: Are there any other ways, that one can limit the amount of memory a process may use? 2014-04-24T20:53:20Z stassats: doubles are returned in r0 and r1, are they not? 2014-04-24T20:53:41Z LiamH quit (Quit: Leaving.) 2014-04-24T20:53:43Z erikvarga left #sbcl 2014-04-24T20:59:25Z fiveop: d0? 2014-04-24T20:59:38Z stassats: no? 2014-04-24T20:59:39Z fiveop: are you talking about the c calling convention? 2014-04-24T20:59:52Z stassats: i am 2014-04-24T21:00:49Z stassats: and the argument are passed on the stack, so, the current c-call doesn't do that 2014-04-24T21:01:33Z fiveop: where do you get that from? 2014-04-24T21:02:16Z stassats: something named "Procedure Call Standard for the the ARM Architecture" 2014-04-24T21:02:17Z fiveop: http://infocenter.arm.com/help/topic/com.arm.doc.ihi0042e/IHI0042E_aapcs.pdf as well as gcc output tell me they are passed/returned in sN/dN registers 2014-04-24T21:02:21Z stassats: plus gcc disassembly 2014-04-24T21:05:49Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-24T21:06:22Z fiveop: https://gist.github.com/anonymous/92009c40e3c4e2ae8598 2014-04-24T21:14:01Z fiveop: added a swap file 2014-04-24T21:14:09Z fiveop: now I get further :) 2014-04-24T21:15:03Z ltbarcly joined #sbcl 2014-04-24T21:17:43Z fiveop: (Though I need to get a swap device, that is not an old SD card, unless I want to shred it) 2014-04-24T21:18:27Z tsuru quit (Remote host closed the connection) 2014-04-24T21:18:35Z stassats: well, your float c-call for sure doesn't work with softfp 2014-04-24T21:18:42Z tsuru joined #sbcl 2014-04-24T21:19:21Z ltbarcly quit (Client Quit) 2014-04-24T21:21:38Z ltbarcly joined #sbcl 2014-04-24T21:27:05Z tsuru quit (Ping timeout: 252 seconds) 2014-04-24T21:28:17Z fiveop: yes it doesn't 2014-04-24T21:32:04Z Bike_ joined #sbcl 2014-04-24T21:33:23Z Bike_ is now known as Bicyclidine 2014-04-24T21:39:20Z fiveop: "internal error #26 (An attempt was made to use an undefined SYMBOL-VALUE.)" 2014-04-24T21:39:21Z fiveop: progress 2014-04-24T21:41:17Z fiveop: gn8 2014-04-24T21:41:20Z fiveop quit (Quit: bye) 2014-04-24T21:42:32Z Hydan joined #sbcl 2014-04-24T21:44:23Z nyef: stassats: Your system has a VFP, but uses soft-float calling conventions? 2014-04-24T21:44:51Z stassats: softfp 2014-04-24T21:45:11Z nyef: If you're getting this far, you have a VFP of one sort or another. 2014-04-24T21:45:16Z nyef: Or a good emulator. 2014-04-24T21:45:30Z stassats: no, i have VFP3, but sotftp is the name of the convention 2014-04-24T21:45:35Z nyef: Ah, okay. 2014-04-24T21:45:45Z stassats: floats are passed in registers/stack, but computed on VFP 2014-04-24T21:45:56Z nyef: Yeah, that sounds about right. 2014-04-24T21:46:13Z nyef: I knew, when I first wrote the c-call stuff, that this was going to be an issue. 2014-04-24T21:46:24Z nyef: I don't know which convention my system uses, actually. 2014-04-24T21:46:37Z stassats: it doesn't look like it's going to be hard to add support 2014-04-24T21:46:43Z nyef: It's not. 2014-04-24T21:47:04Z nyef: It's constrained to the part of c-call north of the VOPs. 2014-04-24T21:47:34Z stassats: can we make different conventions for different shared libraries? 2014-04-24T21:47:55Z nyef: Nothing comes to mind, I'm afraid. 2014-04-24T21:48:15Z nyef: There's also the question of auto-detecting the correct convention at config time. 2014-04-24T21:48:56Z nyef: Oops. Accidentally kicked off a target-2 run when I was trying for the cold repl. 2014-04-24T21:53:17Z nyef: Infinite loop, last filename seen was pcl/boot.lisp. 2014-04-24T21:54:47Z Bicyclidine quit (Read error: Connection reset by peer) 2014-04-24T21:56:52Z nyef: (expt 2.0 3) => 8.0 2014-04-24T21:57:23Z nyef: So, I have hard-float or whatever it's called. 2014-04-24T21:58:54Z stassats: i can already pass the fourth double argument to C 2014-04-24T21:59:18Z nyef: Gotten a warm REPL yet? 2014-04-24T21:59:25Z stassats: nope 2014-04-24T21:59:48Z nyef: How far'd you get? 2014-04-24T22:00:38Z Bike_ joined #sbcl 2014-04-24T22:01:36Z stassats: cpl 2014-04-24T22:01:41Z Bike_ is now known as Bicyclidine 2014-04-24T22:01:52Z nyef: Ooh. That's further than I got. 2014-04-24T22:04:07Z stassats: or even defcombin 2014-04-24T22:04:22Z stassats: it then just hangs 2014-04-24T22:05:19Z nyef: So, same thing I saw at boot? 2014-04-24T22:05:52Z psilord quit (Quit: Leaving.) 2014-04-24T22:06:11Z stassats: apparently 2014-04-24T22:06:34Z stassats: gc? 2014-04-24T22:06:46Z nyef: The need to GC. 2014-04-24T22:06:48Z nyef: Next time it gets there, kick it with a -11 and take a backtrace. 2014-04-24T22:07:12Z nyef: Oh, and first check, I bet the CPU load will be high at that point. 2014-04-24T22:07:13Z oleo quit (Ping timeout: 240 seconds) 2014-04-24T22:07:22Z stassats: it was 2014-04-24T22:07:33Z nyef: Backtrace said receive-pending-interrupt. 2014-04-24T22:07:35Z nyef: Didn't it? 2014-04-24T22:08:10Z stassats: precom2 now 2014-04-24T22:08:30Z stassats: find-rotated-loop-head 2014-04-24T22:08:56Z nyef: Starting a target-2 run. 2014-04-24T22:09:11Z stassats: it goes further each time for me 2014-04-24T22:09:35Z nyef: That seems a little odd... unless it can re-use the old fasls? 2014-04-24T22:09:41Z stassats: it doesn't 2014-04-24T22:09:50Z stassats: but i try to can disable it recompiling already compiled files 2014-04-24T22:09:54Z nyef: Or the mere existence of the old fasls can affect the memory usage pattern or something? 2014-04-24T22:09:56Z nyef: Ahh. 2014-04-24T22:10:41Z nyef: I'm doing a clean build each time (killing the fasls), so... 2014-04-24T22:11:25Z stassats: precom2 now again, didn't get further 2014-04-24T22:11:30Z stassats: ok, let's try reusing fasls 2014-04-24T22:12:13Z stassats: and at find-rotated-loop-head again 2014-04-24T22:12:20Z nyef: Eh, I'm hoping to get through the entire thing in one shot. 2014-04-24T22:12:34Z stassats: "hoping" 2014-04-24T22:12:56Z nyef: Well, I have a theory as to why it locked up... 2014-04-24T22:15:53Z stassats: got past precomp2 that way 2014-04-24T22:16:13Z stassats: does slad work? 2014-04-24T22:16:25Z nyef: I have no idea? 2014-04-24T22:16:25Z drmeister quit (Read error: Connection reset by peer) 2014-04-24T22:16:48Z drmeister joined #sbcl 2014-04-24T22:16:53Z stassats: idling in step 2014-04-24T22:17:00Z stassats: and by idling i mean churning 2014-04-24T22:17:30Z stassats: ash-left-mod32? 2014-04-24T22:17:31Z Hydan is now known as Hydan` 2014-04-24T22:18:02Z nyef: That seems a little odd, unless it's in a largeish loop or something. 2014-04-24T22:18:36Z stassats: i encountered it today already, seemed working, perhaps bignum consing caused a gc? 2014-04-24T22:19:29Z nyef: Hrm. I'm getting a certain amount of cannot-encode-immediate-operand. 2014-04-24T22:20:28Z oleo joined #sbcl 2014-04-24T22:20:45Z stassats: load and dump phase 2014-04-24T22:20:52Z stassats: now my question will be answere 2014-04-24T22:20:53Z stassats: d 2014-04-24T22:20:59Z nyef: Congratulations. 2014-04-24T22:21:39Z nyef: Just saw load-as-fasl for generic-functions. 2014-04-24T22:23:22Z stassats: well, l-a-d phase is hanging right off the bat 2014-04-24T22:25:07Z stassats: [doing purification: roots handlers stack bindings static cleanup done] Memory fault at 7efff000 2014-04-24T22:25:52Z nyef: Lovely. 2014-04-24T22:26:11Z nyef: My build is still going. 2014-04-24T22:26:18Z stassats: but i don't have to! 2014-04-24T22:26:39Z stassats: eh, loading still sometimes hangs too 2014-04-24T22:27:59Z nyef: By this point, I'm fairly sure that I've fixed the random hangs while building PCL. 2014-04-24T22:33:18Z segv- quit (Ping timeout: 240 seconds) 2014-04-24T22:35:04Z nyef: Corruption warning while compiling what looks to be run-program. 2014-04-24T22:35:44Z nyef: ... Some sort of issue with a buffer going missing or a buffer pointer getting corrupted? 2014-04-24T22:36:01Z nyef: I'll think about it more after dinner. 2014-04-24T22:36:40Z stassats: added (setq *print ..) to warm.lisp from make-target.lisp and using (load "warm")instead of --load, now i get a warm REPL 2014-04-24T22:37:14Z stassats: (defclass foo () ()) # * (make-instance 'foo) # 2014-04-24T22:38:10Z stassats: (sb-ext:run-program "uname" '() :search t :output *standard-output*) => Linux Memory fault at 3c 2014-04-24T22:40:01Z drmeister quit (Ping timeout: 240 seconds) 2014-04-24T22:42:42Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-24T22:43:38Z stassats: ok, enough for today, so, slime by tomorrow? 2014-04-24T22:46:33Z cmack: "Lovely"? ;D 2014-04-24T22:51:00Z drmeister joined #sbcl 2014-04-24T22:57:49Z jdz quit (Remote host closed the connection) 2014-04-24T22:58:58Z tmh_ quit (Ping timeout: 252 seconds) 2014-04-24T22:59:05Z drmeister quit (Remote host closed the connection) 2014-04-24T23:00:56Z stassats quit (Ping timeout: 276 seconds) 2014-04-24T23:01:15Z leoc quit (Read error: Operation timed out) 2014-04-24T23:02:13Z leoc joined #sbcl 2014-04-24T23:04:02Z ltbarcly joined #sbcl 2014-04-24T23:06:31Z tmh_ joined #sbcl 2014-04-24T23:16:34Z drmeister joined #sbcl 2014-04-24T23:17:07Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-24T23:29:31Z drmeister quit (Read error: Connection reset by peer) 2014-04-24T23:29:54Z |3b| quit (Ping timeout: 240 seconds) 2014-04-24T23:35:14Z psilord joined #sbcl 2014-04-24T23:36:31Z ltbarcly joined #sbcl 2014-04-24T23:39:17Z cmack quit (Ping timeout: 276 seconds) 2014-04-24T23:43:29Z Bicyclidine quit (Ping timeout: 252 seconds) 2014-04-24T23:50:06Z zRecursive joined #sbcl 2014-04-24T23:54:53Z slyrus: wow. looks like substantial progress is being made on the ARM port. 2014-04-25T00:01:31Z |3b| joined #sbcl 2014-04-25T00:06:27Z nyef: It's not QUITE surviving enough to generate a warm core, but it's close. 2014-04-25T00:13:34Z eudoxia joined #sbcl 2014-04-25T00:13:59Z slyrus: impressive! 2014-04-25T00:14:14Z slyrus: looking forward to trying it on the odroid u3. 2014-04-25T00:14:50Z nyef: Do you want to join in on the fun now? 2014-04-25T00:14:57Z slyrus: no, it's at home. I'm not. 2014-04-25T00:15:01Z nyef: Heh. 2014-04-25T00:15:07Z nyef: Tomorrow, then? (-: 2014-04-25T00:15:28Z slyrus: possibly 2014-04-25T00:23:52Z drmeister joined #sbcl 2014-04-25T00:27:19Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-25T00:28:37Z Hydan` quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-25T00:32:40Z drmeister quit (Read error: Connection reset by peer) 2014-04-25T00:33:10Z drmeister joined #sbcl 2014-04-25T00:34:27Z drmeiste_ joined #sbcl 2014-04-25T00:35:23Z LiamH joined #sbcl 2014-04-25T00:38:25Z drmeister quit (Ping timeout: 276 seconds) 2014-04-25T00:47:32Z slyrus quit (Ping timeout: 255 seconds) 2014-04-25T00:48:09Z Munksgaard quit (Ping timeout: 246 seconds) 2014-04-25T01:02:28Z zRecursive: (random ...) always returns same value if no (setf *random-state* (make-random-state t)) 2014-04-25T01:03:41Z |3b|: do you mean same sequence? 2014-04-25T01:04:44Z zRecursive: IIRC, #'random will return different value even if not changing the default *random-state* 2014-04-25T01:04:58Z zRecursive: |3b|: (random 16) => 12 here 2014-04-25T01:05:35Z |3b|: are you saying (loop repeat 4 collect (random 16)) = (12 12 12 12)? 2014-04-25T01:05:49Z zRecursive: yeah 2014-04-25T01:06:30Z |3b|: odd 2014-04-25T01:06:34Z zRecursive: it is normal after changing *random-state* explicitly 2014-04-25T01:07:13Z |3b|: does *random-state* look reasonable when it is not working correctly? 2014-04-25T01:07:16Z zRecursive: it is the newest sbcl from git repo 2014-04-25T01:08:16Z zRecursive: After (setf *random-state* (make-random-state t)), it will work as before 2014-04-25T01:09:47Z |3b|: before that, is *random-state* a random-state struct, and (sb-kernel::random-state-state *random-state*) is a vector of varying ints? 2014-04-25T01:11:11Z zRecursive: sorry, (loop repeat 4 collect (random 16)) will return same sequence as (12 6 14 9) 2014-04-25T01:11:19Z |3b|: ah, that is expected 2014-04-25T01:11:25Z zRecursive: always (12 6 14 9) 2014-04-25T01:11:44Z |3b|: *random-state* is saved in the image, so whenever load it, it is the same value 2014-04-25T01:12:46Z zRecursive: so in our application, we MUST change *random-state* explicitly 2014-04-25T01:13:07Z |3b|: if you want a different sequence every time, that would probably be a good idea 2014-04-25T01:13:24Z Munksgaard joined #sbcl 2014-04-25T01:13:35Z zRecursive: ok 2014-04-25T01:14:17Z |3b| would expect other CL implementations to behave similarly 2014-04-25T01:16:26Z zRecursive: At least CCL does 2014-04-25T01:16:48Z yacks quit (Ping timeout: 265 seconds) 2014-04-25T01:18:36Z |3b|: though initial value is "implementation dependent", so spec doesn't require either behavior 2014-04-25T01:22:49Z ltbarcly joined #sbcl 2014-04-25T01:48:07Z nyef quit (Quit: G'night all.) 2014-04-25T01:49:21Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-25T02:11:59Z eudoxia quit (Quit: leaving) 2014-04-25T02:26:41Z ivan`` quit (Ping timeout: 264 seconds) 2014-04-25T02:29:14Z nicdev quit (Remote host closed the connection) 2014-04-25T02:30:38Z ivan`` joined #sbcl 2014-04-25T02:31:01Z nicdev joined #sbcl 2014-04-25T02:38:56Z scymtym_ joined #sbcl 2014-04-25T02:39:17Z christoph_debian quit (Ping timeout: 264 seconds) 2014-04-25T02:53:37Z christoph_debian joined #sbcl 2014-04-25T03:18:38Z ltbarcly joined #sbcl 2014-04-25T03:40:44Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-25T03:50:20Z ltbarcly joined #sbcl 2014-04-25T04:01:14Z yacks joined #sbcl 2014-04-25T04:05:08Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-25T04:30:24Z joshee joined #sbcl 2014-04-25T04:31:37Z joshee quit (Client Quit) 2014-04-25T04:34:44Z ehaliewicz joined #sbcl 2014-04-25T04:49:58Z scymtym_ quit (Ping timeout: 276 seconds) 2014-04-25T04:51:36Z attila_lendvai joined #sbcl 2014-04-25T04:51:36Z attila_lendvai quit (Changing host) 2014-04-25T04:51:36Z attila_lendvai joined #sbcl 2014-04-25T05:06:04Z oleo quit (Quit: Leaving) 2014-04-25T05:16:16Z LiamH quit (Ping timeout: 252 seconds) 2014-04-25T05:27:17Z pranavrc joined #sbcl 2014-04-25T05:27:17Z pranavrc quit (Changing host) 2014-04-25T05:27:17Z pranavrc joined #sbcl 2014-04-25T05:33:04Z attila_lendvai quit (Quit: Leaving.) 2014-04-25T07:27:31Z White_Flame quit (Ping timeout: 252 seconds) 2014-04-25T07:48:08Z Hydan joined #sbcl 2014-04-25T07:48:14Z White_Flame joined #sbcl 2014-04-25T07:53:49Z Hydan is now known as Hydan` 2014-04-25T07:57:12Z Hydan` is now known as Hydan 2014-04-25T08:09:02Z zRecursive quit (Remote host closed the connection) 2014-04-25T08:21:26Z hlavaty left #sbcl 2014-04-25T08:29:18Z stassats joined #sbcl 2014-04-25T08:29:20Z Hydan is now known as Hydan` 2014-04-25T08:32:13Z Hydan` is now known as Hydan 2014-04-25T08:32:28Z Hydan quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-25T08:39:12Z jdz joined #sbcl 2014-04-25T08:39:24Z DGASAU joined #sbcl 2014-04-25T09:00:33Z segv- joined #sbcl 2014-04-25T09:33:22Z kludge` quit (Ping timeout: 276 seconds) 2014-04-25T09:36:35Z kludge` joined #sbcl 2014-04-25T09:48:38Z ehaliewicz quit (Ping timeout: 240 seconds) 2014-04-25T10:11:53Z jdz quit (Quit: Leaving...) 2014-04-25T10:28:17Z stassats quit (Ping timeout: 252 seconds) 2014-04-25T10:54:17Z edgar-rft joined #sbcl 2014-04-25T12:26:29Z eudoxia joined #sbcl 2014-04-25T12:32:51Z cmack joined #sbcl 2014-04-25T12:42:04Z oleo joined #sbcl 2014-04-25T12:53:27Z psilord quit (Quit: Leaving.) 2014-04-25T13:01:17Z momo-reina joined #sbcl 2014-04-25T13:09:46Z pranavrc quit 2014-04-25T13:17:52Z momo-reina quit (Remote host closed the connection) 2014-04-25T13:45:58Z ehaliewicz joined #sbcl 2014-04-25T13:50:19Z drmeiste_ is now known as drmeister 2014-04-25T13:50:38Z ehaliewicz quit (Ping timeout: 240 seconds) 2014-04-25T13:53:57Z ams quit (Excess Flood) 2014-04-25T13:54:30Z ams joined #sbcl 2014-04-25T13:56:37Z nyef joined #sbcl 2014-04-25T13:56:50Z nyef: G'morning all. 2014-04-25T13:56:56Z drmeister quit (Remote host closed the connection) 2014-04-25T14:02:59Z drmeister joined #sbcl 2014-04-25T14:06:20Z drmeister quit (Read error: Connection reset by peer) 2014-04-25T14:09:30Z drmeister joined #sbcl 2014-04-25T14:21:52Z oleo: morning! 2014-04-25T14:24:19Z ams: hi 2014-04-25T14:28:13Z pranavrc joined #sbcl 2014-04-25T14:28:14Z pranavrc quit (Changing host) 2014-04-25T14:28:14Z pranavrc joined #sbcl 2014-04-25T14:35:58Z cmack: nyef: I thought I would attempt building on my odroid u3 for fun last night... http://paste.lisp.org/display/142290 is what happens at make-target-2. is that the current expectation or have I failed somewhere in my build process? or is it potentially raspberry pi v. u3 ARM differences? 2014-04-25T14:36:33Z nyef: That's a failed build process. 2014-04-25T14:37:48Z nyef: At a guess, you skipped a step or re-ran a step or did some steps out of order. 2014-04-25T14:39:37Z cmack: :D I'll start over and come back when I figure it out... I'm using the make.sh comments as my guide. I assume they are canonical? 2014-04-25T14:40:25Z nyef: They're for old-style xc-host specification, it looks like. 2014-04-25T14:41:46Z cmack: is "new-style" documented? 2014-04-25T14:42:26Z nyef: You need to run make-config.sh on the target, with an xc-host spec suited to the host, copy everything over to the host, run make-host-1, copy everything back to the target, run make-target-1, copy everything back to the host, run make-host-2, copy everything back to the target, and then it should break differently. 2014-04-25T14:42:36Z nyef: Looks like the new-style options are documented in make-config.sh. 2014-04-25T14:42:49Z nyef: Try sh ./make-config.sh --help 2014-04-25T14:44:13Z cmack: nyef: ok thanks. I will give that a go 2014-04-25T14:48:17Z nyef: At this point, the LATEST that a full build will break is attempting to save the final core file. And that's if you're lucky, persistent, and willing to hack on the build order. 2014-04-25T14:48:51Z nyef: You should get a good way through compiling PCL, though. I think my last run got to RUN-PROGRAM before it died. 2014-04-25T14:55:34Z drmeister quit (Read error: Connection reset by peer) 2014-04-25T14:59:58Z eudoxia quit (Quit: leaving) 2014-04-25T15:06:53Z drmeister joined #sbcl 2014-04-25T15:06:55Z drmeister quit (Remote host closed the connection) 2014-04-25T15:07:29Z drmeister joined #sbcl 2014-04-25T15:21:47Z cmack: well it did break differently but I'm still unsure since I don't get any compiler messages http://paste.lisp.org/display/142290#1 2014-04-25T15:22:45Z nyef: Well, that's progress of a sort. 2014-04-25T15:22:49Z stassats joined #sbcl 2014-04-25T15:23:09Z nyef: Seems like it's failing in cold-init. 2014-04-25T15:24:58Z nyef: What does "uname -m" say on your system? 2014-04-25T15:25:38Z cmack: target? armv7l 2014-04-25T15:26:17Z nyef: Hrm. 2014-04-25T15:28:18Z stassats: sb-show time 2014-04-25T15:28:21Z nyef: "cat /proc/cpuinfo | grep Features"? 2014-04-25T15:28:51Z cmack: Features : swp half thumb fastmult vfp edsp thumbee neon vfpv3 tls 2014-04-25T15:29:29Z nyef: Hrm. Yeah, stassats is right, please build --with-sb-show. 2014-04-25T15:30:42Z cmack: just on make-target-2 ? or with the --xc-host to make-config at step 0? 2014-04-25T15:31:40Z stassats: back to step 0 2014-04-25T15:32:17Z nyef: And instead of make-target-2, the magic incantation for now is "src/runtime/sbcl --core output/cold-sbcl.core --no-sysinit --no-userinit --lose-on-corruption". 2014-04-25T15:33:50Z nyef: stassats: You left off at failing to save a core, didn't you? 2014-04-25T15:33:57Z stassats: right 2014-04-25T15:34:06Z nyef: Did you try the :purify nil option to SLAD? 2014-04-25T15:34:20Z stassats: re fp ABI, i think it would be possible to select different ABI on per-function basis 2014-04-25T15:34:41Z nyef: Why bother? It's not like gcc can do so. 2014-04-25T15:34:52Z stassats: it can 2014-04-25T15:35:04Z nyef: Hunh. Okay, I sit corrected. 2014-04-25T15:36:06Z Munksgaard quit (*.net *.split) 2014-04-25T15:37:28Z stassats: __attribute__((pcs("aapcs"))) vs __attribute__((pcs("aapcs-vfp"))) 2014-04-25T15:37:47Z nyef: Is this the same trick that we use for switching calling conventions between stdcall and cdecl on windows? 2014-04-25T15:38:10Z stassats: i haven't studied that yet 2014-04-25T15:39:22Z stassats: and i tried :purify nil at the bare repl, and it failed about something 2014-04-25T15:39:29Z nyef: Hrm. 2014-04-25T15:39:55Z nyef: Oh, right, so does sbcl on x86-64. 2014-04-25T15:40:10Z nyef: Don't know why, but it does. 2014-04-25T15:40:30Z stassats: but it failed differently with :purify t from the make-target 2014-04-25T15:40:40Z stassats: i'll try with different make-target-2 2014-04-25T15:46:09Z Munksgaard joined #sbcl 2014-04-25T15:47:25Z stassats: well 2014-04-25T15:47:53Z stassats: ./src/runtime/sbcl --core output/sbcl.core Bus error 2014-04-25T15:48:09Z nyef: Hrm. 2014-04-25T15:48:36Z nyef: And no real way to know where that error comes from? 2014-04-25T15:48:54Z stassats: gdb? 2014-04-25T15:49:08Z nyef: I think that's about our only option right now, yes. /-: 2014-04-25T15:49:54Z Hydan joined #sbcl 2014-04-25T15:50:02Z stassats: safe_call_into_lisp 2014-04-25T15:50:15Z stassats: unaligned stack? 2014-04-25T15:50:22Z Hydan is now known as Hydan` 2014-04-25T15:50:27Z Hydan` is now known as Hydan 2014-04-25T15:53:51Z stassats: and safe_call_into_lisp calls something, but i can't find it in the map 2014-04-25T15:53:58Z stassats: i supposes it's #'restart-lisp 2014-04-25T15:54:46Z nyef: Should be calling call_into_lisp in arm-assem.S, which does all of the swizzly bits to call whatever function, which should be #'restart-lisp. 2014-04-25T15:55:11Z stassats: restart-lisp is a local function, its start may not be properly aligned 2014-04-25T15:55:24Z stassats: because it clearly works with cold-init 2014-04-25T15:55:39Z nyef: Ahh... closure? 2014-04-25T15:55:45Z stassats: yeah 2014-04-25T15:56:28Z nyef: But we're calling through something with a FUN-POINTER-LOWTAG, right? 2014-04-25T15:56:40Z stassats: i'm trying a global function now 2014-04-25T15:58:05Z nyef: Yeah, it's the fact that it's a closure. 2014-04-25T15:59:56Z nyef: We need an ldr reg_CODE, [reg_LEXENV+CLOSURE_FUN_OFFSET] in front of the "And, finally, call into Lisp!", and the reg_LEXENV in the following instruction should be reg_CODE. 2014-04-25T16:02:09Z stassats: and if it's not a closure? 2014-04-25T16:02:24Z nyef: Ugh. I'm getting random failures in COPY-UB8-TO-SYSTEM-AREA, called from BUFFER-OUTPUT under ANSI-STREAM-WRITE-SEQUENCE. 2014-04-25T16:02:52Z nyef: If it's not a closure, the CLOSURE_FUN_OFFSET is SIMPLE_FUN_SELF_OFFSET, and just picks the same function pointer back up again. 2014-04-25T16:06:24Z pranavrc quit (Remote host closed the connection) 2014-04-25T16:07:03Z stassats: i am now too 2014-04-25T16:07:14Z stassats: didn't happen before i pulled today's changes 2014-04-25T16:07:31Z nyef: Okay, so what changed? 2014-04-25T16:08:09Z stassats: breakpoint stuff? 2014-04-25T16:08:23Z nyef: Really shouldn't be that. 2014-04-25T16:08:27Z nyef: It's kindof fundamental. 2014-04-25T16:09:08Z nyef: I'm seeing bignum fixes, interrupt-handling fixes, debugger fixes... 2014-04-25T16:09:21Z stassats: i had bignum fixes already 2014-04-25T16:09:23Z nyef: A bunch of float things. 2014-04-25T16:11:25Z ltbarcly joined #sbcl 2014-04-25T16:12:24Z nyef: This could easily have been masked by the receive-pending-interrupt damage. 2014-04-25T16:14:22Z stassats: on the other hand, and on the other failure COPY-UB8-TO-SYSTEM-AREA => SHIFT-TOWARDS-END 2014-04-25T16:14:30Z stassats: i've seen SHIFT-TOWARDS-END yesterday 2014-04-25T16:16:48Z nyef: So, three major issues right now, one of which we have a possible fix for: PURIFY doesn't work, cores with a closure as an initial function don't work, and this stream buffer thing. 2014-04-25T16:17:03Z nyef: Oh, and four, recursive interrupts are bad. 2014-04-25T16:19:34Z erikvarga joined #sbcl 2014-04-25T16:20:50Z stassats is always finding strange that warm.lisp lives in cold/ directory 2014-04-25T16:21:09Z nyef: (On the level of ``I'm fuzzy on the whole good/bad thing. What do you mean, "bad"?'') 2014-04-25T16:24:56Z stassats: (sb-mop:class-slots (find-class 'standard-class)) => (# 2014-04-25T16:24:56Z stassats: # ...) 2014-04-25T16:25:12Z stassats: with non-closure restart-lisp 2014-04-25T16:25:27Z stassats: well, make-target-contribs? 2014-04-25T16:29:23Z stassats: Failed contribs:" asdf-install sb-aclrepl sb-bsd-sockets sb-cover sb-introspect sb-md5 sb-posix sb-simple-streams sb-sprof 2014-04-25T16:29:46Z nyef: Only nine? That's not bad. 2014-04-25T16:31:07Z Hydan is now known as Hydan` 2014-04-25T16:31:39Z stassats: how about some ./run-tests.sh then 2014-04-25T16:32:28Z Hydan` is now known as Hydan 2014-04-25T16:32:53Z stassats: needs sb-posix to run 2014-04-25T16:33:14Z stassats: but not badly 2014-04-25T16:34:08Z stassats: remember run-program being broken yesterday, perhaps that's why all the groveling things fail 2014-04-25T16:34:34Z nyef: Makes a certain amount of sense, yes. 2014-04-25T16:36:05Z Krystof: howdy. How do you feel if I blog about the current state? 2014-04-25T16:36:17Z slyrus joined #sbcl 2014-04-25T16:36:59Z nyef: Now I got a WARNING while compiling dfun.lisp, but the only changes I've made are the closure fix for call_into_lisp and setting QSHOW_SIGNALS. 2014-04-25T16:38:02Z nyef: Krystof: I'm actually sortof half surprised that nobody's blogged about it yet. 2014-04-25T16:38:28Z Krystof: I have an angle, since one of the GSoC projects is improving div-by-mul 2014-04-25T16:38:36Z Krystof: nice that we have a divless platform to make it useful :-) 2014-04-25T16:38:51Z nyef: Ah! found a warning in the scrollback. 2014-04-25T16:39:45Z nyef: http://paste.lisp.org/display/142291 2014-04-25T16:40:12Z stassats: (ASH -129876 -1026) => -32469 2014-04-25T16:41:21Z Hydan quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-25T16:41:58Z nyef: And that's fixnum/fixnum ASH, isn't it? 2014-04-25T16:43:03Z stassats: (disassemble #'list) is still BYTE #X1D, #XE0, #X4F, #XE2 2014-04-25T16:43:28Z nyef: Is that on a bare cold-core, or after loading stuff? 2014-04-25T16:43:41Z stassats: it's on ./run-sbcl.sh 2014-04-25T16:44:04Z nyef: So, no wonder I can't even get that far. (-: 2014-04-25T16:44:20Z Krystof: emblogginated 2014-04-25T16:44:51Z stassats: i'll try to pluck out arith.pure failures 2014-04-25T16:45:30Z stassats: bad math usually leads to all kinds of bad things 2014-04-25T16:45:48Z stassats: and run-program failure too 2014-04-25T16:48:06Z cracauer joined #sbcl 2014-04-25T16:48:09Z nyef: Turns out, the recompile restart got me well past dfun. 2014-04-25T16:50:41Z nyef: Paste annotated with backtrace and whatnot. 2014-04-25T16:59:35Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-25T16:59:54Z ltbarcly joined #sbcl 2014-04-25T17:04:27Z drmeister quit (Read error: Connection reset by peer) 2014-04-25T17:04:47Z drmeister joined #sbcl 2014-04-25T17:05:48Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-25T17:26:00Z nyef: stassats: On that shift/rotate thing, turns out that shift/rotate by register is limited to the low 8 bits of the shift count. 2014-04-25T17:26:14Z ltbarcly joined #sbcl 2014-04-25T17:26:26Z nyef: So it breaks at -256 as well. 2014-04-25T17:27:37Z stassats: i figured that, trying to come up with an optimal sequence 2014-04-25T17:28:12Z stassats: i guess i can just make it work 2014-04-25T17:29:35Z stassats: i created https://github.com/stassats/sbcl-arm for pushing things onto 2014-04-25T17:40:18Z stassats: (funcall (compile nil '(lambda (x) (declare (fixnum x)) (ash x -1024))) -129876) => -1 2014-04-25T17:40:31Z stassats: now to unsigned=>unsigned 2014-04-25T17:43:39Z Hydan joined #sbcl 2014-04-25T17:57:22Z yacks quit (Read error: Operation timed out) 2014-04-25T18:07:43Z stassats: disassemble is being missed 2014-04-25T18:07:53Z cracauer: hey 2014-04-25T18:08:04Z cracauer: anybody here with some knowledge of gencgc? 2014-04-25T18:08:11Z pkhuong: more or less 2014-04-25T18:08:15Z cracauer: I mailed the list a while back and I'm still tinkering with the patch 2014-04-25T18:08:19Z cracauer: Hey Paul. 2014-04-25T18:08:28Z pkhuong: nyef might have a better idea. 2014-04-25T18:08:51Z cracauer: So I am messing with these pages that have conservative pointers to them. 2014-04-25T18:09:25Z cracauer: Instead of moving over the whole page to newspace I don't do that and I let it be scavenged, and I hacked up the scavenging to not do anything to conservatively pointed to objects. 2014-04-25T18:09:29Z cracauer: So far so good. 2014-04-25T18:09:53Z cracauer: But I have the scavenging of *other* pages replace pointers in "my" pages with forwarding pointers. 2014-04-25T18:09:58Z cracauer: Now my question is: 2014-04-25T18:10:13Z cracauer: how does the GC now do that for pages that are in newspace? 2014-04-25T18:10:37Z cracauer: It scavenges them but somehow the code from other pages that puts forwarding pointers into the newspace pages never gets run 2014-04-25T18:12:31Z nyef: cracauer: Which email was this? 2014-04-25T18:12:55Z cracauer: "gencgc not collecting all garbage" 2014-04-25T18:13:11Z cracauer: the problem is that those pages are preserved in their entirety 2014-04-25T18:13:23Z cracauer: we do not wipe out useless pointers in there 2014-04-25T18:13:32Z cracauer: they nail down roots to potentially large trees 2014-04-25T18:13:43Z pkhuong: yeah, nyef is the one you want (: 2014-04-25T18:14:06Z cracauer: So I wrote code that uses the scavenging to mark beginning and size of objects that are conservatively being pointed to 2014-04-25T18:14:20Z cracauer: then I wipe out all the other words in those pages as the last stage of GC 2014-04-25T18:14:36Z cracauer: except for the mental block I describe above 2014-04-25T18:15:07Z nyef: It's going to be about 15-45 minutes before I can really pay attention, I'm sorry. 2014-04-25T18:18:04Z cracauer: Oh I guess it is gc-common.c:229: if (from_space_p(object)) { 2014-04-25T18:19:11Z cracauer: I can't just take those pages out of newspace and stuff them into fromspace either. 2014-04-25T18:19:23Z cracauer: I might have to special-case them entirely. 2014-04-25T18:24:41Z stassats: my ash fix seems to have exposed a problem with ub8-bash-copy 2014-04-25T18:24:47Z stassats: it now fails at cold-init 2014-04-25T18:25:26Z stassats: 198277016064 not of type ub32 2014-04-25T18:29:12Z nyef: Okay, hopefully I now have some mental bandwidth to work with. 2014-04-25T18:31:13Z cracauer: it's more like latency in my case :-) 2014-04-25T18:31:28Z nyef: On my list for gencgc hacks is evacuating the non-pinned objects that are on pinned pages, replacing them with "free-space" objects. 2014-04-25T18:32:02Z cracauer: sounds like what I am doing. 2014-04-25T18:32:13Z nyef: That's the impression I get, yes. 2014-04-25T18:32:54Z cracauer: How would you go about finding beginning and size of the objects being pointed to conservatively? 2014-04-25T18:33:03Z cracauer: I went with marking them during scavenge 2014-04-25T18:33:13Z cracauer: problem is that not all of them go through scavenge 2014-04-25T18:33:46Z nyef: I'm not interested in conservative roots, I'm interested in pinned objects. 2014-04-25T18:34:30Z cracauer: right 2014-04-25T18:35:20Z nyef: On precise gencgc systems (as opposed to conservative systems), we have a per-thread list of pinned objects. 2014-04-25T18:35:21Z jsnell_: walking the pinned pages explicitly should be trivial 2014-04-25T18:35:50Z nyef: The issue isn't walking the pages, it's knowing which objects in the pages are the important ones. 2014-04-25T18:36:10Z nyef: Hrm. 2014-04-25T18:36:17Z cracauer: the pages are being walked during scavenge. However, what I do is get the size of the object when I am about to scavenge 2014-04-25T18:36:19Z nyef: I think I'm looking at an old version of gencgc.c. 2014-04-25T18:36:21Z jsnell_: the object boundaries are trivial to find, and for this purpose the internal structure doesn't matter 2014-04-25T18:36:34Z cracauer: then I check whether that overlaps with what is being pointed to conservatively 2014-04-25T18:36:34Z jsnell_: internal structure of the objects 2014-04-25T18:36:43Z cracauer: and not everything gets covered by the scavenge 2014-04-25T18:38:03Z nyef: So, first, on x86oids, you have actually-conservative pointer detection, but there's a shortcut in the logic that won't check in detail if a page already contains a pinned object. 2014-04-25T18:38:04Z cracauer: but that is the smaller one of my problems 2014-04-25T18:38:46Z cracauer: the bigger one is that I don't understand how to prevent the insertion of forwarding pointers in "my" pages when they are not in newspace 2014-04-25T18:39:07Z nyef: Why can you not relocate them to newspace? 2014-04-25T18:39:29Z cracauer: hmmm 2014-04-25T18:39:40Z cracauer: don't recall. Must have been in the email thread 2014-04-25T18:39:51Z cracauer: now that you ask it sounds like I can leave them there 2014-04-25T18:40:35Z cracauer: let me just run my patch so far leaving them there 2014-04-25T18:40:59Z cracauer: random segfault. Not very helpful 2014-04-25T18:41:38Z nyef: Okay, I think we need another code-path for "pinned" pages: They need a bitmap indicating the header of live objects within them. 2014-04-25T18:41:52Z cracauer: that's what I do 2014-04-25T18:42:09Z cracauer: the "page" has a list of things being pointed to 2014-04-25T18:42:26Z cracauer: then during scavenge I check for overlap between what is to be scavenged and what is marked 2014-04-25T18:42:31Z nyef: Not merely conservative roots, but all known-to-be-live objects? 2014-04-25T18:42:45Z cracauer: oh 2014-04-25T18:43:02Z cracauer: if I put the pages into newspace I will have to do that, right? 2014-04-25T18:43:17Z cracauer: so far I didn't move them to newspace 2014-04-25T18:44:19Z cracauer: actually, no 2014-04-25T18:44:20Z nyef: I'd still move them to newspace, maintain that bitmap, remove the shortcut checks in valid_conservative_root_p() about skipping tests if it's already marked dont_move (because you need to know ALL of the possibly-live objects, not just the first one with a conservative root)... 2014-04-25T18:44:37Z cracauer: what I actually do is move them to newspace at the end of GC 2014-04-25T18:44:42Z cracauer: instead of freeing them 2014-04-25T18:44:55Z nyef: That's sort of what happens with dont_move / pinned pages. 2014-04-25T18:45:15Z cracauer: no, so far they get moved to newspace at the beginning of gc 2014-04-25T18:45:23Z cracauer: my current patch does it at the end 2014-04-25T18:45:38Z nyef: Looks like last time I messed with this stuff was about four months ago. 2014-04-25T18:46:08Z cracauer: I think I do get all roots correctly 2014-04-25T18:46:34Z cracauer: can somebody explain how the scavenging code knows 2014-04-25T18:46:44Z cracauer: not to replace pointers with forwarding pointers 2014-04-25T18:46:50Z cracauer: in pages in new_space? 2014-04-25T18:46:53Z cracauer: I don't get it 2014-04-25T18:49:11Z nyef: It's near the top of scavenge(). There's a from_space_p() check... That you've already found. 2014-04-25T18:49:48Z nyef: Objects in from space get overwritten with forwarding pointers. Objects not in from space don't. 2014-04-25T18:49:50Z cracauer: but that is for the page being scavenged, not possible victims of forwarding pointer inservtions 2014-04-25T18:50:08Z cracauer: oh 2014-04-25T18:50:11Z cracauer: maybe it's not 2014-04-25T18:50:15Z nyef: No, it's for the pointer IN the page being scavenged. 2014-04-25T18:50:26Z cracauer: <== blind 2014-04-25T18:51:10Z cracauer: I wonder whether can I just change that to say from_space_p||dont_move? 2014-04-25T18:51:47Z nyef: Yes, as dont_move is already set up by this point, but you'll want to conditionalize it on gencgc. 2014-04-25T18:51:47Z cracauer: would that fall through to the right code? 2014-04-25T18:52:00Z nyef: There may be further subtleties. 2014-04-25T18:52:13Z cracauer: I already have an interface for gc-common to determine dontmove pages from gencgc 2014-04-25T18:52:25Z cracauer: allright 2014-04-25T18:52:33Z nyef: Oh! If from_space_p and dont_move, you need to mark the extra live object in your page. 2014-04-25T18:52:34Z cracauer: I think that answers my original question 2014-04-25T18:52:46Z cracauer: "extra"? 2014-04-25T18:53:00Z nyef: Well, if it's not already marked as being a live object. 2014-04-25T18:53:23Z nyef: scavenge() is how we typically FIND non-pinned objects, after all. 2014-04-25T18:54:36Z cracauer: do you mean marking the dontmove object or the other objects? 2014-04-25T18:54:46Z nyef: The other objects. 2014-04-25T18:55:21Z nyef: Or, if you only want the pinned objects on that page, we can get more subtle. 2014-04-25T18:55:36Z cracauer: oh, ok 2014-04-25T18:55:48Z cracauer: no, what I intend to do is move the non-pinned objects out 2014-04-25T18:55:54Z cracauer: with the normal scavenging 2014-04-25T18:56:05Z cracauer: and then wipe their slots later 2014-04-25T18:56:26Z cracauer: The code I have lets scavenge run over the page with the pinned object 2014-04-25T18:56:41Z nyef: Okay, so we need more subtlety. 2014-04-25T18:56:42Z cracauer: I then have a function telling scavenge not to scavenge this particular object 2014-04-25T18:56:56Z cracauer: but it does scavenge the other objects normally 2014-04-25T18:57:07Z nyef: You want to set forwarders on everything in that page that isn't pinned. 2014-04-25T18:57:18Z nyef: You also need to check for a forwarder at that point. 2014-04-25T18:57:47Z cracauer: right, so I can't just do the above ||dontmove on the whole page 2014-04-25T18:57:53Z cracauer: I need it for the object only 2014-04-25T18:58:15Z nyef: Right, so instead, where it says "Scavenge that pointer.", add a check against your pin list. 2014-04-25T18:58:34Z cracauer: already have code doing that check, but in a different place 2014-04-25T18:59:12Z nyef: ... Unless that's wrong. Hrm. 2014-04-25T18:59:28Z cracauer: moving that code to the from_space_p place will likely also solve the other problem I mentioned 2014-04-25T18:59:46Z cracauer: that I didn't see everything (attempted) scavenged that was pinned 2014-04-25T19:00:14Z cracauer: But I think I need one more bit: 2014-04-25T19:00:27Z nyef: I must say, this is a nice change of pace from trying to figure out problems with the ARM port. (-: 2014-04-25T19:01:00Z cracauer: if I do it that simply I don't think I mark things pointed to by pinned objects 2014-04-25T19:01:43Z nyef: Right, I'd suggest scavenging your pinned objects first, before you do the normal heap scavenge. 2014-04-25T19:01:54Z cracauer: so for the pinned objects I need to feed them through (scavtab[widetag_of(object)])(object_ptr, object); 2014-04-25T19:02:04Z cracauer: I can't scan them first 2014-04-25T19:02:16Z cracauer: at the beginning of this I don't know beginning and size 2014-04-25T19:02:25Z cracauer: of the pinned objects 2014-04-25T19:03:21Z nyef: In possibly_valid_dynamic_space_pointer(), you know the start of the object. 2014-04-25T19:03:22Z jsnell_: but that can be done as a separate preprocessing step 2014-04-25T19:03:31Z nyef: And we have sizetab[]. 2014-04-25T19:04:22Z nyef: From there, you call scavenge() over the entire object. Has to be done after all of the pinned objects have been marked. 2014-04-25T19:05:09Z cracauer: what I am doing so far is that in scavenge I call sizetab to get the size of the object 2014-04-25T19:05:33Z cracauer: then I call into my helped in gencgc which tells me whether this pointer+size covers the pointed to thing 2014-04-25T19:05:39Z cracauer: pointed-to-thing 2014-04-25T19:07:15Z tmh_ quit (Changing host) 2014-04-25T19:07:15Z tmh_ joined #sbcl 2014-04-25T19:07:18Z cracauer: I'm getting confused now 2014-04-25T19:07:23Z cracauer: brb 2014-04-25T19:07:50Z ehaliewicz joined #sbcl 2014-04-25T19:09:36Z cracauer: in the page with the pinned objects, what do I end up doing differently on the pinned and the non-pinned objects 2014-04-25T19:09:48Z cracauer: I guess nothing different when scavenging *that* page 2014-04-25T19:10:07Z cracauer: only difference is in scavenging *other* pages, then I don't do forwarding pointers 2014-04-25T19:10:21Z cracauer: in that page 2014-04-25T19:10:27Z nyef: Right. But you only don't do forwarding pointers for the pinned object. 2014-04-25T19:10:38Z cracauer: right 2014-04-25T19:10:48Z cracauer: and then I will late be free to wipe out the other words 2014-04-25T19:10:54Z nyef: Which, for a single code-object, might involve several headers. 2014-04-25T19:11:01Z cracauer: because they have been transported out 2014-04-25T19:11:05Z nyef: Right. 2014-04-25T19:11:42Z cracauer: if I call scavtab[] on every object in that page, things stay put, right? 2014-04-25T19:11:56Z cracauer: they only move when calling scavtag on *other* pages 2014-04-25T19:12:27Z nyef: I *think* so? 2014-04-25T19:12:49Z nyef: As I said, it's been a few months since last I was messing with this. 2014-04-25T19:13:08Z cracauer: well it sounds like it gets me several steps ahead from where I am. 2014-04-25T19:13:23Z cracauer: this stuff is a bit annoying to debug :-) 2014-04-25T19:13:37Z nyef: Indeed it is. 2014-04-25T19:13:43Z nyef glares at PURIFY. 2014-04-25T19:13:45Z cracauer: thanks so much 2014-04-25T19:13:55Z nyef: Good luck. 2014-04-25T19:13:56Z cracauer: BTW, I won't make it to Paris for the conference after all. 2014-04-25T19:14:06Z cracauer: And Berlin is cancelled :-( 2014-04-25T19:15:25Z Krystof: that is a shame! 2014-04-25T19:17:58Z cracauer: Montreal in August doesn't suck, though 2014-04-25T19:20:20Z nyef: Hrm... There's SPARC-specific bits in purify, but they don't apply to the specific bit of shared nature between ARM and SPARC... 2014-04-25T19:20:49Z cracauer: What do you want to run the arm port on? 2014-04-25T19:21:02Z stassats: toasters 2014-04-25T19:21:18Z cracauer: but then you also need a working netbsd port 2014-04-25T19:21:20Z nyef: I have a raspberry pi and a beaglebone black... and a galaxy note 10.1... 2014-04-25T19:21:29Z nyef: In all cases, Linux. 2014-04-25T19:21:48Z cracauer: does the raspberry pi have virtual memory and a MMU? 2014-04-25T19:21:57Z nyef: Yes, it does. 2014-04-25T19:22:06Z nyef: And an FPU as well. 2014-04-25T19:22:10Z cracauer: our babies have grown up so much 2014-04-25T19:26:12Z nyef: ... Why is purify using os_zero() instead of scrub_control_stack()? 2014-04-25T19:28:13Z erikvarga left #sbcl 2014-04-25T19:40:33Z Hydan quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-25T19:49:01Z nyef: ... when was the last time someone successfully built a cheneygc port? 2014-04-25T19:49:25Z nyef: At this point I'm entertaining the possibility that purify has bit-rotted. 2014-04-25T19:55:51Z stassats: what's on cheneygc? 2014-04-25T19:55:58Z stassats: sparc? 2014-04-25T19:56:17Z nyef: HPPA, ALPHA, MIPS. PPC and SPARC by a build-time option. 2014-04-25T19:56:32Z nyef: (said option being --without-gencgc --with-cheneygc) 2014-04-25T19:56:36Z stassats: i only ever built on ppc and sparc, besides x86oids 2014-04-25T19:57:02Z nyef: I haven't built on MIPS in a while, because it takes too bloody long. 2014-04-25T19:57:06Z stassats: and an ash fix coming up 2014-04-25T19:57:24Z stassats: took me too long to see the difference between signed and unsigned versions (it's one letter) 2014-04-25T19:57:37Z nyef: On the other hand, I have a faster MIPS box now, and as soon as I've managed to port Linux to it I should be able to do more regular MIPS builds. 2014-04-25T19:58:01Z nyef: Ooh. Does this also fix the UB8 copy thing? 2014-04-25T19:58:06Z stassats: not sure 2014-04-25T19:58:11Z nyef: Fair enough. 2014-04-25T19:58:13Z stassats: i discovered it in arith.pure 2014-04-25T19:58:27Z nyef: Yay for being able to run part of the test suite, huh? 2014-04-25T19:59:34Z nyef: You realize that as soon as we announce this thing, people are all going to be "grar, grar, soft-float, grar, gencgc, grar, iwmmxt, grar, THREADS, grar", right? 2014-04-25T20:00:49Z stassats: back to building the warm core 2014-04-25T20:01:13Z nyef: Did you ever try that closure-initial-function fix? 2014-04-25T20:01:27Z stassats: nope 2014-04-25T20:01:45Z stassats: i'll leave to you 2014-04-25T20:01:52Z nyef: Fair enough. 2014-04-25T20:02:13Z stassats: i'll see what arith.pure has to offer and switch to fixing run-program 2014-04-25T20:02:20Z nyef: It's already in my working tree, but since I haven't gotten to the point of working SLAD yet... 2014-04-25T20:02:47Z nyef: If run-program requires more than one level of interrupt nesting it's not going to work. 2014-04-25T20:02:53Z Krystof: is grar an arm instruction? 2014-04-25T20:03:07Z nyef: ... I hope not. 2014-04-25T20:03:26Z stassats: make-target-2 goes surprisingly far 2014-04-25T20:03:49Z stassats: far as in until the end, let's blame my ash fix 2014-04-25T20:04:10Z stassats: completely without reusing old fasls 2014-04-25T20:04:17Z nyef: Sweet. 2014-04-25T20:06:44Z segv- quit (Ping timeout: 255 seconds) 2014-04-25T20:07:24Z stassats: nyef: pushed to https://github.com/stassats/sbcl-arm 2014-04-25T20:07:56Z nyef: Okay, now I have to figure out how to merge it. 2014-04-25T20:08:04Z stassats: i didn't strive for optimiality, just to get it to work 2014-04-25T20:08:31Z stassats: we can indulge in optimizations once everything's working 2014-04-25T20:11:23Z nyef: Okay, building now. 2014-04-25T20:11:32Z _hmt joined #sbcl 2014-04-25T20:15:11Z drmeister quit (Read error: Connection reset by peer) 2014-04-25T20:15:40Z drmeister joined #sbcl 2014-04-25T20:18:34Z stassats: one test failing due to some state, moved it into the beginning, passes 2014-04-25T20:18:52Z stassats: others failing due to my floating point ABI and another due to missing EQL for ratios 2014-04-25T20:19:38Z stassats: and arith.impure passes 2014-04-25T20:19:47Z stassats: alright then, run-program beckons 2014-04-25T20:21:12Z nyef: cracauer: Last May, there was a thread on sbcl-devel entitled "lisp interface to GC". Would some of the things discussed in that thread be of use with your little toy? 2014-04-25T20:24:58Z stassats: it's getting late and i still don't have working slime 2014-04-25T20:25:14Z nyef: ... And copy-ub8-to-system-area still occasionally dies for me. 2014-04-25T20:25:36Z stassats: perhaps i got lucky 2014-04-25T20:27:56Z edgar-rft quit (Quit: lifetime stopped because experience terminated) 2014-04-25T20:28:47Z stassats: i wonder whether sb-dynamic-core works 2014-04-25T20:29:01Z nyef: Doesn't that require linkage-table? 2014-04-25T20:29:05Z stassats: yes 2014-04-25T20:29:17Z stassats: otherwise i'll have re-slam on changing run-program's c part 2014-04-25T20:29:18Z nyef: Hunh. Now, why, when I PURIFY, does it GC immediately afterwards... and why are there two interrupt contexts at that point? 2014-04-25T20:29:30Z nyef: We don't have linkage-tables yet. 2014-04-25T20:30:02Z nyef: Preliminary conclusion: Purify actually works. 2014-04-25T20:31:57Z stassats: ah, sigchild? 2014-04-25T20:33:21Z nyef: So, we're probably looking at the same issue, and it's on that's been on my list for quite a while. 2014-04-25T20:33:29Z stassats: run-program dies with "signal 4 received" 2014-04-25T20:33:33Z stassats: though, sigchld is 17 2014-04-25T20:33:43Z stassats: 4 is sigill 2014-04-25T20:36:18Z stassats: when i install my own signal handler, it faults at fffffffc 2014-04-25T20:36:47Z nyef: Interrupt handling generally is hosed, and we've known this for quite a while now. 2014-04-25T20:37:38Z nyef: In fact, it's possibly also what's affecting copy-ub8-to-system-area. 2014-04-25T20:42:25Z stassats: (sb-sys::enable-interrupt sb-unix::sigchld :default) mitigates the problem 2014-04-25T20:44:36Z stassats: back to contrib building 2014-04-25T20:46:08Z stassats: can't build ASDF 2014-04-25T20:46:16Z stassats: i could before 2014-04-25T20:48:30Z stassats: copy-ub8-to-system-area 2014-04-25T20:49:56Z cracauer: nyef, checking that thread now 2014-04-25T20:50:38Z stassats: backtrace is strange 2014-04-25T20:51:05Z stassats: write-sequence is 0 to 80, then later it becomes 502 to 80 2014-04-25T20:52:11Z stassats: maybe they have a different meaning 2014-04-25T20:52:37Z nyef: Could also be a doubled interrupt scribbling on the context. 2014-04-25T20:52:59Z stassats: 502 seems legit 2014-04-25T20:53:11Z nyef: Oh, for another register. 2014-04-25T20:54:00Z nyef: ... Here's a crazy thought: Do we need to keep the LISP stack pointer in the register set? 2014-04-25T20:54:28Z stassats: compiled asdf.fasl on the nth try 2014-04-25T20:56:28Z nyef: If I can put CSP into a static symbol, interrupt handling basically sorts itself out. 2014-04-25T21:01:47Z stassats: maybe qshow can shed some lite 2014-04-25T21:01:52Z stassats: or light 2014-04-25T21:03:36Z stassats: blimey, now i compile asdf fine three times in a row 2014-04-25T21:05:57Z stassats: i now get The value 43646266416563712 is not of type (UNSIGNED-BYTE 32). 2014-04-25T21:06:03Z stassats: in (SB-KERNEL:SHIFT-TOWARDS-END 1729300840 -24) 2014-04-25T21:06:43Z nyef: So... Still more fun and games with the shifter? 2014-04-25T21:06:54Z stassats: but it works fine! (SB-KERNEL:SHIFT-TOWARDS-END 1729300840 -24) => 319383552 2014-04-25T21:08:00Z stassats: maybe it doesn't survive gc? 2014-04-25T21:08:20Z stassats: due to all the bignum consing 2014-04-25T21:08:41Z nyef: Pushed the call_into_lisp funcallable-instance/closure fix. 2014-04-25T21:10:39Z stassats: i'll try to make a bignum thrasher 2014-04-25T21:11:29Z nyef: I'm finding the idea of moving CSP out to a symbol and moving NSP into a register to be a very attractive, yet horrifying, thing to attempt. 2014-04-25T21:15:03Z stassats: can't trigger any problems 2014-04-25T21:16:59Z nyef: I can't even think about trying this now. 2014-04-25T21:17:39Z nyef: It's right there, in the heart of our register pressure problems, and I suddenly need another register in order to do anything with CSP? 2014-04-25T21:17:46Z stassats: but i get some kind of modern art http://i.imgur.com/Xexstzc.png 2014-04-25T21:18:15Z nyef: Heh. Mine eyes glazeth over. 2014-04-25T21:21:41Z nyef breathes a sigh of relief. 2014-04-25T21:22:07Z nyef: I can do it. The uses of CSP-TN directly around the areas of highest register pressure can ALL be finessed. 2014-04-25T21:22:49Z nyef: I am not, however, doing it NOW. 2014-04-25T21:23:11Z drmeister quit (Remote host closed the connection) 2014-04-25T21:26:08Z slyrus: stassats: that could be the cover for your new album! 2014-04-25T21:26:39Z stassats: the second will be the same, but with CORRUPTION WARNINGCORRUPTION WARNING 2014-04-25T21:28:25Z _hmt quit (Remote host closed the connection) 2014-04-25T21:32:03Z stassats: nyef: what about object pinning? 2014-04-25T21:32:45Z stassats: SAP bashing should be using, does it work? 2014-04-25T21:33:26Z drmeister joined #sbcl 2014-04-25T21:34:15Z stassats: though, buffer-output doesn't pin anything 2014-04-25T21:35:09Z stassats: but it works everywhere else 2014-04-25T21:36:53Z nyef: Object pinning should be a non-issue, it just does a without-gcing. 2014-04-25T21:37:05Z stassats: maybe that's the issue? 2014-04-25T21:37:15Z nyef: On the other hand, if we're in a signal handler already and then a without-gcing, and a GC gets called for we're sunk. 2014-04-25T21:37:43Z nyef: Yeah, at this point I suspect that a lot of stuff will clear up and start working as soon as I fix our interrupt-handling issues. 2014-04-25T21:38:00Z stassats: can't wait 2014-04-25T21:38:11Z stassats: i need to find me more arithmetic failures 2014-04-25T21:38:23Z stassats: or just start implementing softfp calling convention 2014-04-25T21:59:23Z drmeister quit (Read error: Connection reset by peer) 2014-04-25T22:16:58Z drmeister joined #sbcl 2014-04-25T22:32:42Z cmack: finally was able to get sb-show working for the u3: http://paste.lisp.org/display/142290#3 2014-04-25T22:32:48Z eudoxia joined #sbcl 2014-04-25T22:33:40Z stassats: i don't see sb-show output 2014-04-25T22:34:28Z stassats: is your image and sbcl binary in sync? 2014-04-25T22:35:09Z cmack: hmm.. ok well I that's all I get :( where the ':' lines are repeating a long time. I guess I missed a line where the control stack exhausted 2014-04-25T22:35:45Z cmack: ... does that mean I can't use sbcl 1.1.17 as a xc-host? 2014-04-25T22:35:57Z stassats: you can 2014-04-25T22:38:35Z cmack: I'm not sure how to check if image and binary are in sync 2014-04-25T22:40:00Z stassats: compile it again and make sure sbcl.nm and stuff-groveled-from-headers.lisp are copied to the host 2014-04-25T22:42:04Z cmack: hmm well I've been rsyncing everything back and forth... I'll have to investigate when I get back; just realize I'm late for my CSA veggies pickup 2014-04-25T23:00:34Z eudoxia quit (Quit: leaving) 2014-04-25T23:05:08Z stassats: my VOP apparently works but doesn't pass the cold-init 2014-04-25T23:05:20Z stassats: and i can't seen anything wrong with it 2014-04-25T23:09:35Z nyef: Dare I ask? 2014-04-25T23:10:24Z stassats: shift-towards-end/start as VOPs 2014-04-25T23:11:02Z nyef: Ah. 2014-04-25T23:11:09Z stassats: the most trivial of things 2014-04-25T23:11:39Z nyef: ... OCFP isn't live during call-in, is it? 2014-04-25T23:12:14Z stassats: is it important for temporaries to have non-descriptor-reg sc? 2014-04-25T23:12:22Z stassats: instead of just signed/unsigned? 2014-04-25T23:13:01Z nyef: No, the important distinction for temporaries is boxed/unboxed. 2014-04-25T23:13:26Z stassats: i just see a lot of non-descriptor-reg 2014-04-25T23:13:44Z nyef: Because that's the important attribute. 2014-04-25T23:13:57Z stassats: i suppose it doesn't really matter if a temporary is unsigned or signed 2014-04-25T23:14:05Z stassats: except for the debugger 2014-04-25T23:14:34Z nyef: signed-reg/unsigned-reg overlap the non-descriptor-reg space, but have specific implication for PRIMTYPE. 2014-04-25T23:15:50Z stassats: maybe my vop is misvopified by the host 2014-04-25T23:16:00Z stassats: at the REPL, it works whatever i throw at it 2014-04-25T23:16:33Z nyef: Paste? 2014-04-25T23:17:29Z stassats: http://paste.lisp.org/display/142293 2014-04-25T23:19:13Z nyef: ... AND? Really? 2014-04-25T23:19:44Z stassats: what else is there? 2014-04-25T23:19:53Z ltbarcly_ joined #sbcl 2014-04-25T23:19:56Z nyef: I thought the point was that it should have an upper bound, not that it wrap sooner? 2014-04-25T23:20:22Z stassats: no, that's what it does 2014-04-25T23:20:34Z stassats: in the lisp version: (ldb (byte (1- (integer-length sb!vm:n-word-bits)) 0) count) 2014-04-25T23:20:35Z nyef: Hrm. 2014-04-25T23:21:57Z ltbarcly quit (Ping timeout: 252 seconds) 2014-04-25T23:23:10Z stassats: and it works at the repl, no problem 2014-04-25T23:23:55Z nyef: And the amount is signed? 2014-04-25T23:24:09Z stassats: yes 2014-04-25T23:24:18Z nyef: Hrm. 2014-04-25T23:25:49Z nyef: I'd ask if the "lower 5 bits" thing were a red herring, but it looks right from what I remember of x86 assembly. 2014-04-25T23:26:24Z stassats: the lisp version works like this too 2014-04-25T23:27:58Z nyef: Heh! For a laugh, check out the commentary around the x86-64 version of these VOPs. 2014-04-25T23:28:35Z stassats: well, it has 64 bits 2014-04-25T23:28:47Z nyef: For a 32-bit operation? 2014-04-25T23:29:02Z stassats: is it 32-bit? 2014-04-25T23:29:11Z nyef: According to the COMMENT it is. (-: 2014-04-25T23:29:22Z stassats: ah, that 2014-04-25T23:31:36Z nyef: Nothing seems wrong about these. 2014-04-25T23:31:51Z stassats: maybe they just break something else 2014-04-25T23:31:58Z stassats: they are used as an argument to logior, for instance 2014-04-25T23:42:34Z stassats: and i don't really know what it breaks, so can't insert print statements 2014-04-25T23:44:38Z nyef: Was the non-VOP version returning incorrect results? 2014-04-25T23:44:49Z White_Flame quit (Ping timeout: 265 seconds) 2014-04-25T23:45:44Z stassats: it didn't crash the cold-init, so, i suppose it didn't 2014-04-25T23:51:27Z White_Flame joined #sbcl 2014-04-25T23:56:49Z stassats: ok, that's a puzzle for tomorrow 2014-04-25T23:57:26Z nyef: I will possibly be unavailable for much of tomorrow. 2014-04-25T23:57:49Z nyef: I may-or-may-not have something in terms of fixed interrupt handling by then. 2014-04-26T00:02:42Z oleo quit (Ping timeout: 265 seconds) 2014-04-26T00:02:56Z oleo joined #sbcl 2014-04-26T00:03:26Z stassats quit (Ping timeout: 255 seconds) 2014-04-26T00:04:28Z psilord joined #sbcl 2014-04-26T00:05:22Z drmeister quit (Remote host closed the connection) 2014-04-26T00:19:24Z nyef: ... On the one hand, I find it hard to believe that I managed to scare up another register. On the other, where the heck am I going to scare up ANOTHER one? 2014-04-26T00:41:24Z ltbarcly_ quit (Quit: Computer has gone to sleep.) 2014-04-26T00:54:00Z ltbarcly joined #sbcl 2014-04-26T01:25:04Z ltbarcly_ joined #sbcl 2014-04-26T01:26:55Z ltbarcly quit (Ping timeout: 276 seconds) 2014-04-26T01:30:23Z ltbarcly_ quit (Quit: Computer has gone to sleep.) 2014-04-26T01:30:45Z nyef: Ugh. Trying to bisect something in my working tree is a pain. 2014-04-26T01:38:54Z joshe: I keep an extra clone or two around for bisecting 2014-04-26T01:39:49Z nyef: No, I mean, I have a bunch of changes in ONE FILE in my working tree, and some portion of them are wrong, but I can't really tell just by looking. 2014-04-26T01:40:13Z joshe: oh, right 2014-04-26T01:47:29Z nyef: Okay, got it down to one hunk that is broken. That took too much doing. /-: 2014-04-26T01:54:12Z drmeister joined #sbcl 2014-04-26T01:58:38Z drmeister quit (Ping timeout: 240 seconds) 2014-04-26T02:10:38Z ivan`` quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2014-04-26T02:15:22Z ivan`` joined #sbcl 2014-04-26T02:15:57Z ivan`` quit (Remote host closed the connection) 2014-04-26T02:18:37Z zRecursive joined #sbcl 2014-04-26T02:19:21Z ivan`` joined #sbcl 2014-04-26T02:20:44Z ivan`` quit (Remote host closed the connection) 2014-04-26T02:26:22Z ivan`` joined #sbcl 2014-04-26T02:30:43Z nyef quit (Quit: G'night all.) 2014-04-26T02:39:08Z christoph_debian quit (Ping timeout: 255 seconds) 2014-04-26T02:47:59Z ltbarcly joined #sbcl 2014-04-26T02:52:16Z christoph_debian joined #sbcl 2014-04-26T03:17:55Z drmeister joined #sbcl 2014-04-26T03:19:27Z drmeister quit (Read error: Connection reset by peer) 2014-04-26T03:19:37Z drmeiste_ joined #sbcl 2014-04-26T03:46:13Z zRecursive quit (Remote host closed the connection) 2014-04-26T04:04:35Z attila_lendvai joined #sbcl 2014-04-26T04:04:36Z attila_lendvai quit (Changing host) 2014-04-26T04:04:36Z attila_lendvai joined #sbcl 2014-04-26T04:21:53Z ehaliewicz quit (Read error: No route to host) 2014-04-26T04:25:52Z pranavrc joined #sbcl 2014-04-26T04:25:53Z pranavrc quit (Changing host) 2014-04-26T04:25:53Z pranavrc joined #sbcl 2014-04-26T04:29:36Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-26T04:32:14Z drmeiste_ quit (Remote host closed the connection) 2014-04-26T05:48:59Z Intensity quit (Ping timeout: 246 seconds) 2014-04-26T05:59:58Z edgar-rft joined #sbcl 2014-04-26T06:11:10Z PuercoPop joined #sbcl 2014-04-26T06:34:56Z yacks joined #sbcl 2014-04-26T06:50:03Z edgar-rft quit (Quit: existence interrupted because all hope lost) 2014-04-26T06:52:24Z edgar-rft joined #sbcl 2014-04-26T06:53:29Z cmack quit (Ping timeout: 252 seconds) 2014-04-26T06:54:16Z cmack joined #sbcl 2014-04-26T08:00:06Z stassats joined #sbcl 2014-04-26T08:00:55Z scymtym_ joined #sbcl 2014-04-26T08:29:12Z pranavrc quit (Remote host closed the connection) 2014-04-26T09:14:18Z scymtym_ quit (Ping timeout: 252 seconds) 2014-04-26T09:28:08Z Hydan joined #sbcl 2014-04-26T09:33:12Z kludge` quit (Ping timeout: 240 seconds) 2014-04-26T09:38:52Z kludge` joined #sbcl 2014-04-26T09:55:35Z Hydan is now known as Hydan` 2014-04-26T10:05:44Z erikvarga joined #sbcl 2014-04-26T10:21:32Z pnpuff joined #sbcl 2014-04-26T10:22:05Z pnpuff left #sbcl 2014-04-26T10:37:49Z stassats` joined #sbcl 2014-04-26T10:40:16Z Hydan` is now known as Hydan 2014-04-26T10:41:36Z stassats quit (Ping timeout: 240 seconds) 2014-04-26T10:57:25Z Hydan is now known as Hydan` 2014-04-26T11:27:24Z Hydan` is now known as Hydan 2014-04-26T11:29:49Z nyef joined #sbcl 2014-04-26T11:30:05Z nyef: G'morning all. 2014-04-26T11:31:57Z stassats`: i got my shift-towards- problem, it misallocated registers 2014-04-26T11:32:22Z stassats`: and i somehow got it to work, no idea how 2014-04-26T11:32:36Z stassats`: pushed onto https://github.com/stassats/sbcl-arm 2014-04-26T11:32:41Z nyef: And I haven't had any more trouble from the ub8 copy thing. 2014-04-26T11:32:55Z nyef: Or run-program, for that matter. 2014-04-26T11:32:59Z stassats`: good 2014-04-26T11:33:16Z stassats`: i have problem with immediate encoding, i fixed it, but it breaks again something else 2014-04-26T11:33:19Z nyef: Purify still doesn't work, though. )-: 2014-04-26T11:33:58Z stassats`: alloc-number-stack-space blindly trusts the amount to allocate 2014-04-26T11:34:13Z stassats`: i split it to 8 bit chunks, something else fails 2014-04-26T11:34:43Z nyef: I'm about to affect number stack allocation, but not that aspect of it. 2014-04-26T11:35:09Z stassats`: it says something about segment size mismatching 2014-04-26T11:35:17Z stassats`: rebuilding to get back to it 2014-04-26T11:36:35Z stassats`: maybe you see something wrong here: http://paste.lisp.org/display/142293#1 2014-04-26T11:37:37Z stassats`: and it prevents building ASDF, which really should be unrelated 2014-04-26T11:38:06Z stassats`: code-length=460, n-written=456 2014-04-26T11:38:11Z nyef: Can't use COMPOSITE-IMMEDIATE-INSTRUCTION? 2014-04-26T11:38:13Z stassats`: SB-FASL::DUMP-SEGMENT 2014-04-26T11:38:27Z stassats`: i don't know what composite-immediate-instruction does 2014-04-26T11:38:48Z nyef: It's for precisely this. 2014-04-26T11:39:16Z stassats`: i thought it was just for loading immedites into a register 2014-04-26T11:39:28Z stassats`: so, SB-FASL::DUMP-SEGMENT fails with code-length=460, n-written=456 2014-04-26T11:39:47Z nyef: At XC time, or on the target? 2014-04-26T11:39:53Z stassats`: building ASDF 2014-04-26T11:40:12Z stassats`: without alloc-number-stack-space changes, it builds fine 2014-04-26T11:40:42Z nyef: But you need these changes elsewhere? 2014-04-26T11:40:51Z stassats`: to build sb-bsd-sockets 2014-04-26T11:41:24Z nyef: Have a look at arith.lisp, there's some trickier use of composite-immediate-instruction there. 2014-04-26T11:41:58Z nyef: And I'm prepping a commit series that should affect these VOPs anyway, so one of use will need to merge. 2014-04-26T11:42:26Z stassats`: i can hold back committing it 2014-04-26T11:42:29Z stassats`: if i get it to work 2014-04-26T11:44:49Z stassats`: and i don't mind rebasing my repo 2014-04-26T11:45:03Z nyef: I'll pull/rebase first if I can. 2014-04-26T11:47:15Z stassats`: you can pull the two arith changes for now 2014-04-26T11:47:50Z nyef: And pushed. 2014-04-26T11:49:55Z stassats`: i can now build asdf, well, weird 2014-04-26T11:50:07Z nyef: Smaller generated code, maybe? 2014-04-26T11:50:19Z stassats`: maybe, but why would it care? 2014-04-26T11:50:39Z nyef: Alternately, the stack and interrupt games probably damaged a lot of stuff. 2014-04-26T11:52:02Z stassats`: sb-bsd-sockets are in now too 2014-04-26T11:52:07Z stassats`: i see slime waving around the corner 2014-04-26T11:52:28Z stassats`: even no tests failed 2014-04-26T11:53:10Z stassats`: i'm a bit confused are you using arm-port-2 or the master branch on? 2014-04-26T11:53:20Z stassats`: i thought you switched yesterday to the master, now it's back onto arm-port-2 2014-04-26T11:53:26Z stassats`: or was that just a mistake? 2014-04-26T11:53:29Z nyef: Should be arm-port-2. 2014-04-26T11:53:38Z stassats`: (i switched to master too) 2014-04-26T11:53:41Z nyef: Pushing to master would have been a mistake, but an easy one to make. 2014-04-26T11:53:49Z nyef: (At least for me.) 2014-04-26T11:54:59Z nyef: Wow. I'm having trouble following the flow for composite-immediate-instruction. 2014-04-26T11:56:52Z stassats`: ok, now building with your changes 2014-04-26T11:59:14Z nyef: Looks like you should just be able to (composite-immediate-instruction dest nsp-tn delta) and it just work, unless someone plays sillybuggers with negative deltas, in which case you'd need :neg-op as well. 2014-04-26T11:59:42Z stassats`: that's what i did 2014-04-26T12:00:47Z stassats`: the only problem with getting that far is that rebuilding takes longer 2014-04-26T12:00:56Z stassats`: (than failing immediately at cold-init) 2014-04-26T12:01:54Z nyef: Yeah, the penalty of not being able to slam changes, generally hot-patch things, and so on. The penalty of having a NEARLY working system. 2014-04-26T12:05:34Z stassats`: waiting for the contribs to build and then trying slime 2014-04-26T12:06:22Z stassats`: and pushed alloc-number-stack-space to https://github.com/stassats/sbcl-arm 2014-04-26T12:07:31Z stassats`: a faster cpu wouldn't hurt 2014-04-26T12:08:54Z stassats`: but i have two cores, parallel contrib building! 2014-04-26T12:09:21Z stassats`: failing contribs: sb-aclrepl sb-cover sb-introspect sb-md5 sb-posix sb-simple-streams sb-sprof 2014-04-26T12:09:41Z nyef: Hrm. That doesn't match my last list. 2014-04-26T12:11:52Z nyef: sb-posix passed on my last build, for example. 2014-04-26T12:12:13Z stassats`: it was something with open.1 2014-04-26T12:12:25Z stassats`: my os peculiarity, maybe 2014-04-26T12:12:29Z nyef: Maybe. 2014-04-26T12:15:56Z stassats`: connected to slime, not long till an error 2014-04-26T12:16:06Z nyef: Have you run into anywhere that the %NIP-VALUES thing causes problems? 2014-04-26T12:16:35Z stassats`: make-compiled-code-location The value -1757171084 is not of type (MOD 536870909). 2014-04-26T12:17:11Z nyef: I'm impressed, though, that you can even try to get SLIME running. 2014-04-26T12:18:43Z Hydan is now known as Hydan` 2014-04-26T12:22:51Z stassats`: i may have a too new slime 2014-04-26T12:27:33Z nyef: We're not far from attempting a forward-port. 2014-04-26T12:27:45Z stassats`: well, i can evaluate things 2014-04-26T12:27:59Z stassats`: but something upon start casues an error, which cannot be backtraced 2014-04-26T12:28:44Z Hydan` quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-26T12:29:26Z stassats`: machine-type is undefined 2014-04-26T12:29:50Z stassats`: not that that's what fails, but i tried to evaluate it 2014-04-26T12:31:01Z stassats`: would ARM be a suitable machine-type? 2014-04-26T12:31:13Z stassats`: or we need more detail, like amr7abcdefghz 2014-04-26T12:31:29Z nyef: ARM is probably good enough for now. 2014-04-26T12:32:27Z stassats`: pushed 2014-04-26T12:32:32Z nyef: There's a huge variation within processor families even on other architectures, but machine-type typically returns a constant string. 2014-04-26T12:32:51Z stassats`: don't want to rebuild, let's hot-patch 2014-04-26T12:33:58Z stassats`: i can imagine slime wanting to call machine-type 2014-04-26T12:35:05Z stassats`: that was it 2014-04-26T12:35:28Z stassats`: that was an accidental fix 2014-04-26T12:36:24Z stassats`: we have a slime off 2014-04-26T12:37:46Z stassats`: http://i.imgur.com/DiXl3i0.png 2014-04-26T12:38:35Z nyef: I'm building contribs now. 2014-04-26T12:39:06Z nyef: With your number-stack allocation fixes, not with machine-type. 2014-04-26T12:40:03Z antoszka: wow 2014-04-26T12:40:07Z stassats`: undefined function error was causing that type error 2014-04-26T12:40:48Z stassats`: but C-x C-e worked, and i wanted to show that at least something worked, and tried (machine-type), and bingo, without investigating anything 2014-04-26T12:53:12Z oleo quit (Read error: Operation timed out) 2014-04-26T12:53:39Z oleo joined #sbcl 2014-04-26T12:59:21Z nyef: Failed contribs: sb-aclrepl, sb-cover, sb-introspect, sb-simple-streams, sb-sprof 2014-04-26T13:01:09Z stassats`: fcntl.flock.1 failed, something with the FS, i guess 2014-04-26T13:01:11Z nyef: And sprof is failing because of a lack of stepping support. 2014-04-26T13:01:15Z stassats`: it's an android kernel, after all 2014-04-26T13:02:04Z stassats`: introspect too due to stepping 2014-04-26T13:02:36Z stassats`: sb-md5 fails here due to floats 2014-04-26T13:03:18Z nyef: Ooh. (without-gcing (purify)) shows a different failure mode. 2014-04-26T13:11:43Z stassats` quit (Ping timeout: 252 seconds) 2014-04-26T13:24:08Z attila_lendvai quit (Quit: Leaving.) 2014-04-26T13:27:27Z PuercoPop quit (Ping timeout: 265 seconds) 2014-04-26T13:27:39Z PuercoPop joined #sbcl 2014-04-26T13:42:02Z drmeister joined #sbcl 2014-04-26T13:53:32Z nyef: I'm fairly sure I have EQL fixed. 2014-04-26T14:28:35Z stassats` joined #sbcl 2014-04-26T14:31:14Z nyef: http://paste.lisp.org/display/142304 2014-04-26T14:31:48Z stassats`: at least it finished 2014-04-26T14:31:56Z nyef: Yes, that too. 2014-04-26T14:32:58Z nyef notes that it helps to spell "regster" with an i. 2014-04-26T14:43:12Z stassats`: do you have an idea what's wrong with disassemble? 2014-04-26T14:47:23Z nyef: Yes. 2014-04-26T14:55:42Z stassats`: and fcntl.flock.1 doesn't fail from the repl 2014-04-26T14:57:31Z stassats`: nyef: what does grep Hardware /proc/cpuinfo say for you? 2014-04-26T14:58:19Z stassats`: trying to make get-machine-version 2014-04-26T15:00:10Z stassats`: and grep grep Processor 2014-04-26T15:01:09Z drmeister quit (Remote host closed the connection) 2014-04-26T15:01:18Z stassats`: Hardware seems no good 2014-04-26T15:01:33Z stassats`: Processor it is then 2014-04-26T15:03:53Z stassats`: (machine-version) => "ARMv7 Processor rev 1 (v7l)" 2014-04-26T15:05:04Z stassats`: and pushed 2014-04-26T15:05:05Z nyef: http://paste.lisp.org/display/142304#1 2014-04-26T15:05:31Z stassats`: nyef: looks good 2014-04-26T15:17:55Z drmeister joined #sbcl 2014-04-26T15:19:00Z nyef: Also, http://paste.lisp.org/display/142304#2 2014-04-26T15:23:13Z nyef: Clearly the output needs a bit of improvement, but it's a start. (-: 2014-04-26T15:25:56Z stassats`: eql fails because we don't have generic-eql 2014-04-26T15:26:11Z nyef: Not quite. And I do have a commit to fix EQL in my local tree. 2014-04-26T15:27:40Z nyef: Pull now? 2014-04-26T15:32:54Z stassats`: what about self building? 2014-04-26T15:33:03Z stassats`: will probably take quite some time 2014-04-26T15:33:35Z nyef: It'd take time, but it'd make SLAM easier to use... 2014-04-26T15:34:28Z stassats`: working eql should resolve a number of test issues 2014-04-26T15:34:47Z nyef: I think I'm going to bring my BBB with me this week, ISTR it is either faster or has more RAM... Or both. 2014-04-26T15:43:32Z slyrus quit (Ping timeout: 252 seconds) 2014-04-26T15:46:06Z stassats`: introspect fails because of DEBUG 3 2014-04-26T15:46:27Z nyef: So, two places where we don't support stepping? 2014-04-26T15:46:48Z stassats`: i'm thinking maybe disabled insert-debug-catch for the meantime with debug 3 2014-04-26T15:47:33Z stassats`: or was it insert-step-conditions 2014-04-26T15:47:36Z nyef: You're just scared of messing with DEFINE-FULL-CALL, aren't you? (-: 2014-04-26T15:52:21Z stassats`: i'll commit the disablement unless you already have a fix 2014-04-26T15:52:35Z nyef: No fix yet. 2014-04-26T15:52:45Z nyef: Concentrating on getting the first part of the disassembler in shape. 2014-04-26T15:52:52Z stassats`: i'm just too lazy to touch test-passed for sb-introspect 2014-04-26T15:53:23Z stassats`: did you try running slime? 2014-04-26T15:53:38Z nyef: No, I'm too lazy to try and install slime. 2014-04-26T15:54:37Z stassats`: i just scpied my local version, and run-sbcl.sh --load start-swank.lisp 2014-04-26T15:54:54Z stassats`: only i also added :style nil to start-server 2014-04-26T15:55:04Z nyef: I'm too lazy to go LOOKING for a local version. Or to download one. 2014-04-26T15:55:25Z stassats`: maybe :fd-handler works too 2014-04-26T15:55:32Z stassats`: (the default) 2014-04-26T15:55:35Z stassats`: but NIL is simpler 2014-04-26T15:55:43Z nyef: Besides, I'm concentrating on stuff that requires a full build cycle, not stuff to test interactively much. (-: 2014-04-26T16:02:22Z stassats`: it still complains about insert-step-conditions even though DEBUG 3 no longer triggers it 2014-04-26T16:06:18Z Intensity joined #sbcl 2014-04-26T16:10:48Z stassats`: i may need a full build 2014-04-26T16:11:29Z stassats`: indeed 2014-04-26T16:12:25Z stassats`: or rather, touch the files querying for it 2014-04-26T16:13:19Z stassats`: pushed 2014-04-26T16:20:26Z stassats`: now only sb-cover sb-md5 sb-posix and sb-simple-streams are failing 2014-04-26T16:25:45Z nyef: ; 6D0: 20908BE2 ADD NFP, CFP, 32 2014-04-26T16:26:12Z nyef: Unfortunately, we also get crap like this: 2014-04-26T16:26:13Z nyef: ; 730: 0090A0E1 MOV NFP, R0, R0 LSL 0 2014-04-26T16:26:42Z stassats`: i can live with that 2014-04-26T16:26:58Z nyef: Only works for DP instructions so far, though. 2014-04-26T16:30:45Z stassats`: the bloody sb-posix doesn't fail when i run its tests 2014-04-26T16:31:16Z stassats`: make when launched from make-target-contrib, it does fail 2014-04-26T16:31:33Z nyef: Something slightly different about the environment? 2014-04-26T16:31:56Z stassats`: only the current directory may seem to affect it 2014-04-26T16:33:04Z stassats`: and sb-rt of course doesn't show the error, just a ~s printed condition 2014-04-26T16:33:37Z stassats`: because it's ignored 2014-04-26T16:36:10Z xymox quit (Ping timeout: 258 seconds) 2014-04-26T16:38:39Z stassats`: no, sb-rt is just that bad 2014-04-26T16:38:53Z stassats`: error opening "test-file": file exists 2014-04-26T16:40:38Z xymox joined #sbcl 2014-04-26T16:41:33Z stassats`: something had been broken the previous run, so it didn't get deleted when NLXing from W-O-F 2014-04-26T16:42:01Z stassats`: now failing, sb-cover, sb-md5 and sb-simple-streams. sb-md5 due to my floats 2014-04-26T16:42:41Z Hydan joined #sbcl 2014-04-26T16:44:07Z nyef: Might I suggest solving for your floats with a build-time conditional for now? make-config --with-arm-softfp? 2014-04-26T16:44:38Z stassats`: i need to implement it first 2014-04-26T16:44:56Z nyef: True. 2014-04-26T16:45:04Z stassats`: i procrastinate by fixing other stuff 2014-04-26T16:45:09Z nyef: Fair enough. 2014-04-26T16:45:33Z stassats`: so, forward porting, when, how? 2014-04-26T16:46:03Z stassats`: adding too much stuff outside of src/compiler/arm now will only increase the amount 2014-04-26T16:46:15Z nyef: When is after we've removed the last of the ARM-specific kludges from src/code/ and IR2TRAN, I think. 2014-04-26T16:46:35Z nyef: I've been working to decrease the overhead off-and-on. 2014-04-26T16:47:11Z stassats`: grar, grar, threads! 2014-04-26T16:47:21Z nyef: The two big pieces that I want to see fixed are float-SAPs and float-unboxed-structure-slots. 2014-04-26T16:47:40Z nyef: gencgc first, then we can discuss trying to scare up a TLS register. 2014-04-26T16:50:07Z stassats`: i can try building everything on the target 2014-04-26T16:51:00Z nyef: I'm just about out of time for now, but I should be back in a few hours. 2014-04-26T16:51:13Z stassats`: compiler.pure fails due to Don't know how to %NIP-VALUES 2014-04-26T16:51:27Z nyef: Wonderful! I've been waiting for that one. 2014-04-26T16:52:01Z stassats`: our stack analysis is broken anyhow 2014-04-26T16:52:04Z nyef: If you can make up a list of what needs fixing, and where to check if it fails, I'll see what I can do about it tonight. 2014-04-26T16:52:06Z nyef: It is? 2014-04-26T16:52:19Z stassats`: yeah, lot of tickets on launchpad 2014-04-26T16:52:30Z nyef: Oh, fun. 2014-04-26T16:53:02Z nyef: Oh, wait, that cluster of ones involving multiple-values and non-local-exit? 2014-04-26T16:53:09Z stassats`: i had a go at it a year a go, but came up short 2014-04-26T16:53:18Z stassats`: yeah 2014-04-26T16:53:30Z nyef: I came up short as well, the fix breaks the tail-position detector. 2014-04-26T16:53:36Z stassats`: well, that's basically what stack-analysis does, stack and NLX 2014-04-26T16:54:27Z nyef: And it's not the analysis that's broken, it's an illegal change to the IR1 graph somewhere along the line. 2014-04-26T16:54:28Z Hydan is now known as Hydan` 2014-04-26T16:55:57Z nyef: (As I said, I came up short because the fix breaks important further optimization, not because I couldn't figure out why the analysis was broken.) 2014-04-26T16:56:35Z nyef: Anyway, time's up. I'll be back in a few hours. 2014-04-26T16:56:45Z nyef quit (Quit: Bye all, back in a few hours.) 2014-04-26T17:00:58Z ltbarcly joined #sbcl 2014-04-26T17:08:42Z jhao joined #sbcl 2014-04-26T17:09:36Z slyrus joined #sbcl 2014-04-26T17:14:12Z krzysz00 joined #sbcl 2014-04-26T17:21:57Z erikvarga quit (Ping timeout: 252 seconds) 2014-04-26T17:23:42Z jaimef quit (Ping timeout: 258 seconds) 2014-04-26T17:31:01Z jaimef joined #sbcl 2014-04-26T17:45:55Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-26T17:51:05Z Hydan` is now known as Hydan 2014-04-26T17:55:47Z Hydan quit (Ping timeout: 255 seconds) 2014-04-26T18:00:05Z jhao quit (Ping timeout: 252 seconds) 2014-04-26T18:08:04Z jhao joined #sbcl 2014-04-26T18:13:20Z jhao quit (Ping timeout: 255 seconds) 2014-04-26T18:29:40Z ltbarcly joined #sbcl 2014-04-26T18:29:54Z ltbarcly quit (Client Quit) 2014-04-26T18:34:32Z drmeister quit (Read error: Connection reset by peer) 2014-04-26T18:34:53Z drmeister joined #sbcl 2014-04-26T18:38:53Z jhao joined #sbcl 2014-04-26T18:43:33Z ltbarcly joined #sbcl 2014-04-26T19:01:55Z jhao quit (Ping timeout: 265 seconds) 2014-04-26T19:17:10Z attila_lendvai joined #sbcl 2014-04-26T19:17:10Z attila_lendvai quit (Changing host) 2014-04-26T19:17:10Z attila_lendvai joined #sbcl 2014-04-26T19:17:46Z phf joined #sbcl 2014-04-26T19:30:56Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-26T19:39:29Z nyef joined #sbcl 2014-04-26T19:39:42Z nyef: Okay, I'm back. 2014-04-26T19:44:12Z stassats`: nyef: does (loop for d of-type double-float downfrom 10d0 to 0d0 by 2d0 collect d) finish for you? 2014-04-26T19:44:45Z nyef: ... Let me find the sdcard with my system image on it. 2014-04-26T19:46:58Z nyef: Killed after five GCs. I'm not waiting for it. 2014-04-26T19:47:09Z nyef: (That is, I killed it after five GCs.) 2014-04-26T19:48:29Z stassats`: (loop for d downfrom 2d0 to 0d0 by 2d0 do (print d)) it's always 2d0 2014-04-26T19:49:30Z nyef: Yeah, looks like something isn't quite right there. 2014-04-26T19:49:37Z stassats`: (funcall (compile nil `(lambda (x y) (declare (double-float x y)) (- x y))) 2d0 1d0) => 0.0d0 2014-04-26T19:49:54Z nyef: (- 2d0 2d0) => 2.0d0 2014-04-26T19:50:18Z nyef: Something with generic arithmetic? 2014-04-26T19:50:36Z stassats`: with non generic too 2014-04-26T19:50:55Z nyef: Ah, so I now see. 2014-04-26T19:50:57Z nyef: Hrm. 2014-04-26T19:51:39Z nyef: I'm... going to get back to what I was working on earlier. 2014-04-26T19:52:45Z nyef: I figure if I cover load/store, branch, traps, and maybe some of the floaty bits, it'll cover most of what we'll need to disassemble. 2014-04-26T19:55:42Z nyef: Would you like me to pull, merge, and push each time I get another chunk of the disassembler working, or do you want to wait until I have most of it done? 2014-04-26T19:56:17Z jhao joined #sbcl 2014-04-26T19:56:41Z stassats`: i can wait 2014-04-26T19:56:52Z stassats`: not much use of a partially working disassembler 2014-04-26T19:57:35Z nyef: Actually, given that it can't get badly confused about varying instruction lengths, a partially working disassembler can give you a good idea of what's going on if it covers a good chunk of the instructions, including branches. 2014-04-26T19:58:16Z stassats`: _i_ can still get confused 2014-04-26T19:58:33Z nyef: Heh. Fair enough. 2014-04-26T19:59:49Z stassats`: i can get trace files right on the device, from slime 2014-04-26T19:59:59Z nyef: There is that, I guess. 2014-04-26T20:01:05Z jhao quit (Ping timeout: 252 seconds) 2014-04-26T20:02:31Z stassats`: -/double-float is using fsubS, that should explain 2014-04-26T20:03:20Z nyef: Ouch. 2014-04-26T20:03:33Z stassats`: and it fixes it 2014-04-26T20:06:07Z stassats`: (eql 0d0 (- 1d0 1d0)) is NIL, though 2014-04-26T20:09:00Z stassats`: pushed the fsubd fix 2014-04-26T20:14:32Z nyef: The possibly-saving grace of the single-step-around trap thing is that it's a static symbol value being compared to null, so I can use an unboxed register. 2014-04-26T20:15:21Z nyef: On the other hand, I may not HAVE an unboxed register available at that point. 2014-04-26T20:15:40Z nyef: Oh! 2014-04-26T20:16:02Z nyef realizes that he doesn't need to do the main target-2 pass with what he's doing, he can skip straight to LOAD if he keeps the fasls. 2014-04-26T20:19:13Z stassats`: (eql nil a) does not work, because it passes NIL TN to CMP, and it doesn't like that 2014-04-26T20:19:41Z stassats`: i'm not sure if testing for (sc-is x null) and then replacing with nil-tn is the right thing 2014-04-26T20:20:13Z nyef: No, that's not quite right... An SC-case is. And it's NULL-TN. 2014-04-26T20:20:33Z stassats`: shouldn't CMP do that itself? 2014-04-26T20:20:51Z stassats`: why is the TN saying nil and not an NULL-TN already? 2014-04-26T20:21:31Z nyef: Hrm. 2014-04-26T20:21:44Z nyef: I don't know at this point. 2014-04-26T20:21:54Z nyef: If I DID know at some point, it's been long since paged out. 2014-04-26T20:22:14Z stassats`: ldr/str do check for nulls 2014-04-26T20:22:46Z stassats`: though (eq 'null (sc-name (tn-sc reg))) is a strange way to write (sc-is reg null) 2014-04-26T20:23:59Z nyef: Too many precedents, too easily found. 2014-04-26T20:24:03Z stassats`: i'll just turn nil-tn into a null-tn 2014-04-26T20:24:12Z stassats`: for now 2014-04-26T20:28:44Z stassats`: pushed that too 2014-04-26T20:29:50Z attila_lendvai quit (Quit: Leaving.) 2014-04-26T20:30:56Z stassats`: and there goes my another evening movie 2014-04-26T20:36:59Z nyef: I think I've figured out how to get rid of the LSL 0 when there's no shift to deal with, and to detect the RRX case. 2014-04-26T20:40:34Z stassats`: and mop.impure and seq.impure now pass 2014-04-26T20:41:27Z stassats`: bit-vector.impure-cload.lisp fails because SB-EXT:DYNAMIC-SPACE-SIZE is undefined on -gencgc 2014-04-26T20:43:40Z stassats`: (eql (the double-float x) (the double-float y)) is compiled into something doing double-float-high/low-bits and comparing them 2014-04-26T20:44:05Z stassats`: while (= (the double-float x) (the double-float y)) uses FCMPD 2014-04-26T20:44:26Z stassats`: is that due to -0d0 = 0d0? 2014-04-26T20:46:16Z Krystof: yes 2014-04-26T20:46:29Z Krystof: and NaNs 2014-04-26T20:46:32Z nyef: Oh, good. I didn't have an answer for that one. (-: 2014-04-26T20:46:46Z Krystof: I spent some of this evening installing debian on my BBB 2014-04-26T20:47:31Z stassats`: i don't see no more easy test failures 2014-04-26T20:48:03Z stassats`: i guess my todo list would be a) float ABI b) more modular arithmetic 2014-04-26T20:48:03Z nyef: stassats`: So... time for you to fix float FFI for softfp? 2014-04-26T20:48:35Z Krystof: which branch is the current fun happening on? 2014-04-26T20:48:40Z nyef: Okay, and my list is disassembler, stepping, then probably something else. 2014-04-26T20:48:47Z stassats`: i think that's for tomorrow, i spend too much time on this ARM thing already 2014-04-26T20:49:01Z nyef: Yeah, I had other plans for this past week myself. 2014-04-26T20:49:15Z nyef: (Plans that I can't really get back to for ANOTHER week, at that.) 2014-04-26T20:49:21Z stassats`: Krystof: http://repo.or.cz/w/sbcl/nyef.git/shortlog/refs/heads/arm-port-2 is the main one 2014-04-26T20:49:39Z stassats`: https://github.com/stassats/sbcl-arm has a couple of unmerged important fixes 2014-04-26T20:49:46Z krzysz00 quit (Ping timeout: 276 seconds) 2014-04-26T20:50:10Z nyef: I'll merge and push once I finish with this part of the disassembler. 2014-04-26T20:50:47Z stassats`: i now have only sb-cover and sb-md5 failing 2014-04-26T20:51:25Z nyef: And sb-cover is definitely *STEPPING*-related. 2014-04-26T20:51:35Z stassats`: sb-md5 is due to my floats being different 2014-04-26T20:51:42Z nyef: ... Seriously? 2014-04-26T20:51:49Z stassats`: tests 2014-04-26T20:51:51Z nyef: Ah. 2014-04-26T20:52:13Z stassats`: or not 2014-04-26T20:52:15Z nyef: We also still need float SAP operations and float unboxed struct slots. 2014-04-26T20:52:33Z nyef: ... And to figure out what's up with PURIFY. 2014-04-26T20:52:37Z stassats`: it uses (abs (sin ..)) to build a table 2014-04-26T20:52:49Z stassats`: and sin is a library function 2014-04-26T20:53:18Z nyef: And then we'll need someone to start putting together a soft-float implementation... 2014-04-26T20:53:39Z stassats`: all my targets have hard float, screw soft float 2014-04-26T20:54:16Z nyef: Yeah, soft-float should be stashed on the wishlist. 2014-04-26T20:54:48Z stassats`: my n810 only has vfp (and a broken screen too) 2014-04-26T20:55:36Z nyef: Add audit for VFPv2 instructions to the list? 2014-04-26T20:55:47Z nyef: I think I might only have VFP and not VFPv2 on my Pi. 2014-04-26T20:56:07Z stassats`: i seem to have vfp3 everywhere else 2014-04-26T21:02:23Z stassats`: the current iphone has some nice ARM hardware, too bad it's too closed down 2014-04-26T21:06:07Z stassats`: i may need to update my storage library to work better on SBCL arm 2014-04-26T21:08:05Z nyef: So, here's a crazy question: If we have a thread structure, can we move the first three arg-passing locations into it? 2014-04-26T21:11:07Z nyef: Krystof: In case you missed it, the amusement to be gained from the ARM interrupt handling turned out to be pushing the control stack pointer to a static symbol so that we could leave the NSP in a register. 2014-04-26T21:11:42Z Krystof: I saw! 2014-04-26T21:12:22Z Krystof: meanwhile, we haven't had an agreed fix for the COMPLEX regression, have we? 2014-04-26T21:12:48Z stassats`: the current patch is the way to fix it during the freeze, but i haven't review it 2014-04-26T21:13:03Z Krystof: ok. Tomorrow. 2014-04-26T21:13:15Z stassats`: i blame the arm port 2014-04-26T21:13:49Z nyef: So do I. If you hadn't kept hacking on it, I'd've taken a break already. d-: 2014-04-26T21:14:03Z nyef: (Damned positive feedback loops, right?) 2014-04-26T21:14:50Z stassats`: i joined at the right moment, most of the work done, but still plenty fun stuff to go around 2014-04-26T21:15:11Z nyef: How about, say, July or August for Itanium? 2014-04-26T21:15:33Z stassats`: i don't have a phone with an itanium cpu 2014-04-26T21:15:33Z nyef: (Okay, maybe later than that.) 2014-04-26T21:15:44Z nyef: Heh! 2014-04-26T21:15:57Z stassats`: and if i had, i hope it hadn't a broken screen to 2014-04-26T21:16:29Z nyef: With a bit of luck, I should have most of the bits for an SBCL build farm by next weekend. 2014-04-26T21:16:40Z nyef: Modulo a Linux kernel that works on my MIPS box. 2014-04-26T21:16:56Z stassats`: who will be for all the electricity? 2014-04-26T21:17:05Z stassats`: paying 2014-04-26T21:17:11Z nyef: For now, it's probably going to be offline most of the time. 2014-04-26T21:17:34Z stassats`: the gcc build farm has a nice and fast ppc box 2014-04-26T21:18:09Z nyef: I'm also thinking to maybe set up boinkmarks, which would want a lightly-loaded box. 2014-04-26T21:20:21Z stassats`: argh, i depend on sb-md5 somehow 2014-04-26T21:20:42Z stassats`: i can just accept its failure 2014-04-26T21:23:32Z nyef: ; D00: 0090A0E1 MOV NFP, R0 2014-04-26T21:23:43Z nyef: I think I finally got this bit working. 2014-04-26T21:24:44Z nyef: Hrm. 2014-04-26T21:26:04Z nyef: Except that the octets are backwards...? 2014-04-26T21:26:50Z nyef: Or is that expected on a little-endian system with a fixed instruction width? 2014-04-26T21:27:06Z nyef: Something for later, maybe. 2014-04-26T21:34:00Z scymtym_ joined #sbcl 2014-04-26T21:37:43Z stassats`: it takes 6 seconds to load a data file, compared to 0.041 seconds on the desktop 2014-04-26T21:38:04Z stassats`: run times consists of 5.295 seconds of GC time 2014-04-26T21:38:15Z stassats`: that explains why it's so slow 2014-04-26T21:38:28Z nyef: Yeah, working purify should help with that. 2014-04-26T21:40:03Z stassats`: but it loads the data correctly and saves too 2014-04-26T21:42:40Z stassats`: now everything i'm trying to do is working 2014-04-26T21:42:50Z stassats`: this things is too solid, let's start a new port 2014-04-26T21:43:06Z nyef: I told you, Itanium, in a couple of months. 2014-04-26T21:43:20Z stassats`: what about ppc64? 2014-04-26T21:43:29Z nyef: There's a huge backlog of work to do on PPC, SPARC, MIPS, and HPPA as well. 2014-04-26T21:44:31Z stassats`: there's a lot of work to do which doesn't include backends 2014-04-26T21:45:30Z stassats`: imagine how SBCL would look like if it were always developed at a pace such as this week 2014-04-26T21:46:08Z nyef: Sure, but this week was mostly easy and obvious problems. 2014-04-26T21:48:16Z stassats`: everything is obvious once you know what it is 2014-04-26T21:48:45Z nyef: Yes and no. 2014-04-26T21:48:55Z Hydan joined #sbcl 2014-04-26T21:55:49Z krzysz00 joined #sbcl 2014-04-26T21:56:15Z les joined #sbcl 2014-04-26T21:56:15Z les quit (Changing host) 2014-04-26T21:56:15Z les joined #sbcl 2014-04-26T21:56:32Z krzysz00 quit (Client Quit) 2014-04-26T22:04:34Z stassats` quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2014-04-26T22:12:51Z edgar-rft quit (Quit: lifetime corrupted because memory access abandoned) 2014-04-26T22:21:38Z les quit (Quit: leaving) 2014-04-26T22:28:52Z Hydan quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-26T22:34:43Z Hydan joined #sbcl 2014-04-26T22:34:56Z Hydan is now known as Hydan` 2014-04-26T22:34:57Z Hydan` is now known as Hydan 2014-04-26T22:41:30Z nyef: ; D74: L1: L2: L3: L4:00C04CE2 SUB NARGS, NARGS, 0 2014-04-26T22:41:37Z nyef: I'm thinking that this isn't quite right. 2014-04-26T22:42:28Z nyef: Especially since all four instructions referring to this location use L4. 2014-04-26T23:06:35Z ltbarcly joined #sbcl 2014-04-26T23:08:35Z Hydan is now known as Hydan` 2014-04-26T23:11:29Z cmack: so this is my cpuinfo: http://paste.lisp.org/display/142290#4 ... will this cause problems for 0b324659916a8e8bd794d47a0a5786fe988c711d ? I'm hoping it's not a bogus implementation. I'm still having no joy at make-target-2 step 2014-04-26T23:12:07Z nyef: Looks good. 2014-04-26T23:12:12Z samskulls joined #sbcl 2014-04-26T23:12:19Z nyef: Are you on a hard-float or softfp system? 2014-04-26T23:12:59Z cmack: I /think/ it's hard float; will double check 2014-04-26T23:13:27Z nyef: Also, that commit hash is SO several hours ago. 2014-04-26T23:14:05Z nyef: Pull stassats' repository, it's currently most advanced (I'm still working on my next commit series). 2014-04-26T23:15:05Z Hydan` quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-26T23:15:28Z cmack: I'm on your 0a6bc8e58ab4a1d5f300... I was just curious about the one above because it is "Processor"; mine is "processor"... and it doesn't have the model info 2014-04-26T23:15:46Z nyef: Ah... I don't know about that, then. 2014-04-26T23:16:21Z nyef: I'm sure stassats will have an opinion tomorrow. 2014-04-26T23:19:01Z cmack: alright, I'll check out his and test. It looked like he was working with a similar arm chip, so I'm doubly miffed where mine is failing. 2014-04-26T23:19:44Z nyef: When did you last try a build? 2014-04-26T23:20:34Z nyef: Oh, and if you're going to try and do a full make-target-2, first edit make-target-2-load.lisp and add :PURIFY NIL to the SAVE-LISP-AND-DIE call at the end. 2014-04-26T23:21:14Z cmack: well I'm doing what you suggested with the src/runtime/sbcl --core output/cold-sbcl.core --no-sysinit --no-userinit --lose-on-corruption 2014-04-26T23:21:36Z nyef: Okay, that's a good start. 2014-04-26T23:21:40Z cmack: I built a few minutes ago with your current HEAD 2014-04-26T23:23:23Z cmack: I get a lot of these messages, then segfault:http://paste.lisp.org/display/142290#5 2014-04-26T23:24:12Z nyef: Hrm. 2014-04-26T23:24:13Z cmack: last I mentioned, stassats thought maybe my sbcl.nm was out of sync... but I'm rsyncing everything between xc-host and target 2014-04-26T23:26:02Z nyef: In local-target-features.lisp-expr, do you see :big-endian or :little-endian? 2014-04-26T23:27:19Z cmack: heh I just put that in http://paste.lisp.org/+31SI/6 I see little-endian 2014-04-26T23:27:32Z nyef: Hrm. Okay, so that's not it... 2014-04-26T23:28:13Z nyef: Yeah, that's the same as my ltf. 2014-04-26T23:29:51Z nyef: Do you have gdb installed? 2014-04-26T23:30:12Z samskulls quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2014-04-26T23:31:12Z cmack: yes 2014-04-26T23:32:35Z nyef: Okay, gdb src/runtime/sbcl, set args --core output/cold-sbcl.core, handle SIGUSR1 nostop, break call_into_lisp, and run. 2014-04-26T23:39:30Z cmack: sorry, took a sec to get fresh stassats HEAD done, still having prob. Here is gdb output: 2014-04-26T23:39:32Z cmack: Breakpoint 1, call_into_lisp () at arm-assem.S:31 2014-04-26T23:39:33Z cmack: 31 ldr reg_NULL, .known_nil 2014-04-26T23:39:33Z cmack: 2014-04-26T23:41:58Z nyef: Okay, use "stepi" until you catch a signal or just after "add reg_PC, reg_CODE, #SIMPLE_FUN_CODE_OFFSET". 2014-04-26T23:44:43Z cmack: ok do you want that all in a paste? or just the last line? 2014-04-26T23:45:01Z nyef: Neither. 2014-04-26T23:45:17Z nyef: Do an "x/20i $pc" and paste that. 2014-04-26T23:46:41Z cmack: http://paste.lisp.org/display/142309 2014-04-26T23:48:22Z nyef: Hrm. Code generation is a bit iffy, optimization-wise, but otherwise this looks reasonable. 2014-04-26T23:49:25Z nyef: Try "run", and see what happens? 2014-04-26T23:49:51Z cmack: from beginning? 2014-04-26T23:49:56Z cmack: it asks 2014-04-26T23:49:58Z nyef: Sorry, good point. 2014-04-26T23:50:03Z nyef: I meant "cont", not "run". 2014-04-26T23:50:12Z nyef: I keep making that mistake, and I'm glad it asks. 2014-04-26T23:50:16Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-26T23:50:34Z cmack: Program received signal SIGSEGV, Segmentation fault. 2014-04-26T23:50:35Z cmack: 0x8ffffffa in ?? () 2014-04-26T23:50:35Z cmack: 2014-04-26T23:50:51Z cmack: I said N to prompt, then cont... 2014-04-26T23:51:34Z cmack: and that appears none too helpful :( 2014-04-26T23:52:33Z nyef: Hrm. 2014-04-26T23:52:43Z nyef: info registers 2014-04-26T23:53:17Z nyef: Then... call lisp_backtrace(5) 2014-04-26T23:53:46Z cmack: http://paste.lisp.org/display/142309#1 doh...ok about to callback trace 2014-04-26T23:54:47Z cmack: annotated the above with backtrace 2014-04-26T23:54:57Z nyef: Ouch. 2014-04-26T23:55:22Z nyef: Also do a "disas funcall0" 2014-04-26T23:56:48Z cmack: annotated 2014-04-26T23:56:58Z nyef: Yeah, that'd do it. 2014-04-26T23:57:59Z cmack: gcc version just in case: gcc (Ubuntu/Linaro 4.8.1-10ubuntu9) 4.8.1 2014-04-26T23:57:59Z nyef: Now, how do we make this actually work...? 2014-04-27T00:01:00Z nyef: Back up to make-target-1 and invoke as "CFLAGS=-marm sh ./make-target-1.sh", copy back to the host, invoke "(. output/build-config; sh make-genesis-2.sh)", then copy back to the target and try again. 2014-04-27T00:06:26Z cmack: same behavior: Memory fault at 8ffffffa (pc=0x8ffffffa, sp=(nil)) 2014-04-27T00:06:29Z cmack: gdb? 2014-04-27T00:06:45Z nyef: Hrm. 2014-04-27T00:07:10Z nyef: Yeah, see if the address increment in "disas funcall0" is still 2 or if it's jumped to 4. 2014-04-27T00:09:50Z ltbarcly joined #sbcl 2014-04-27T00:10:54Z cmack: I see no difference 2014-04-27T00:12:08Z cmack: annotated just in case I missed something but it looks the same to my eyes 2014-04-27T00:13:29Z nyef: Yeah. Damn. 2014-04-27T00:13:55Z nyef: rm src/runtime/*.o and try the target-1 step again? 2014-04-27T00:14:13Z cmack: the cflags one? 2014-04-27T00:14:16Z nyef: Yeah. 2014-04-27T00:14:23Z nyef: Actually, hang on a sec, let me check something. 2014-04-27T00:15:00Z nyef: Damnit. That's not going to work. Edit src/runtime/Config.arm-linux and edit the CFLAGS= line there. 2014-04-27T00:16:22Z cmack: CFLAGS = -g -marm ? or no -g 2014-04-27T00:16:33Z nyef: Still -g. 2014-04-27T00:17:06Z cmack: ok. make-target-1? del the *.o files too? 2014-04-27T00:18:10Z nyef: Yeah, just in case. 2014-04-27T00:19:44Z cmack: whoa. 2014-04-27T00:19:54Z nyef: Different? 2014-04-27T00:19:57Z cmack: different output for sure.. 2014-04-27T00:20:04Z nyef: Increments by 4 instead of 2? 2014-04-27T00:20:04Z cmack: sb-show I think is working 2014-04-27T00:20:18Z cmack: I'm at a REPL 2014-04-27T00:20:51Z nyef: The hard parts were figuring out why it didn't work, and figuring out how to persuade the compiler to make the change required. 2014-04-27T00:21:42Z cmack: the puppet will have no comment :D 2014-04-27T00:21:57Z cmack: but thanks! 2014-04-27T00:22:24Z nyef: You don't happen to have a PPC or SPARC system over there, do you? 2014-04-27T00:22:44Z cmack: sadly, no. 2014-04-27T00:22:51Z nyef: Okay, fair enough. 2014-04-27T00:23:04Z nyef: I have a SPARC here, but I managed to forget to bring a power cord for it. 2014-04-27T00:24:26Z nyef: Worst-case scenario, I have to wait a week before I can run the test that I want. 2014-04-27T00:25:39Z cmack: so was this a vagary of my particular arm / gcc or is something that needs to be set explicitly going forward? 2014-04-27T00:26:07Z nyef: We're probably going to have to do something about thumb interworking going forward. 2014-04-27T00:27:39Z cmack: ah it was thumb? 2014-04-27T00:28:04Z nyef: Yeah, gcc on armv7 defaults to thumb, but SBCL can't interwork with it yet. 2014-04-27T00:41:12Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-27T01:00:38Z zRecursive joined #sbcl 2014-04-27T01:02:41Z eudoxia joined #sbcl 2014-04-27T01:03:53Z krzysz00 joined #sbcl 2014-04-27T01:07:12Z krzysz00 quit (Read error: Connection reset by peer) 2014-04-27T01:07:52Z krzysz00 joined #sbcl 2014-04-27T01:08:22Z slyrus_ joined #sbcl 2014-04-27T01:08:27Z slyrus quit (Ping timeout: 276 seconds) 2014-04-27T01:10:18Z slyrus_ is now known as slyrus 2014-04-27T01:13:49Z slyrus_ joined #sbcl 2014-04-27T01:14:59Z slyrus quit (Ping timeout: 255 seconds) 2014-04-27T01:15:13Z slyrus_ is now known as slyrus 2014-04-27T01:31:32Z nyef quit (Quit: G'night a..) 2014-04-27T01:34:28Z oleo quit (Ping timeout: 276 seconds) 2014-04-27T01:45:18Z slyrus quit (Ping timeout: 240 seconds) 2014-04-27T01:45:22Z slyrus_ joined #sbcl 2014-04-27T01:47:22Z slyrus_ is now known as slyrus 2014-04-27T01:47:43Z oleo joined #sbcl 2014-04-27T01:50:00Z ltbarcly joined #sbcl 2014-04-27T02:38:10Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-27T02:38:40Z christoph_debian quit (Ping timeout: 265 seconds) 2014-04-27T02:45:41Z pranavrc joined #sbcl 2014-04-27T02:45:41Z pranavrc quit (Changing host) 2014-04-27T02:45:41Z pranavrc joined #sbcl 2014-04-27T02:51:59Z christoph_debian joined #sbcl 2014-04-27T03:14:03Z zRecursive quit (Remote host closed the connection) 2014-04-27T03:28:42Z attila_lendvai joined #sbcl 2014-04-27T03:28:43Z attila_lendvai quit (Changing host) 2014-04-27T03:28:43Z attila_lendvai joined #sbcl 2014-04-27T03:33:58Z tmh_ quit (Ping timeout: 240 seconds) 2014-04-27T03:36:46Z eudoxia quit (Quit: leaving) 2014-04-27T04:12:03Z tmh_ joined #sbcl 2014-04-27T04:24:37Z asedeno quit (Ping timeout: 252 seconds) 2014-04-27T04:26:25Z asedeno joined #sbcl 2014-04-27T05:24:27Z pranavrc quit (Ping timeout: 265 seconds) 2014-04-27T05:28:56Z Intensity quit (Ping timeout: 246 seconds) 2014-04-27T05:40:29Z Hydan joined #sbcl 2014-04-27T06:03:53Z krzysz00 quit (Ping timeout: 252 seconds) 2014-04-27T06:46:44Z zRecursive joined #sbcl 2014-04-27T07:05:29Z asedeno quit (Ping timeout: 252 seconds) 2014-04-27T07:07:35Z asedeno joined #sbcl 2014-04-27T07:34:18Z asedeno quit (Ping timeout: 240 seconds) 2014-04-27T07:36:30Z asedeno joined #sbcl 2014-04-27T07:41:57Z zRecursive quit (Remote host closed the connection) 2014-04-27T08:29:04Z stassats joined #sbcl 2014-04-27T08:41:05Z edgar-rft joined #sbcl 2014-04-27T08:42:48Z erikvarga joined #sbcl 2014-04-27T08:44:33Z yacks quit (Ping timeout: 265 seconds) 2014-04-27T08:46:08Z yacks joined #sbcl 2014-04-27T09:11:43Z attila_lendvai quit (Quit: Leaving.) 2014-04-27T09:55:37Z pranavrc joined #sbcl 2014-04-27T09:55:38Z pranavrc quit (Changing host) 2014-04-27T09:55:38Z pranavrc joined #sbcl 2014-04-27T10:01:02Z asedeno quit (Ping timeout: 255 seconds) 2014-04-27T10:01:18Z asedeno joined #sbcl 2014-04-27T10:10:21Z asedeno quit (Read error: Operation timed out) 2014-04-27T10:12:56Z asedeno joined #sbcl 2014-04-27T10:42:11Z asedeno quit (Ping timeout: 252 seconds) 2014-04-27T10:43:54Z asedeno joined #sbcl 2014-04-27T10:44:58Z attila_lendvai joined #sbcl 2014-04-27T10:44:58Z attila_lendvai quit (Changing host) 2014-04-27T10:44:58Z attila_lendvai joined #sbcl 2014-04-27T10:46:21Z edgar-rft quit (Ping timeout: 265 seconds) 2014-04-27T10:50:10Z edgar-rft joined #sbcl 2014-04-27T11:08:04Z krzysz00 joined #sbcl 2014-04-27T11:59:23Z pranavrc quit (Ping timeout: 252 seconds) 2014-04-27T12:19:32Z LiamH joined #sbcl 2014-04-27T12:21:03Z pranavrc joined #sbcl 2014-04-27T12:21:03Z pranavrc quit (Changing host) 2014-04-27T12:21:03Z pranavrc joined #sbcl 2014-04-27T12:34:04Z Krystof: stassats: thanks! 2014-04-27T12:50:01Z attila_lendvai quit (Quit: Leaving.) 2014-04-27T13:04:46Z kludge` quit (Quit: leaving) 2014-04-27T13:05:42Z kludge` joined #sbcl 2014-04-27T13:16:55Z stassats: passing doubles in two registers poses a bit of a problem 2014-04-27T13:17:10Z stassats: say i define a new storage class for that, how do i tell that it takes up two registers? 2014-04-27T13:17:45Z Krystof: can you not crib from x86 and its overlapping registers? 2014-04-27T13:19:20Z stassats: but aren't they taking up one physical register? 2014-04-27T13:20:24Z Krystof: I meant eax taking up ah and al 2014-04-27T13:21:07Z stassats: i got that part, but you mean that it prevents ah and al from being allocated to byte-regs? 2014-04-27T13:22:36Z Krystof: I don't know 2014-04-27T13:26:00Z stassats: i think i see what x86 does 2014-04-27T13:26:24Z stassats: but i can't apply it to arm because the registers i need to pair up are not sequential 2014-04-27T13:27:22Z Krystof: ok, next try 2014-04-27T13:27:38Z Krystof: some 32-bit architectures go through hoops to pass long longs (64-bits) in two 32-bit registers 2014-04-27T13:27:48Z Krystof: in src/compiler//c-call.lisp 2014-04-27T13:29:09Z stassats: at least all the pairs are the same, i.e. 0-1, 2-3, not 0-1, 1-2 2014-04-27T13:29:37Z stassats: except that it's not 0-1, but 7-12, 4-6 2014-04-27T13:34:04Z stassats: i guess i need to hack up the %funcall transform 2014-04-27T13:36:03Z Hydan quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-27T13:38:13Z Krystof: like the %alien-funcall in ppc/c-call.lisp? 2014-04-27T13:39:13Z stassats: right 2014-04-27T13:39:24Z stassats: everyone hacks %alien-funcall besides arm 2014-04-27T13:50:47Z Hydan joined #sbcl 2014-04-27T14:01:56Z Hydan quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-27T14:03:13Z stassats: maybe i can use new sc after all, i haven't tried it 2014-04-27T14:04:26Z stassats: or even without, ppc-darwin do have the same problem 2014-04-27T14:09:52Z pranavrc quit (Remote host closed the connection) 2014-04-27T14:28:29Z oleo quit (Ping timeout: 264 seconds) 2014-04-27T14:28:53Z oleo joined #sbcl 2014-04-27T14:33:59Z erikvarga left #sbcl 2014-04-27T15:08:07Z fe[nl]ix quit (Remote host closed the connection) 2014-04-27T15:08:07Z Blkt_ quit (Remote host closed the connection) 2014-04-27T15:08:16Z Blkt joined #sbcl 2014-04-27T15:08:18Z fe[nl]ix joined #sbcl 2014-04-27T15:19:55Z eudoxia joined #sbcl 2014-04-27T15:31:18Z pkhuong: new sc seems the like the simplest and correct way 2014-04-27T15:32:25Z pkhuong: now I want to try a new SC for double-word bignums 2014-04-27T15:40:53Z attila_lendvai joined #sbcl 2014-04-27T15:40:53Z attila_lendvai quit (Changing host) 2014-04-27T15:40:53Z attila_lendvai joined #sbcl 2014-04-27T15:40:58Z nyef joined #sbcl 2014-04-27T15:41:09Z nyef: G'morning all. 2014-04-27T15:42:23Z pkhuong: morning. Have to run, but do you see any reason why we can't have an SC of register pairs for double-wide bignums? 2014-04-27T15:43:05Z pkhuong: *double-word 2014-04-27T15:43:32Z nyef: Nothing immediately comes to mind, other than a possible shortage of SC numbers. 2014-04-27T15:44:00Z pkhuong: I took care of that for SSE stuff 2014-04-27T15:44:08Z pkhuong: we ran over the 32 SC limit. It's now 64. 2014-04-27T15:44:43Z pkhuong: which breaks half the platforms' debug stuff 2014-04-27T15:45:01Z nyef: Heh. 2014-04-27T15:45:11Z erikvarga joined #sbcl 2014-04-27T15:45:33Z nyef: What's it going to take to support, say AltiVec on PPC, or any other SIMD hardware? 2014-04-27T15:45:38Z pkhuong: not much. 2014-04-27T15:45:43Z nyef: Good. 2014-04-27T15:45:45Z pkhuong: Neon is weird 2014-04-27T15:45:49Z pkhuong: altivec would be an easy fit. 2014-04-27T15:45:55Z pkhuong: the hard part is the SC and type system stuff 2014-04-27T15:46:09Z pkhuong: you'll have completely different primitives and VOPs, but that's expected 2014-04-27T15:46:14Z pkhuong: ARM has weird stuff. 2014-04-27T15:46:39Z pkhuong: like array-addressable SIMD register file 2014-04-27T15:46:43Z nyef: Yeah, the first clue that ARM was weird was having the program counter as a GPR... 2014-04-27T15:47:09Z pkhuong: but you can probably pretend it's a regular register file and get most of the gains. 2014-04-27T15:47:36Z pkhuong: I don't know how we'd regalloc the full extenf of the thing anyway. 2014-04-27T15:48:43Z pkhuong: I think vfp is more "normal" 2014-04-27T15:48:58Z nyef: Yeah, not going there right now. 2014-04-27T15:49:16Z nyef: I'm sure somewhere along the line we'll have to support iwmmxt as well. 2014-04-27T15:49:21Z pkhuong: (or perhaps it's Neon's that normal and vfp that's weird... either way, I decided a while ago that the cross product of ARM configs was just too ugly a space to think about) 2014-04-27T15:49:33Z nyef: And FPA. 2014-04-27T15:49:43Z nyef: That it is. 2014-04-27T15:50:45Z |3b|: hmm, (defun foo() (loop do)) prints error, while (defmethod foo() (loop do)) ends up in debugger 2014-04-27T15:51:14Z stassats: defmethod invokes a code-walker 2014-04-27T15:55:16Z |3b| was confused that the defmethod behaved differently depending on whether i hit C-c C-c on it or C-M-x 2014-04-27T15:56:06Z stassats: this double passed in int registers is full of corner cases 2014-04-27T15:56:43Z nyef: softfp support? 2014-04-27T15:56:57Z stassats: yep 2014-04-27T15:57:08Z stassats: i got single float to work already 2014-04-27T15:59:24Z stassats: maybe i can claim the two registers in the move vop, with some cunning life-times 2014-04-27T16:09:57Z drmeister quit (Remote host closed the connection) 2014-04-27T16:11:29Z Intensity joined #sbcl 2014-04-27T16:11:41Z stassats: "no unique move-arg-vop for moves in SC DOUBLE-TO-UNSIGNED" 2014-04-27T16:14:33Z stassats: maybe i need a full build 2014-04-27T16:23:58Z stassats: ok, it's a :move-arg define-move-vop 2014-04-27T16:26:36Z stassats: nothing works 2014-04-27T16:26:46Z stassats: and i'm lost 2014-04-27T17:07:04Z ltbarcly joined #sbcl 2014-04-27T17:39:57Z les joined #sbcl 2014-04-27T17:39:57Z les quit (Changing host) 2014-04-27T17:39:57Z les joined #sbcl 2014-04-27T17:43:31Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-27T17:51:44Z les quit (Quit: Lost terminal) 2014-04-27T17:52:09Z les joined #sbcl 2014-04-27T17:52:09Z les quit (Client Quit) 2014-04-27T17:52:32Z les joined #sbcl 2014-04-27T17:52:33Z les quit (Changing host) 2014-04-27T17:52:33Z les joined #sbcl 2014-04-27T17:53:51Z les quit (Client Quit) 2014-04-27T17:55:09Z les joined #sbcl 2014-04-27T17:55:09Z les quit (Changing host) 2014-04-27T17:55:09Z les joined #sbcl 2014-04-27T17:58:30Z ltbarcly joined #sbcl 2014-04-27T18:01:05Z eudoxia quit (Quit: leaving) 2014-04-27T18:07:35Z erikvarga left #sbcl 2014-04-27T18:48:18Z reb quit (Ping timeout: 240 seconds) 2014-04-27T18:58:29Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-27T19:11:20Z stassats quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2014-04-27T19:38:55Z Hydan joined #sbcl 2014-04-27T19:46:04Z drmeister joined #sbcl 2014-04-27T20:40:03Z deepestthought42 joined #sbcl 2014-04-27T21:42:46Z attila_lendvai quit (Quit: Leaving.) 2014-04-27T21:52:21Z Xof joined #sbcl 2014-04-27T21:52:21Z ChanServ has set mode +o Xof 2014-04-27T23:18:35Z deepestthought42 quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2014-04-27T23:44:19Z zRecursive joined #sbcl 2014-04-28T00:21:37Z ltbarcly joined #sbcl 2014-04-28T00:24:27Z edgar-rft quit (Quit: memory access corrupted because all hope lost) 2014-04-28T00:32:55Z nyef quit (Quit: G'night all.) 2014-04-28T00:36:41Z jumper joined #sbcl 2014-04-28T00:40:38Z jumper: Hello 2014-04-28T00:41:21Z jumper: Do any of you know if progn is a keyword in sbcl? 2014-04-28T00:41:58Z Bike: a keyword in what sense? 2014-04-28T00:42:21Z jumper: A token of the language. Like print, print is a token 2014-04-28T00:43:11Z Bike: i think you should probably get a better idea of how lisp parses. like, see http://www.lispworks.com/documentation/lw51/CLHS/Body/02_.htm. and this sounds like a #lisp question, not an #sbcl question 2014-04-28T00:44:24Z jumper: thank you Bike. I will definitely check it out 2014-04-28T01:12:57Z jumper left #sbcl 2014-04-28T01:23:33Z ltbarcly_ joined #sbcl 2014-04-28T01:25:06Z ltbarcly quit (Ping timeout: 276 seconds) 2014-04-28T01:55:13Z ltbarcly_ is now known as ltbarcly 2014-04-28T02:22:29Z slyrus quit (Read error: Connection reset by peer) 2014-04-28T02:25:33Z slyrus joined #sbcl 2014-04-28T02:33:22Z slyrus quit (Read error: Connection reset by peer) 2014-04-28T02:34:18Z slyrus_ joined #sbcl 2014-04-28T02:36:19Z slyrus_ is now known as slyrus 2014-04-28T02:38:19Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-28T02:38:38Z ltbarcly joined #sbcl 2014-04-28T02:39:35Z christoph_debian quit (Ping timeout: 255 seconds) 2014-04-28T02:50:54Z scymtym_ quit (Ping timeout: 276 seconds) 2014-04-28T02:52:37Z christoph_debian joined #sbcl 2014-04-28T03:02:32Z ltbarcly quit (Ping timeout: 255 seconds) 2014-04-28T03:03:27Z ltbarcly joined #sbcl 2014-04-28T03:18:47Z oleo quit (Ping timeout: 252 seconds) 2014-04-28T03:19:08Z oleo joined #sbcl 2014-04-28T03:48:02Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-28T04:08:43Z ltbarcly joined #sbcl 2014-04-28T05:13:19Z LiamH quit (Quit: Leaving.) 2014-04-28T05:20:20Z ltbarcly_ joined #sbcl 2014-04-28T05:21:53Z ltbarcly quit (Ping timeout: 265 seconds) 2014-04-28T05:27:28Z attila_lendvai joined #sbcl 2014-04-28T05:27:28Z attila_lendvai quit (Changing host) 2014-04-28T05:27:28Z attila_lendvai joined #sbcl 2014-04-28T05:27:48Z pranavrc joined #sbcl 2014-04-28T05:27:48Z pranavrc quit (Changing host) 2014-04-28T05:27:48Z pranavrc joined #sbcl 2014-04-28T05:32:28Z oleo quit (Quit: Leaving) 2014-04-28T05:42:36Z ltbarcly_ quit (Quit: Computer has gone to sleep.) 2014-04-28T06:04:20Z krzysz00 quit (Ping timeout: 255 seconds) 2014-04-28T06:06:06Z ltbarcly joined #sbcl 2014-04-28T06:29:55Z rudi joined #sbcl 2014-04-28T06:55:44Z Hydan quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-28T06:55:55Z ltbarcly quit (Quit: Textual IRC Client: www.textualapp.com) 2014-04-28T06:59:51Z flip214 quit (Remote host closed the connection) 2014-04-28T07:03:48Z flip214 joined #sbcl 2014-04-28T07:03:48Z flip214 quit (Changing host) 2014-04-28T07:03:48Z flip214 joined #sbcl 2014-04-28T07:40:04Z Xof quit (Remote host closed the connection) 2014-04-28T07:46:02Z zRecursive quit (Remote host closed the connection) 2014-04-28T08:43:40Z momo-reina joined #sbcl 2014-04-28T08:50:15Z momo-reina quit (Remote host closed the connection) 2014-04-28T09:50:30Z jdz joined #sbcl 2014-04-28T10:52:42Z hlavaty joined #sbcl 2014-04-28T11:15:13Z krzysz00 joined #sbcl 2014-04-28T11:26:41Z jdz quit (Quit: Leaving...) 2014-04-28T11:28:55Z Krystof: ok, release probably made 2014-04-28T11:28:56Z Krystof: go! 2014-04-28T12:11:15Z attila_lendvai quit (Quit: Leaving.) 2014-04-28T12:45:30Z pranavrc quit 2014-04-28T12:49:18Z eudoxia joined #sbcl 2014-04-28T12:53:59Z nyef joined #sbcl 2014-04-28T12:54:41Z nyef: G'morning all. 2014-04-28T13:01:29Z drmeister quit (Remote host closed the connection) 2014-04-28T13:34:00Z stassats joined #sbcl 2014-04-28T13:46:12Z oleo joined #sbcl 2014-04-28T13:59:49Z segv- joined #sbcl 2014-04-28T14:21:46Z stassats quit (Ping timeout: 265 seconds) 2014-04-28T14:22:42Z rudi quit (Quit: Client exciting.) 2014-04-28T14:30:32Z brown joined #sbcl 2014-04-28T14:30:56Z brown is now known as Guest19089 2014-04-28T14:34:02Z Guest19089 is now known as reb 2014-04-28T14:38:03Z eudoxia quit (Quit: leaving) 2014-04-28T14:51:53Z drmeister joined #sbcl 2014-04-28T14:59:35Z drmeister quit (Remote host closed the connection) 2014-04-28T15:02:17Z stassats joined #sbcl 2014-04-28T15:35:37Z krzysz00 quit (Ping timeout: 252 seconds) 2014-04-28T15:48:52Z drmeister joined #sbcl 2014-04-28T15:55:32Z drmeister quit (Ping timeout: 265 seconds) 2014-04-28T16:00:48Z oleo quit (Read error: Operation timed out) 2014-04-28T16:01:00Z oleo joined #sbcl 2014-04-28T16:07:43Z drmeister joined #sbcl 2014-04-28T16:27:52Z edgar-rft joined #sbcl 2014-04-28T16:35:14Z Intensity quit (Ping timeout: 246 seconds) 2014-04-28T16:49:27Z cmack` joined #sbcl 2014-04-28T16:50:25Z ltbarcly joined #sbcl 2014-04-28T16:52:00Z cmack quit (Ping timeout: 276 seconds) 2014-04-28T16:53:06Z sdemarre joined #sbcl 2014-04-28T17:04:59Z drmeister quit (Quit: Leaving...) 2014-04-28T17:07:12Z drmeister joined #sbcl 2014-04-28T17:10:47Z Hydan joined #sbcl 2014-04-28T17:20:56Z segv- quit (Remote host closed the connection) 2014-04-28T17:21:57Z segv- joined #sbcl 2014-04-28T17:22:19Z slyrus quit (Ping timeout: 252 seconds) 2014-04-28T17:27:01Z segv- quit (Remote host closed the connection) 2014-04-28T17:27:48Z segv- joined #sbcl 2014-04-28T17:34:16Z segv- quit (Remote host closed the connection) 2014-04-28T17:40:09Z segv- joined #sbcl 2014-04-28T17:44:20Z segv- quit (Read error: Connection reset by peer) 2014-04-28T17:51:16Z segv- joined #sbcl 2014-04-28T18:07:30Z Intensity joined #sbcl 2014-04-28T18:31:59Z pentagram565465 joined #sbcl 2014-04-28T18:32:45Z attila_lendvai joined #sbcl 2014-04-28T18:32:45Z attila_lendvai quit (Changing host) 2014-04-28T18:32:45Z attila_lendvai joined #sbcl 2014-04-28T18:38:35Z attila_lendvai quit (Ping timeout: 252 seconds) 2014-04-28T18:40:47Z attila_lendvai joined #sbcl 2014-04-28T18:40:47Z attila_lendvai quit (Changing host) 2014-04-28T18:40:47Z attila_lendvai joined #sbcl 2014-04-28T18:44:11Z sdemarre quit (Read error: Operation timed out) 2014-04-28T19:15:34Z pentagram565465 quit (Quit: pentagram565465) 2014-04-28T19:16:04Z Hydan quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-28T19:17:34Z Hydan joined #sbcl 2014-04-28T19:18:39Z prxq joined #sbcl 2014-04-28T19:27:36Z pentagram565465 joined #sbcl 2014-04-28T19:32:00Z pentagram565465 quit (Ping timeout: 240 seconds) 2014-04-28T19:34:38Z stassats quit (Ping timeout: 240 seconds) 2014-04-28T19:39:01Z stassats joined #sbcl 2014-04-28T19:39:09Z stassats quit (Remote host closed the connection) 2014-04-28T20:07:38Z drmeister quit (Read error: Connection reset by peer) 2014-04-28T20:08:17Z drmeister joined #sbcl 2014-04-28T20:33:06Z fikusz quit (Ping timeout: 240 seconds) 2014-04-28T20:36:28Z fikusz joined #sbcl 2014-04-28T20:52:57Z Bike quit (Read error: Connection reset by peer) 2014-04-28T20:53:15Z Bike joined #sbcl 2014-04-28T21:10:49Z segv- quit (Remote host closed the connection) 2014-04-28T21:11:05Z attila_lendvai quit (Ping timeout: 255 seconds) 2014-04-28T21:22:16Z segv- joined #sbcl 2014-04-28T21:26:06Z drmeister quit (Remote host closed the connection) 2014-04-28T21:38:17Z drmeister joined #sbcl 2014-04-28T22:01:38Z drmeister quit (Read error: Connection reset by peer) 2014-04-28T22:04:29Z drmeister joined #sbcl 2014-04-28T22:11:23Z edgar-rft quit (Quit: session ended into perpetual bleeding) 2014-04-28T22:23:59Z eudoxia joined #sbcl 2014-04-28T22:29:42Z drmeister quit (Read error: Connection reset by peer) 2014-04-28T22:39:10Z segv- quit (Remote host closed the connection) 2014-04-28T22:45:14Z slyrus joined #sbcl 2014-04-28T22:53:29Z milosn_ joined #sbcl 2014-04-28T22:54:24Z milosn quit (Ping timeout: 240 seconds) 2014-04-28T23:22:04Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-28T23:30:54Z ltbarcly joined #sbcl 2014-04-28T23:34:45Z drmeister joined #sbcl 2014-04-28T23:41:02Z eudoxia quit (Quit: wow such dinnertime) 2014-04-28T23:43:08Z drmeister quit (Read error: Connection reset by peer) 2014-04-28T23:52:48Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-29T00:04:36Z eudoxia joined #sbcl 2014-04-29T00:05:45Z zRecursive joined #sbcl 2014-04-29T00:37:42Z eudoxia quit (Quit: Lost terminal) 2014-04-29T01:30:47Z Hydan_ joined #sbcl 2014-04-29T01:34:02Z Hydan quit (Ping timeout: 252 seconds) 2014-04-29T01:34:09Z prxq_ joined #sbcl 2014-04-29T01:37:28Z prxq quit (Ping timeout: 265 seconds) 2014-04-29T01:52:47Z nyef quit (Quit: G'night all.) 2014-04-29T02:15:56Z ltbarcly joined #sbcl 2014-04-29T02:38:38Z christoph_debian quit (Ping timeout: 240 seconds) 2014-04-29T02:52:13Z christoph_debian joined #sbcl 2014-04-29T03:02:21Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-29T03:19:40Z ehaliewicz joined #sbcl 2014-04-29T03:33:35Z yacks quit (Read error: Connection reset by peer) 2014-04-29T03:38:20Z yacks joined #sbcl 2014-04-29T04:37:53Z oleo quit (Read error: Operation timed out) 2014-04-29T04:38:23Z oleo joined #sbcl 2014-04-29T04:38:50Z attila_lendvai joined #sbcl 2014-04-29T04:38:50Z attila_lendvai quit (Changing host) 2014-04-29T04:38:50Z attila_lendvai joined #sbcl 2014-04-29T05:01:31Z oleo quit (Quit: Leaving) 2014-04-29T05:29:52Z _8hzp joined #sbcl 2014-04-29T05:31:12Z _8hzp` quit (Ping timeout: 240 seconds) 2014-04-29T05:32:06Z pranavrc joined #sbcl 2014-04-29T05:32:06Z pranavrc quit (Changing host) 2014-04-29T05:32:06Z pranavrc joined #sbcl 2014-04-29T05:33:59Z _8hzp quit (Ping timeout: 240 seconds) 2014-04-29T05:52:38Z drmeister joined #sbcl 2014-04-29T06:09:55Z ehaliewicz quit (Remote host closed the connection) 2014-04-29T07:25:42Z attila_lendvai quit (Quit: Leaving.) 2014-04-29T07:28:06Z attila_lendvai joined #sbcl 2014-04-29T07:28:06Z attila_lendvai quit (Changing host) 2014-04-29T07:28:06Z attila_lendvai joined #sbcl 2014-04-29T07:35:58Z attila_lendvai quit (Ping timeout: 240 seconds) 2014-04-29T07:40:33Z attila_lendvai joined #sbcl 2014-04-29T07:40:33Z attila_lendvai quit (Changing host) 2014-04-29T07:40:33Z attila_lendvai joined #sbcl 2014-04-29T07:43:57Z zRecursive quit (Remote host closed the connection) 2014-04-29T07:57:54Z erikvarga joined #sbcl 2014-04-29T08:23:31Z attila_lendvai quit (Quit: Leaving.) 2014-04-29T08:41:13Z edgar-rft joined #sbcl 2014-04-29T08:48:46Z DGASAU quit (Quit: reboot) 2014-04-29T08:58:19Z attila_lendvai joined #sbcl 2014-04-29T08:58:19Z attila_lendvai quit (Changing host) 2014-04-29T08:58:19Z attila_lendvai joined #sbcl 2014-04-29T08:59:04Z DGASAU joined #sbcl 2014-04-29T09:31:52Z kludge` quit (Ping timeout: 252 seconds) 2014-04-29T09:36:44Z kludge` joined #sbcl 2014-04-29T09:40:26Z erikvarga left #sbcl 2014-04-29T10:05:04Z ams quit (Read error: Connection reset by peer) 2014-04-29T10:09:38Z ams` joined #sbcl 2014-04-29T10:10:11Z ams` quit (Remote host closed the connection) 2014-04-29T10:11:02Z ams joined #sbcl 2014-04-29T11:27:26Z attila_lendvai quit (Ping timeout: 255 seconds) 2014-04-29T11:28:11Z attila_lendvai joined #sbcl 2014-04-29T11:28:12Z attila_lendvai quit (Changing host) 2014-04-29T11:28:12Z attila_lendvai joined #sbcl 2014-04-29T12:01:23Z jdz joined #sbcl 2014-04-29T12:01:49Z prxq_ is now known as prxq 2014-04-29T12:26:10Z drmeister quit (Remote host closed the connection) 2014-04-29T12:28:26Z drmeister joined #sbcl 2014-04-29T12:34:25Z drmeister quit (Remote host closed the connection) 2014-04-29T12:43:23Z pranavrc quit 2014-04-29T12:47:15Z nyef joined #sbcl 2014-04-29T12:47:24Z nyef: G'morning all. 2014-04-29T12:58:11Z stassats joined #sbcl 2014-04-29T12:58:16Z drmeister joined #sbcl 2014-04-29T12:58:33Z stassats: adding new storage classes isn't easy 2014-04-29T12:58:55Z eudoxia joined #sbcl 2014-04-29T12:59:01Z stassats: i guess i'll just add up enough different move vops until it compiles, then put correct code in them 2014-04-29T13:00:00Z stassats: i can also cheat and do what ppc-darwin does in %alien-funcall ir2 transform 2014-04-29T13:05:56Z psilord quit (Quit: Leaving.) 2014-04-29T13:08:46Z nyef: Yeah, getting a new SC going involves a bit of doing. You can somewhat see what's involved if you look back at when we enabled complex floats in registers. 2014-04-29T13:09:06Z grc joined #sbcl 2014-04-29T13:09:22Z nyef: But yeah, the IR2 transform is probably the better way to go, especially if it doesn't need a new SC. 2014-04-29T13:09:48Z stassats: i just need to stuff a double into two registers 2014-04-29T13:11:11Z stassats: new SC potentially doesn't require any changes outside of compiler/arm 2014-04-29T13:12:08Z stassats: i guess i'll try to do the transform hack and then, when it works, experiment with different approaches 2014-04-29T13:12:34Z stassats: nyef: any ARM news on your side? 2014-04-29T13:13:14Z nyef: Still slogging through the disassembler. 2014-04-29T13:14:04Z nyef: My SunBlade 150 just lost it's idprom within the past couple of weeks, so I haven't been able to cross-check the behavior of PURIFY against a known-working reference yet. 2014-04-29T13:15:45Z nyef: Disassembler is already ahead of Alpha in one respect, though: It can decode error-trap information. 2014-04-29T13:16:22Z stassats: how about 1.1.19 target for a public release? 2014-04-29T13:18:23Z drmeister quit (Read error: Connection reset by peer) 2014-04-29T13:19:37Z nyef: What's our current version? 2014-04-29T13:19:53Z nyef: Nevermind, I'll do a git pull. 2014-04-29T13:21:29Z nyef: Hrm. So, a forward-port and upstream-commit within the next couple of weeks? 2014-04-29T13:22:37Z grc quit (Remote host closed the connection) 2014-04-29T13:25:48Z nyef: Still outstanding, then: Purify, the disassembler, softfp calling convention, thumb interworking (especially for the runtime), stepping, trace :encapsulate nil, unboxed float structure slots, float SAP references. 2014-04-29T13:26:15Z nyef: Beyond that, I think we're down to nice-to-haves. 2014-04-29T13:26:44Z nyef: Sound about right? 2014-04-29T13:28:03Z nyef: Oh, and some auto-detection stuff in make-config to pick out default float calling convention. 2014-04-29T13:28:38Z nyef: ... and cleanup of various bits of dead code, stale comments, et cetera. 2014-04-29T13:31:24Z nyef notes that with his recent changes upstream, the cumulative ARM changes to compiler/generic/array disappear. 2014-04-29T13:32:42Z nyef: About the only thing that looks like a potential showstopper is purify, because I don't really know what's wrong with it. 2014-04-29T13:36:44Z stassats: most of the outstanding issues can be resolved after the merge 2014-04-29T13:37:10Z nyef: I suppose, though it'd be nice to nail them before. 2014-04-29T13:37:12Z stassats: and the platform table can cell be colored yellow 2014-04-29T13:37:33Z nyef: There's also the question of how much we rewrite (or outright discard) history during the merge. 2014-04-29T13:40:43Z segv- joined #sbcl 2014-04-29T13:43:56Z attila_lendvai quit (Quit: Leaving.) 2014-04-29T13:47:49Z oleo joined #sbcl 2014-04-29T13:49:18Z nyef: ; 0254: L20: L21: L22: L23: L24: L25: L26: L27: L28: L29: L30: L31: L32: L33: L34: L35: L36: L37: L38: L39: L40:00709BE5 LDR OCFP, [CFP] 2014-04-29T13:49:26Z nyef: Sometimes I wonder about this disassembler. 2014-04-29T13:49:50Z stassats: ppc has something like that too 2014-04-29T13:49:57Z stassats: i couldn't figure out why 2014-04-29T13:50:06Z nyef: I haven't noticed it on x86-64, and I couldn't figure out why not. 2014-04-29T13:50:14Z stassats: never seen on x86oids 2014-04-29T13:50:17Z nyef: The why is easy, the why not isn't. 2014-04-29T13:53:17Z drmeister joined #sbcl 2014-04-29T13:53:19Z nyef: Oh, and there's this quirk with ARM that makes disassembly just that extra bit more obnoxious: The condition code is sometimes placed INSIDE the instruction mnemonic. 2014-04-29T13:53:47Z nyef: And I haven't figured out how to tell the disassembler that a given arg type (or field) may have any value OTHER than 15. 2014-04-29T13:57:39Z Krystof: has it built itself yet? 2014-04-29T13:57:55Z nyef: That's a good question, actually. 2014-04-29T13:58:05Z nyef: stassats: Have you tried a native build yet? 2014-04-29T13:59:08Z stassats: not yet 2014-04-29T13:59:15Z stassats: i'll try after the floats 2014-04-29T13:59:23Z nyef: I guess I'll give that a try next, then. 2014-04-29T14:12:33Z psilord joined #sbcl 2014-04-29T14:17:55Z yacks quit (Read error: Connection reset by peer) 2014-04-29T14:27:30Z yacks joined #sbcl 2014-04-29T14:29:11Z michael_lee joined #sbcl 2014-04-29T14:47:28Z michael_lee quit (Remote host closed the connection) 2014-04-29T14:53:58Z jdz quit (Quit: Leaving...) 2014-04-29T14:54:14Z nyef: Trying a full build now. 2014-04-29T14:56:08Z grc joined #sbcl 2014-04-29T14:56:52Z michael_lee joined #sbcl 2014-04-29T14:59:40Z foom: https://plus.sandbox.google.com/111282580643669107165/posts/Uwh9W3XiZTQ <- I wonder if sbcl runs on that. :P 2014-04-29T15:01:40Z stassats: i don't see why not, except 32-bit only 2014-04-29T15:02:33Z nyef: So... Think we could get ITA to switch platforms? 2014-04-29T15:03:08Z stassats: it's not like PPC support is better than Intel currently 2014-04-29T15:03:32Z nyef: Slightly better GC behavior, maybe... 2014-04-29T15:05:14Z jsnell_: any bets on whether running in little endian mode actually works out of the box? 2014-04-29T15:05:47Z stassats: are the instructions reversed too? 2014-04-29T15:08:11Z foom: QPX would (have to) switch platforms if most of the machines Google became Power. 2014-04-29T15:08:33Z foom: But, I don't actually know anything about this, except what is released externally. 2014-04-29T15:15:43Z nyef: jsnell_: No bet. It almost certainly won't for us. 2014-04-29T15:50:04Z nyef: Build failed, it ran to a certain point and then looped with a corruption-warning enough to push any useful information out of my scrollback buffer before I caught it. 2014-04-29T15:50:14Z nyef: Trying again with --lose-on-corruption for the build host. 2014-04-29T15:58:38Z stassats: i can now call things with doubles 2014-04-29T15:58:43Z stassats: not yet returning 2014-04-29T15:59:02Z nyef: Congratulations. 2014-04-29T15:59:23Z cracauer: how much is that POWER8 board? 2014-04-29T16:00:16Z stassats: the problem is that call_into_c doesn't expect there to be two return registers 2014-04-29T16:00:28Z stassats: i need to save r1 2014-04-29T16:01:28Z stassats: nl3 is free 2014-04-29T16:02:03Z erikvarga joined #sbcl 2014-04-29T16:03:05Z grc quit (Ping timeout: 252 seconds) 2014-04-29T16:06:31Z nyef: I might need to adjust call_into_c anyway so that we can interwork with a thumb runtime. 2014-04-29T16:09:11Z slyrus quit (Ping timeout: 240 seconds) 2014-04-29T16:10:54Z Krystof: erikvarga: hi! 2014-04-29T16:12:31Z erikvarga: Hi, Krystof. You're the other GSoC student, if I'm correct? 2014-04-29T16:15:31Z nyef: I have a hard time imagining Krystof as a GSoC student. 2014-04-29T16:17:18Z nyef: Hrm. I wonder if it just runs out of heap or something, and the overflow behavior is simply that bad. 2014-04-29T16:17:36Z nyef: We might need working purify before we can do a native build. 2014-04-29T16:18:04Z nyef: (My build is currently in host-1, compiler/target/insts.) 2014-04-29T16:18:27Z erikvarga: Oh, sorry then. I just thought so because the name of the other student was Krzysztof Drewniak, and your username seemed similar. 2014-04-29T16:19:31Z nyef: ... If we had Mr. Crampsie in here, maybe the three of us (Krystof, him, and myself) could pretend? 2014-04-29T16:20:02Z stassats: nyet 2014-04-29T16:20:38Z jsnell_: erikvarga: Krystof is Christophe, i.e. your gsoc mentor ;-) 2014-04-29T16:20:40Z jsnell_: but welcome 2014-04-29T16:20:49Z mood: I believe the nick of the "other GSoC student" is krzysz200 2014-04-29T16:22:51Z Krystof: yeah, it's going to be confusing 2014-04-29T16:22:53Z Krystof: sorry 2014-04-29T16:23:27Z erikvarga: Seems I need to learn which username belongs to who :) 2014-04-29T16:23:55Z stassats: at least they're pronounced differently 2014-04-29T16:24:05Z Krystof: I hope that Tom Emerson (the mentor for the other project) will hang out here too 2014-04-29T16:24:17Z Krystof: also, advance warning: I might well end up being `Xof' some of the time 2014-04-29T16:24:21Z Krystof: you'll get used to it :) 2014-04-29T16:25:37Z drmeiste_ joined #sbcl 2014-04-29T16:26:17Z drmeister quit (Read error: Connection reset by peer) 2014-04-29T16:26:48Z stassats: a tn with primitive-type NIL, what does that mean? 2014-04-29T16:27:24Z nyef: ... that something is messed up, or that it's a MAKE-RANDOM-TN ? 2014-04-29T16:29:32Z nyef: I'm beginning to think that we should pass all integer arguments on the stack and load the appropriate set up in call_into_c, allowing us to pass the function to call in an unboxed-reg. 2014-04-29T16:29:42Z nyef: Err... non-descriptor-reg. 2014-04-29T16:29:57Z nyef: erikvarga: So, which project are you working on? 2014-04-29T16:31:21Z erikvarga: I'll be working on speeding up division by constants. 2014-04-29T16:31:29Z nyef: Oh, cool! 2014-04-29T16:31:50Z nyef: That's the one that intersects nicely with what I've been doing recently, isn't it? 2014-04-29T16:32:16Z nyef: (ARM has no integer division instruction.) 2014-04-29T16:34:35Z erikvarga: Then I hope I'll do a good job :) 2014-04-29T16:43:24Z nyef: Oh, hell. Purify again. 2014-04-29T16:44:01Z nyef: Host-1 builds through to the end of pcl/walk and then runs purify. 2014-04-29T16:44:42Z stassats: nyef: what does gcc -v show you for Target: 2014-04-29T16:45:36Z nyef: arm-linux-gnueabihf. 2014-04-29T16:45:50Z nyef: ISTR going slightly out of my way to get a hard-float system, at that. 2014-04-29T16:45:54Z stassats: that's what i expected 2014-04-29T16:48:39Z stassats: nyef: and echo __SOFTFP__ | cpp 2014-04-29T16:49:15Z nyef: A bunch of # 1 "" lines followed by __SOFTFP__ 2014-04-29T16:49:19Z stassats: ok, good 2014-04-29T16:49:25Z stassats: i think that's the test 2014-04-29T16:51:37Z michael_lee quit (Quit: Ex-Chat) 2014-04-29T16:52:58Z nyef tries Yet Another full build, and notes that purify damage apparently trumps --lose-on-corruption. 2014-04-29T16:57:39Z erikvarga left #sbcl 2014-04-29T16:58:27Z Hydan_ quit (Quit: Textual IRC Client: www.textualapp.com) 2014-04-29T17:01:45Z stassats: still returning some garbage 2014-04-29T17:01:57Z stassats: i think i forgot to sync the assem file 2014-04-29T17:08:22Z eudoxia quit (Ping timeout: 265 seconds) 2014-04-29T17:09:49Z stassats: still doesn't work, though in a different fashion 2014-04-29T17:13:13Z eudoxia joined #sbcl 2014-04-29T17:14:40Z stassats: or not 2014-04-29T17:15:58Z ASau joined #sbcl 2014-04-29T17:16:00Z stassats: no, broken definitely 2014-04-29T17:19:44Z stassats: NARGS appears to have the wrong value 2014-04-29T17:29:14Z oleo quit (Ping timeout: 255 seconds) 2014-04-29T17:29:37Z oleo joined #sbcl 2014-04-29T17:36:31Z stassats: i don't get it anymore 2014-04-29T17:37:00Z stassats: if i switch input low-bits with high-bits, i get the correct result on the lisp side, but the printf prints the wrong input value 2014-04-29T17:37:14Z stassats: and if i switch them around, i get the right pritnfed value and the wrong result 2014-04-29T17:37:19Z stassats: doens't make any sense 2014-04-29T17:39:28Z drmeister joined #sbcl 2014-04-29T17:40:09Z ASau quit (Remote host closed the connection) 2014-04-29T17:41:59Z drmeiste_ quit (Ping timeout: 240 seconds) 2014-04-29T17:44:57Z ASau joined #sbcl 2014-04-29T17:50:33Z ltbarcly joined #sbcl 2014-04-29T17:56:45Z drmeister quit (Ping timeout: 252 seconds) 2014-04-29T17:56:55Z stassats: and things like atan and sqrt work correctly 2014-04-29T17:56:59Z stassats: maybe pritnf is just broken 2014-04-29T18:03:32Z Krystof: printf has an odd calling convention, remember 2014-04-29T18:04:16Z stassats: i'm calling it from C 2014-04-29T18:04:28Z Krystof: oh 2014-04-29T18:04:49Z stassats: minion: advice on compiler 2014-04-29T18:04:49Z minion: #11907: Looking for a compiler bug is the strategy of LAST resort. LAST resort. 2014-04-29T18:07:29Z ltbarcly quit (Ping timeout: 264 seconds) 2014-04-29T18:09:04Z nyef: That advice doesn't count if you're hacking a compiler backend in the first place. d-: 2014-04-29T18:09:36Z stassats: i applied that to gcc 2014-04-29T18:09:46Z nyef: Fair enough. 2014-04-29T18:09:58Z ltbarcly joined #sbcl 2014-04-29T18:16:11Z nyef: Well, it's really slow so far, but it's chugging through, cross-compiling code/print already. 2014-04-29T18:16:14Z slyrus joined #sbcl 2014-04-29T18:16:41Z nyef: Thus far the only big gotcha has been the call to PURIFY. 2014-04-29T18:17:20Z drmeister joined #sbcl 2014-04-29T18:28:31Z pkhuong: stassats: re new SC, the SSE changes are probably the simplest source. 2014-04-29T18:31:19Z pkhuong: chugging through a self build? Amazing progress. 2014-04-29T18:33:30Z nyef: Could probably have done this over the weekend, were I not so all-over-the-place. 2014-04-29T18:37:28Z stassats: ok, got to generate :arm-softfp by make-config 2014-04-29T18:38:25Z stassats: (sqrt 9) => 3.0 2014-04-29T18:38:31Z stassats: sounds right 2014-04-29T18:45:59Z stassats: i think i can't try a self build because i don't have enough space 2014-04-29T18:46:30Z nyef: Mmm. Android linux installers are confusing. /-: 2014-04-29T18:46:42Z stassats: tell me about it 2014-04-29T18:46:49Z stassats: that's why i have not enough space 2014-04-29T18:47:23Z nyef: And trying to get networking functional in your chroot? Ugh. 2014-04-29T18:51:04Z stassats: no contribs failed 2014-04-29T18:52:09Z nyef: Congratulations. 2014-04-29T18:52:34Z nyef: ... Does this mean that I might see that as well in an hour and a half or so? 2014-04-29T18:53:10Z stassats: some float tests still fail 2014-04-29T18:53:17Z yacks quit (Quit: Leaving) 2014-04-29T18:54:22Z stassats: pushed arm-softfp changes 2014-04-29T18:54:46Z stassats: (not really fond of them) 2014-04-29T18:55:36Z Krystof: http://permalink.gmane.org/gmane.lisp.openmcl.devel/8006 2014-04-29T18:55:40Z Krystof: might be interesting reading 2014-04-29T18:59:32Z eudoxia quit (Quit: leaving) 2014-04-29T19:13:20Z fikusz quit (Quit: Leaving) 2014-04-29T19:17:52Z attila_lendvai joined #sbcl 2014-04-29T19:17:52Z attila_lendvai quit (Changing host) 2014-04-29T19:17:53Z attila_lendvai joined #sbcl 2014-04-29T19:22:14Z fikusz joined #sbcl 2014-04-29T19:25:41Z slyrus_ joined #sbcl 2014-04-29T19:26:41Z slyrus quit (Ping timeout: 255 seconds) 2014-04-29T19:27:49Z slyrus_ is now known as slyrus 2014-04-29T19:40:28Z nyef: It's building the external-format code now. 2014-04-29T19:47:05Z irsol quit (Ping timeout: 264 seconds) 2014-04-29T19:48:45Z hzp joined #sbcl 2014-04-29T19:48:58Z irsol joined #sbcl 2014-04-29T19:50:44Z nyef: Oh! Did we ever do the character-comparison VOPs? 2014-04-29T19:51:17Z stassats: not really 2014-04-29T19:51:42Z stassats: not that they are necessary 2014-04-29T19:51:56Z nyef: ... Yeah, I found the lisp-side implementation. They'd be a nice optimization, though. 2014-04-29T19:52:06Z stassats: that's true for many things 2014-04-29T19:52:14Z stassats: there's no * vop either 2014-04-29T19:53:22Z nyef: ... Is that why I can't find any use of MUL instructions in the disassembler? 2014-04-29T19:54:00Z stassats: bignum vops have some 2014-04-29T19:54:14Z stassats: umull 2014-04-29T19:54:20Z nyef: Ah, right. 2014-04-29T19:55:41Z drmeister quit (Remote host closed the connection) 2014-04-29T19:56:18Z fiveop joined #sbcl 2014-04-29T19:56:36Z nyef: Oh, good. fiveop is here. 2014-04-29T19:57:00Z fiveop: hi 2014-04-29T19:57:22Z fiveop: how is that good? 2014-04-29T19:57:28Z stassats: a simple define-binop doesn't completely work 2014-04-29T19:57:32Z stassats: (* 3 2) => 24 2014-04-29T19:57:50Z nyef: We still need someone to do the float SAP VOPs and the float unboxed slot VOPs. 2014-04-29T19:58:10Z nyef: stassats: Looks like you need to compensate for fixtags. 2014-04-29T19:58:52Z drmeister joined #sbcl 2014-04-29T19:58:57Z attila_lendvai quit (Quit: Leaving.) 2014-04-29T19:59:38Z fiveop: nyef: what distribution and what settings are you running on your pi, in particular with respect to memory management? 2014-04-29T20:00:04Z fiveop: I wanted to do something else. And now I boot up my pi again ... :) 2014-04-29T20:00:21Z fiveop: stassats: you maintain common-qt, right? 2014-04-29T20:00:27Z nyef: I seem to be running some form of raspbian, no particular extra settings. 2014-04-29T20:00:38Z nyef: Haven't even hit swap yet, TBH. 2014-04-29T20:01:24Z fiveop: I couldn't get past the very early stages without swap 2014-04-29T20:01:54Z nyef: proc/sys/vm/overcommit_memory is 0 for me. 2014-04-29T20:02:13Z stassats: and MUL can't shift 2014-04-29T20:02:44Z stassats: fiveop: i do 2014-04-29T20:02:58Z stassats: and there's no dash 2014-04-29T20:04:26Z nyef: I realized that ARM can do a great number of small integer constant multiplies with an instruction. Things like multiplication by 7 are easy, for example. 2014-04-29T20:04:37Z fiveop: stassats: would you consider checking for qmake-qt4 in addition to qmake, as you do for gmake and make? 2014-04-29T20:04:51Z nyef: (Any power of two, minus one, is easy.) 2014-04-29T20:05:25Z stassats: fiveop: i may 2014-04-29T20:05:44Z ltbarcly quit (Ping timeout: 252 seconds) 2014-04-29T20:07:35Z nyef: Into genesis. 2014-04-29T20:07:42Z nyef: (And, therefore, peak memory usage.) 2014-04-29T20:08:38Z ltbarcly joined #sbcl 2014-04-29T20:11:50Z ltbarcly_ joined #sbcl 2014-04-29T20:13:58Z ltbarcly quit (Ping timeout: 240 seconds) 2014-04-29T20:15:14Z stassats: why do binops have both arguments targeted to the same result? 2014-04-29T20:15:34Z nyef: That seems odd, but not quite likely to cause problems. 2014-04-29T20:16:42Z stassats: it causes problems to me for *-c 2014-04-29T20:17:29Z nyef: fiveop: Is your Pi a model A or a model B? 2014-04-29T20:17:40Z fiveop: I have a theory, why I run out of memory. Linux uses unused memory as a file system cache, I have a huge hd mounted via nfs when I'm running it. That might cause the problems :) 2014-04-29T20:17:42Z fiveop: It's a B. 2014-04-29T20:17:50Z nyef: Hrm. 2014-04-29T20:18:04Z fiveop: you both don't use nfs, right? 2014-04-29T20:18:06Z nyef: No, filesystem cache shrinks as memory demand increases. 2014-04-29T20:18:19Z nyef: I'm running purely off of an sdcard. 2014-04-29T20:18:30Z stassats: rsync/scp here 2014-04-29T20:18:32Z fiveop: but target-2 is running ON the nfs mount 2014-04-29T20:21:17Z stassats: okay, the problem is not two targets actually 2014-04-29T20:21:38Z stassats: it's just that i can't load an immediate into R, since it's mapped to X 2014-04-29T20:25:37Z nyef: In target-2. (-: 2014-04-29T20:28:12Z stassats: i can't seem to convince it to give me different registers for arg and result 2014-04-29T20:28:51Z stassats: :from :eval to the result helped 2014-04-29T20:28:56Z stassats: no idea what that means 2014-04-29T20:30:54Z nyef: Ugh, and my website is down, so the article I wrote on VOP lifetimes is gone. Maybe the wayback machine has it? 2014-04-29T20:31:13Z fiveop: When the NFS is not mounted, it runs without a problem. 2014-04-29T20:31:13Z fiveop: 2014-04-29T20:32:04Z fiveop: pkhuong explained it nicely 2014-04-29T20:32:21Z fiveop: /doing first GC 2014-04-29T20:32:22Z fiveop: make-target-2.sh: line 39: 785 Killed 2014-04-29T20:32:26Z fiveop: So this is where I'm at 2014-04-29T20:32:29Z fiveop: Is that expected? :) 2014-04-29T20:32:45Z nyef: Umm... No, when was the last time you did a pull? 2014-04-29T20:33:01Z fiveop: right about after I join the channel 2014-04-29T20:33:06Z nyef: Hrm. 2014-04-29T20:33:21Z fiveop: Hm 2014-04-29T20:33:24Z fiveop: oom killer again :/ 2014-04-29T20:33:25Z nyef: From where did you pull? 2014-04-29T20:33:30Z nyef: Ah. 2014-04-29T20:33:30Z fiveop: from your repo 2014-04-29T20:33:43Z nyef: stassats' is actually more current right now. 2014-04-29T20:33:54Z stassats: but not for make-target-2 to not get killed 2014-04-29T20:33:55Z nyef: By two or three commits, IIRC. 2014-04-29T20:34:14Z nyef: Yeah, we're both through target-2. 2014-04-29T20:34:32Z stassats: i'm hacking away in slime 2014-04-29T20:34:39Z nyef: Into target-contrib. 2014-04-29T20:35:44Z Krystof: if there's something like the EA hack you can crib from the x86 backend 2014-04-29T20:35:53Z Krystof: for multiplications 2014-04-29T20:36:33Z nyef: We have a barrel shifter on one ALU input, and both add and "reverse-subtract" instructions. 2014-04-29T20:36:34Z Krystof: I have to say I liked the umlal instruction 2014-04-29T20:36:46Z nyef: Any power of two, plus or minus one, we can hit. 2014-04-29T20:37:11Z stassats: i'm only slightly hampered by the lack of disassemble, but i need to see which VOPs are selected anywya 2014-04-29T20:37:32Z nyef: Umm... My disassembler stuff is mostly complete for the integer stuff. 2014-04-29T20:38:27Z nyef: Still needs CLZ, the "misc" load/store, and bl/blx before it's done for integers. 2014-04-29T20:38:57Z nyef: And it doesn't annotate much yet, just the error traps and possible lra headers (the latter because it's automatic). 2014-04-29T20:39:30Z nyef: fiveop: You're not running a full desktop environment on your Pi, are you? 2014-04-29T20:41:16Z fiveop: no 2014-04-29T20:41:32Z fiveop: It runs sshd and make-target-2 :) 2014-04-29T20:41:34Z fiveop: mainly 2014-04-29T20:42:06Z nyef: Mine runs sshd and the full build process, it seems. 2014-04-29T20:42:53Z nyef: Admittedly, it's still building contribs, but the mere fact that it got that far is a good sign. 2014-04-29T20:43:48Z fiveop: If any of you have enough knowledge of how the linux kernel behaves, feel free to comment on why it even happens: https://gist.github.com/anonymous/02564535f3cfaebeef2a 2014-04-29T20:44:01Z fiveop: As far as I can decifer there's plenty of free memory. 2014-04-29T20:44:32Z nyef: "decipher"? 2014-04-29T20:44:51Z fiveop: probably 2014-04-29T20:44:53Z fiveop: :) 2014-04-29T20:46:01Z Krystof: the point at which you do the first GC is the point at which a bunch of things stop being mmap()ed and are copied into non-file-backed memory 2014-04-29T20:47:10Z nyef: That's a fair point. 2014-04-29T20:48:29Z nyef: ... avahi-daemon? 2014-04-29T20:49:21Z nyef: And what's "haveged", other than a statement of some level of education certification? 2014-04-29T20:49:53Z fiveop: :) 2014-04-29T20:50:13Z fiveop: Since I have no idea, probably default things arch linux installs. 2014-04-29T20:50:30Z fiveop: avahi seems to about detecting things like printers on the network 2014-04-29T20:50:34Z nyef: So, an arch vs. raspbian issue? 2014-04-29T20:50:39Z nyef: Yeah, zeroconf. 2014-04-29T20:51:10Z fiveop: haveged is for entropy (I suppose they put it in, since most raspberries are run headless) 2014-04-29T20:51:48Z Krystof: does this still happen if you don't run cold-init on nfs? 2014-04-29T20:52:26Z fiveop: nfs is not in the picture any longer 2014-04-29T20:52:47Z fiveop: I rsynced the stuff and unmounted the nfs volume 2014-04-29T20:52:55Z stassats: got some modular * arithmetic going 2014-04-29T20:54:16Z nyef: Heh. A full build didn't even touch swap. 2014-04-29T20:54:39Z milosn_ quit (Quit: Lost terminal) 2014-04-29T20:55:03Z milosn joined #sbcl 2014-04-29T20:55:19Z fiveop: where do you have your swap? 2014-04-29T20:55:33Z fiveop: do you have a swap file? 2014-04-29T20:55:37Z Krystof: wait, why is fiveop's trying to fork()? 2014-04-29T20:55:38Z nyef: I don't know, this looks like it's a fairly default raspbian setup. 2014-04-29T20:57:50Z fiveop: Krystof: where do I fork? 2014-04-29T20:58:05Z Krystof: the stack trace shows do_fork() at the bottom 2014-04-29T20:58:11Z drmeister quit (Remote host closed the connection) 2014-04-29T20:58:13Z fiveop: thats from watch 2014-04-29T20:58:15Z nyef: Might be something else forking? 2014-04-29T20:58:40Z nyef: I'm not seeing watch in my process list either. 2014-04-29T20:59:27Z fiveop: that's because I ran watch in a second terminal to look at proc/meminfo 2014-04-29T20:59:34Z fiveop: but it happens either way ;) 2014-04-29T20:59:39Z nyef: Ahh. 2014-04-29T21:00:03Z fiveop: Whatever process requests memory at the time sbcl hits first gc, will trigger the oom killer 2014-04-29T21:00:48Z nyef: /build started: Tue Apr 29 16:53:16 UTC 2014 2014-04-29T21:00:49Z nyef: /build finished: Tue Apr 29 21:00:19 UTC 2014 2014-04-29T21:00:55Z nyef: All contribs passed. 2014-04-29T21:02:05Z stassats: that's some slow build 2014-04-29T21:02:20Z nyef: I'll at least partly blame GC time because we can't purify. 2014-04-29T21:02:44Z nyef: But yeah, I'm not quite impressed with the speed here, and may well dig out my BBB when I get back to the house. 2014-04-29T21:03:47Z drmeister joined #sbcl 2014-04-29T21:04:31Z stassats: more arithmetic VOPs should make things faster 2014-04-29T21:04:43Z nyef: So should D-X. 2014-04-29T21:04:49Z fiveop: D-X? 2014-04-29T21:04:54Z nyef: dynamic-extent. 2014-04-29T21:05:03Z stassats: what could MLA be used for? 2014-04-29T21:05:07Z nyef shuts down his PI. 2014-04-29T21:05:49Z stassats: can we really optimize (+ (* x y) z)? 2014-04-29T21:06:16Z fiveop: for sufficiently small x, y and z :) 2014-04-29T21:07:17Z stassats: some ir-snooping and rewriting, similar to modular arithmetic can achieve that 2014-04-29T21:09:45Z fiveop: what do you mean when you talk about modular arithmatic? You can't just talk about the mathematical concept of rings, when you use it in that context. 2014-04-29T21:10:12Z stassats: pushed * vops 2014-04-29T21:11:19Z edgar-rft quit (Quit: existence terminated into eternal something) 2014-04-29T21:11:21Z nyef: Didn't Krystof write a paper on modular arithmetic? Or am I misremembering? 2014-04-29T21:11:28Z Krystof: I half-wrote a paper about it 2014-04-29T21:11:29Z stassats: he did 2014-04-29T21:11:42Z Krystof: fiveop: we exploit the fact that the cpu does arithmetic mod 2^32 2014-04-29T21:11:47Z nyef: I remember it being the new big thing right about when I started using Lisp, and was the main reason I transitioned from CMUCL to SBCL. 2014-04-29T21:12:09Z Krystof: so if we do (logand #xffffffff) we can do that complicated arithmetic mod 2^32 2014-04-29T21:12:27Z Krystof: even if intermediate results would normally be bignums 2014-04-29T21:12:38Z nyef: Or with any smaller mask at the end, or up to mod 2^64 on suitable platforms. 2014-04-29T21:13:08Z stassats: x86 could do 8/16-bit modular arithmetic 2014-04-29T21:13:13Z fiveop: Krystof: depending on the 2014-04-29T21:13:17Z Krystof: yes 2014-04-29T21:13:31Z Krystof: you're not allowed to have high bits affecting lower bits 2014-04-29T21:14:08Z Krystof: but +, -, *, logical operations, left shifts are all ok for these purposes 2014-04-29T21:14:54Z fiveop: (/ (* 2 2^31) 2^31) :) 2014-04-29T21:15:21Z fiveop: oh 2014-04-29T21:15:26Z fiveop: you didn't mention / :) 2014-04-29T21:15:28Z fiveop: lets see 2014-04-29T21:15:44Z stassats: / produces ratios 2014-04-29T21:16:00Z stassats: truncate would be more suitable 2014-04-29T21:16:54Z Krystof: but no, truncate doesn't work because high bits in the operands affect low bits in the answer 2014-04-29T21:18:03Z Krystof: I'd like to play with arm-port-N tomorrow, if possible -- let me know whose branch is most useful in 12 hours' time or so 2014-04-29T21:18:34Z fiveop: target-2 is still running ;) 2014-04-29T21:18:44Z stassats: depends on what happens in those 12 hours, so far, mine is the currentest 2014-04-29T21:19:19Z nyef: I might do a merge-and-push of my disassembler bits tonight. 2014-04-29T21:20:45Z fiveop: nyef: where the 5+ hours just make-target-2? 2014-04-29T21:21:38Z nyef: No, that was from typing bash ./make.sh --xc-host="sh ../fiveop-sbcl-arm-port/run-sbcl.sh --disable-debugger --lose-on-corruption" 2014-04-29T21:22:48Z fiveop: so a full build from itself 2014-04-29T21:23:06Z nyef: Yup. 2014-04-29T21:24:04Z nyef: There are things that we can do to knock that time down a bit, and there are things that we'd like to do but have yet to figure out why they go wrong... 2014-04-29T21:24:54Z fiveop: threads! 2014-04-29T21:25:01Z fiveop: (won't help though :)) 2014-04-29T21:25:53Z drmeister quit (Remote host closed the connection) 2014-04-29T21:27:50Z nyef: I don't have an angle on threads for ARM yet. 2014-04-29T21:28:54Z nyef: Before we do threads, we need gencgc anyway. 2014-04-29T21:30:17Z Hydan joined #sbcl 2014-04-29T21:31:42Z Hydan quit (Client Quit) 2014-04-29T21:33:23Z fiveop: ##linux tells me to report a bug with archlinuxarm and #archlinux-arm tells me, they have no idea why it happens (from the kernel output) 2014-04-29T21:34:42Z drmeister joined #sbcl 2014-04-29T21:35:25Z fiveop: and make-target crashes, but this time not because of the oom killer 2014-04-29T21:35:39Z fiveop: [doing purification: roots handlers stack bindings static cleanup done] 2014-04-29T21:35:39Z fiveop: CORRUPTION WARNING in SBCL pid 8426: 2014-04-29T21:35:39Z fiveop: 2014-04-29T21:36:19Z stassats: use :purify nil 2014-04-29T21:36:41Z fiveop: where? 2014-04-29T21:36:55Z stassats: make-target-2-load.lisp 2014-04-29T21:38:04Z fiveop: in the sb-ext:gc call? 2014-04-29T21:38:26Z stassats: slad 2014-04-29T21:38:28Z ltbarcly_ quit (Quit: Computer has gone to sleep.) 2014-04-29T21:40:07Z fiveop: and I'll run it again tomorrow 2014-04-29T21:40:14Z fiveop: good night 2014-04-29T21:40:19Z fiveop: (or thursday, probably thursday) 2014-04-29T21:41:34Z fiveop: stassats: I'll write a patch for qmake/qmake-qt4, if you're willing to apply it 2014-04-29T21:41:38Z fiveop quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2014-04-29T21:42:41Z stassats: if you do, open a ticket on github 2014-04-29T21:43:22Z Hydan joined #sbcl 2014-04-29T21:44:58Z stassats: gah, i want to use a new feature for VOPs, and it's too new 2014-04-29T21:45:31Z stassats: (a satisfies type for constant args) 2014-04-29T21:45:49Z Bicyclidine joined #sbcl 2014-04-29T21:46:02Z nyef: Tomorrow, we can discuss forward-porting once more. 2014-04-29T21:46:26Z stassats: i added character vops 2014-04-29T21:46:37Z nyef: Cool. 2014-04-29T21:47:03Z stassats: except for the constant ones, without (satisfies char-encodable-as-immediate) it's pretty useless 2014-04-29T21:47:17Z nyef: base-char? 2014-04-29T21:47:58Z stassats: char-encodable-as-immediate covers more than base-char, and it can wait until after forward-porting 2014-04-29T21:49:14Z nyef: Okay, I'm going to head out, possibly get some more done on the disassembler. If I get the rest of the integer bits that we use done tonight I'll merge and push. 2014-04-29T21:49:35Z stassats: you can take my branch, it has MULs 2014-04-29T21:50:09Z nyef: I might, if I can't find anything in terms of bignum stuff. 2014-04-29T21:50:19Z stassats: bignum only uses umull 2014-04-29T21:50:47Z nyef: It counts, I just want to make sure that at least one instruction is decoded properly. 2014-04-29T21:51:17Z nyef: Anyway, I'm gone. I'll be back either later today, or tomorrow. 2014-04-29T21:51:22Z nyef quit (Quit: G'night all.) 2014-04-29T21:57:19Z stassats: also implemented branch-if 2014-04-29T21:57:31Z stassats: move-if needs to be implemented too 2014-04-29T22:08:16Z drmeister quit (Read error: Connection reset by peer) 2014-04-29T22:15:34Z drmeister joined #sbcl 2014-04-29T22:18:07Z eudoxia joined #sbcl 2014-04-29T22:37:15Z pkhuong: stassats: have you looked at FP comparisons already? 2014-04-29T22:37:35Z pkhuong: they tend to complexify flag specs 2014-04-29T22:37:51Z stassats quit (Ping timeout: 265 seconds) 2014-04-29T22:43:29Z drmeister quit (Ping timeout: 264 seconds) 2014-04-29T22:51:49Z psilord quit (Quit: Leaving.) 2014-04-29T22:54:25Z drmeister joined #sbcl 2014-04-29T22:57:57Z nukemdukey joined #sbcl 2014-04-29T22:58:15Z nukemdukey quit (Remote host closed the connection) 2014-04-29T22:58:15Z drmeister quit (Read error: Connection reset by peer) 2014-04-29T22:58:44Z drmeister joined #sbcl 2014-04-29T23:41:19Z psilord joined #sbcl 2014-04-29T23:46:44Z eudoxia quit (Quit: wow such dinnertime) 2014-04-29T23:52:44Z zRecursive joined #sbcl 2014-04-30T00:01:31Z eudoxia joined #sbcl 2014-04-30T00:14:28Z ltbarcly joined #sbcl 2014-04-30T00:19:17Z ltbarcly quit (Client Quit) 2014-04-30T00:27:06Z ltbarcly joined #sbcl 2014-04-30T00:27:46Z jhao joined #sbcl 2014-04-30T00:34:47Z slyrus quit (Ping timeout: 240 seconds) 2014-04-30T01:04:48Z pkhuong: not sure if I want to add arbitrary location lists to scs. 2014-04-30T01:11:42Z Bicyclidine quit (Ping timeout: 265 seconds) 2014-04-30T01:17:31Z eudoxia quit (Quit: leaving) 2014-04-30T01:29:59Z prxq_ joined #sbcl 2014-04-30T01:33:27Z prxq quit (Ping timeout: 265 seconds) 2014-04-30T01:40:59Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-30T01:54:47Z ltbarcly joined #sbcl 2014-04-30T02:16:24Z momo-reina joined #sbcl 2014-04-30T02:39:21Z christoph_debian quit (Ping timeout: 276 seconds) 2014-04-30T02:44:24Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-30T02:51:53Z christoph_debian joined #sbcl 2014-04-30T03:09:11Z momo-reina quit (Remote host closed the connection) 2014-04-30T03:22:38Z ltbarcly joined #sbcl 2014-04-30T03:27:25Z slyrus joined #sbcl 2014-04-30T04:01:39Z drmeister quit (Remote host closed the connection) 2014-04-30T04:16:13Z michael_lee joined #sbcl 2014-04-30T04:27:58Z jhao quit (Ping timeout: 240 seconds) 2014-04-30T04:42:14Z attila_lendvai joined #sbcl 2014-04-30T04:42:14Z attila_lendvai quit (Changing host) 2014-04-30T04:42:14Z attila_lendvai joined #sbcl 2014-04-30T04:55:23Z ltbarcly quit (Ping timeout: 252 seconds) 2014-04-30T05:05:22Z ltbarcly joined #sbcl 2014-04-30T05:08:16Z oleo quit (Quit: Leaving) 2014-04-30T05:12:05Z attila_lendvai quit (Quit: Leaving.) 2014-04-30T05:17:29Z pranavrc joined #sbcl 2014-04-30T05:17:30Z pranavrc quit (Changing host) 2014-04-30T05:17:30Z pranavrc joined #sbcl 2014-04-30T05:17:34Z Hydan quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-30T05:41:01Z sdemarre joined #sbcl 2014-04-30T05:50:33Z drmeister joined #sbcl 2014-04-30T05:51:06Z sdemarre quit (Ping timeout: 276 seconds) 2014-04-30T05:51:33Z prxq_ is now known as prxq 2014-04-30T05:54:47Z drmeister quit (Ping timeout: 240 seconds) 2014-04-30T07:21:56Z ASau quit (Ping timeout: 265 seconds) 2014-04-30T07:28:10Z zRecursive quit (Remote host closed the connection) 2014-04-30T07:39:26Z drmeister joined #sbcl 2014-04-30T07:44:10Z drmeister quit (Ping timeout: 265 seconds) 2014-04-30T08:02:05Z yacks joined #sbcl 2014-04-30T08:05:49Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-30T08:06:11Z ltbarcly joined #sbcl 2014-04-30T08:21:05Z kludge` quit (Ping timeout: 252 seconds) 2014-04-30T08:46:50Z kludge` joined #sbcl 2014-04-30T08:48:30Z attila_lendvai joined #sbcl 2014-04-30T08:48:31Z attila_lendvai quit (Changing host) 2014-04-30T08:48:31Z attila_lendvai joined #sbcl 2014-04-30T09:19:59Z segv- quit (Ping timeout: 240 seconds) 2014-04-30T09:20:43Z Hydan joined #sbcl 2014-04-30T09:21:33Z prxq quit (Quit: Leaving) 2014-04-30T09:27:17Z prxq joined #sbcl 2014-04-30T09:28:05Z drmeister joined #sbcl 2014-04-30T09:31:28Z kludge` quit (Ping timeout: 265 seconds) 2014-04-30T09:33:24Z drmeister quit (Ping timeout: 276 seconds) 2014-04-30T09:34:04Z attila_lendvai1 joined #sbcl 2014-04-30T09:34:04Z attila_lendvai quit (Disconnected by services) 2014-04-30T09:34:04Z attila_lendvai1 quit (Changing host) 2014-04-30T09:34:04Z attila_lendvai1 joined #sbcl 2014-04-30T09:34:48Z erikvarga joined #sbcl 2014-04-30T09:35:57Z kludge` joined #sbcl 2014-04-30T09:46:51Z stassats joined #sbcl 2014-04-30T10:00:11Z erikvarga left #sbcl 2014-04-30T10:12:16Z Hydan quit (Quit: My MacBook Pro has gone to sleep. ZZZzzz…) 2014-04-30T10:25:38Z grc joined #sbcl 2014-04-30T10:44:08Z attila_lendvai joined #sbcl 2014-04-30T10:44:08Z attila_lendvai quit (Changing host) 2014-04-30T10:44:08Z attila_lendvai joined #sbcl 2014-04-30T10:47:27Z attila_lendvai1 quit (Ping timeout: 245 seconds) 2014-04-30T11:06:39Z grc quit (Remote host closed the connection) 2014-04-30T11:12:49Z stassats quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2014-04-30T11:14:33Z stassats joined #sbcl 2014-04-30T11:16:50Z drmeister joined #sbcl 2014-04-30T11:21:22Z drmeister quit (Ping timeout: 252 seconds) 2014-04-30T11:23:14Z edgar-rft joined #sbcl 2014-04-30T11:57:18Z pranavrc quit (Remote host closed the connection) 2014-04-30T11:57:35Z pranavrc joined #sbcl 2014-04-30T11:57:36Z pranavrc quit (Changing host) 2014-04-30T11:57:36Z pranavrc joined #sbcl 2014-04-30T11:58:24Z stassats quit (Ping timeout: 252 seconds) 2014-04-30T12:00:21Z attila_lendvai1 joined #sbcl 2014-04-30T12:00:21Z attila_lendvai quit (Disconnected by services) 2014-04-30T12:00:21Z attila_lendvai1 quit (Changing host) 2014-04-30T12:00:21Z attila_lendvai1 joined #sbcl 2014-04-30T12:05:27Z drmeister joined #sbcl 2014-04-30T12:11:55Z drmeister quit (Remote host closed the connection) 2014-04-30T12:13:58Z flip214 quit (Ping timeout: 240 seconds) 2014-04-30T12:24:55Z DGASAU quit (Remote host closed the connection) 2014-04-30T12:24:56Z jdz joined #sbcl 2014-04-30T12:30:50Z drmeister joined #sbcl 2014-04-30T12:31:36Z DGASAU joined #sbcl 2014-04-30T12:32:16Z flip214 joined #sbcl 2014-04-30T12:32:16Z flip214 quit (Changing host) 2014-04-30T12:32:16Z flip214 joined #sbcl 2014-04-30T12:35:59Z drmeister quit (Remote host closed the connection) 2014-04-30T12:38:00Z attila_lendvai1 quit (Quit: Leaving.) 2014-04-30T12:52:38Z DGASAU quit (Ping timeout: 240 seconds) 2014-04-30T12:53:14Z eudoxia joined #sbcl 2014-04-30T12:53:47Z prxq_ joined #sbcl 2014-04-30T12:54:55Z DGASAU joined #sbcl 2014-04-30T12:55:37Z eudoxia_ joined #sbcl 2014-04-30T12:55:40Z eudoxia_ quit (Client Quit) 2014-04-30T12:58:20Z eudoxia quit (Ping timeout: 265 seconds) 2014-04-30T13:03:17Z michael_lee quit (Ping timeout: 264 seconds) 2014-04-30T13:09:06Z drmeister joined #sbcl 2014-04-30T13:15:01Z michael_lee joined #sbcl 2014-04-30T13:15:42Z pranavrc quit 2014-04-30T13:20:03Z drmeister quit (Read error: Connection reset by peer) 2014-04-30T13:26:42Z stassats joined #sbcl 2014-04-30T13:36:08Z oleo joined #sbcl 2014-04-30T13:37:49Z segv- joined #sbcl 2014-04-30T13:38:03Z prxq quit (Quit: Leaving) 2014-04-30T13:52:43Z Krystof: well, that wasn't a flawless success 2014-04-30T13:52:55Z Krystof: stassats: building from your arm tree, I seem to have got myself into infinite corruption in cold-init 2014-04-30T13:53:37Z stassats: i always passed --lose-on-corruption 2014-04-30T13:54:36Z Krystof: well that gets me single-corruption on cold-init 2014-04-30T13:54:47Z stassats: does it get you ldb? 2014-04-30T13:55:00Z Krystof: yes 2014-04-30T13:55:22Z Krystof: backtrace gets me more corruption 2014-04-30T13:55:30Z stassats: that's unfortunate 2014-04-30T13:55:30Z Krystof: Memory fault at fb6ffffe (pc=0xfb6ffffe, sp=(nil)) 2014-04-30T13:55:47Z Krystof: I can't debug this now :-( childcare beckons 2014-04-30T13:55:59Z Krystof: this is a Beaglebone Black running Debian 2014-04-30T13:56:06Z Krystof: I might be able to look at it some more this evening 2014-04-30T13:56:15Z stassats: armel or armhf? 2014-04-30T13:56:45Z Krystof: hf 2014-04-30T13:56:49Z Krystof: ah, but I have an idea what it might be 2014-04-30T13:56:58Z Krystof: I bet half of this was built with :sb-show and the other half not 2014-04-30T13:57:12Z Krystof: ok, childcare doesn't just beckon it says "don't be late to school" 2014-04-30T13:57:33Z drmeister joined #sbcl 2014-04-30T13:58:31Z Krystof: (not totally convinced by my hypothesis, but...) 2014-04-30T13:58:58Z stassats: ensuring that everything is synced correctly is a first step 2014-04-30T13:59:22Z stassats: i spent some time debugging a thing and i was putting the new core into the wrong directory 2014-04-30T14:00:08Z stassats: so that sbcl, sbcl.nm and sbcl-cold.core all match 2014-04-30T14:02:08Z drmeister quit (Ping timeout: 265 seconds) 2014-04-30T14:08:03Z kludge` quit (Remote host closed the connection) 2014-04-30T14:11:39Z kludge` joined #sbcl 2014-04-30T14:12:39Z drmeister joined #sbcl 2014-04-30T14:15:35Z Krystof: Breakpoint 2, debug_print (string=2824) at interr.c:228 2014-04-30T14:15:42Z Krystof: that is a very implausible looking string 2014-04-30T14:15:49Z Krystof: (or even lispobj) 2014-04-30T14:16:54Z Krystof: it is 0xb08 2014-04-30T14:27:27Z nyef joined #sbcl 2014-04-30T14:27:41Z nyef: G'morning all. 2014-04-30T14:28:52Z nyef: stassats: I can't quite tell, does your FFI change affect non-ARM platforms somehow? One of the hunks seems to mess with reader conditionals in a way that doesn't quite make sense to me. 2014-04-30T14:35:32Z stassats: it shouldn't 2014-04-30T14:35:35Z stassats: let me check 2014-04-30T14:36:00Z ltbarcly quit (Quit: Computer has gone to sleep.) 2014-04-30T14:36:52Z pranavrc joined #sbcl 2014-04-30T14:36:52Z pranavrc quit (Changing host) 2014-04-30T14:36:53Z pranavrc joined #sbcl 2014-04-30T14:36:56Z stassats: everything seems fine 2014-04-30T14:37:55Z stassats: or not 2014-04-30T14:38:09Z nyef: Which is it? 2014-04-30T14:38:25Z stassats: i removed the code, but left the reader conditional 2014-04-30T14:39:16Z nyef: That's what I thought I saw. 2014-04-30T14:40:24Z stassats: pushed 2014-04-30T14:42:25Z nyef: My working tree includes single-step changes, but I haven't the slightest idea of how to test them. 2014-04-30T14:43:37Z stassats: i actually have no idea why non-x86oids move things through a temporary tn, wouldn't the move arrange things like that? 2014-04-30T14:43:51Z nyef: And once more I find that the test suite contains an obnoxious number of bare asserts, making it more difficult than necessary to track down what's going wrong. /-: 2014-04-30T14:44:03Z nyef: I don't know either, TBH. 2014-04-30T14:44:28Z stassats: i can see the idea, RISC and all, but it shouldn't be needed 2014-04-30T14:44:39Z nyef: We can look into it later. 2014-04-30T14:44:55Z nyef: I'll try and shake commits out of my working tree later today, but I'm supposed to be working right now. /-: 2014-04-30T14:45:26Z stassats: it was that way since cmucl, apparently 2014-04-30T14:46:14Z nyef: There's a reason why I keep a cmucl git tree on my local disk. 2014-04-30T14:46:44Z stassats: i have it too 2014-04-30T14:50:11Z ltbarcly joined #sbcl 2014-04-30T14:50:40Z stassats: turns out, it was that way in cmucl, then they made so that x86 uses temps too 2014-04-30T14:51:40Z stassats: something about the temp-tn being on the stack 2014-04-30T15:16:53Z xymox quit (Ping timeout: 252 seconds) 2014-04-30T15:19:07Z jdz quit (Quit: Leaving...) 2014-04-30T15:25:09Z xymox joined #sbcl 2014-04-30T15:28:26Z christoph_debian quit (Ping timeout: 252 seconds) 2014-04-30T15:29:45Z Hydan joined #sbcl 2014-04-30T15:31:21Z xymox quit (Ping timeout: 250 seconds) 2014-04-30T15:32:02Z prxq_ quit (Ping timeout: 265 seconds) 2014-04-30T15:33:20Z christoph_debian joined #sbcl 2014-04-30T15:33:50Z xymox joined #sbcl 2014-04-30T15:34:18Z Hydan quit (Ping timeout: 240 seconds) 2014-04-30T15:34:24Z foom quit (Quit: Leaving) 2014-04-30T15:36:34Z Hydan joined #sbcl 2014-04-30T15:38:58Z foom joined #sbcl 2014-04-30T15:40:23Z Krystof quit (Ping timeout: 240 seconds) 2014-04-30T15:40:28Z nyef: Okay, I found that the trace changes that I made are insufficient on the lisp side and require a runtime-side component as well, but I should be able to get them working... Though having a known test case or two would be nice. 2014-04-30T15:41:52Z Krystof joined #sbcl 2014-04-30T15:41:52Z ChanServ has set mode +o Krystof 2014-04-30T15:46:42Z Krystof: oh, wait. Maybe this problem is "thumb" 2014-04-30T15:47:20Z Krystof: how do I make my gcc be non-thumb? 2014-04-30T15:47:53Z nyef: Add -marm to CFLAGS in the Config file. 2014-04-30T15:49:21Z nyef: I have a possible line on allowing the runtime to be thumb without breaking anything, but it's not at the top of my list right now. 2014-04-30T15:51:14Z Krystof: a repl 2014-04-30T15:51:15Z Krystof: excellent 2014-04-30T16:03:30Z pranavrc quit (Remote host closed the connection) 2014-04-30T16:48:30Z slyrus: thumb? 2014-04-30T16:52:09Z nyef: slyrus: The 16-bit-wide ARM instruction set, as opposed to the 32-bit wide one. 2014-04-30T16:53:46Z slyrus: ah, thanks 2014-04-30T16:54:23Z stassats: why didn't they call it "pinky"? 2014-04-30T16:54:27Z stassats: if it's smaller 2014-04-30T16:55:06Z nyef: Because then they'd have to rename the big one to "brain". 2014-04-30T17:40:50Z stassats: and i got myself linkage table 2014-04-30T17:42:18Z slyrus quit (Ping timeout: 240 seconds) 2014-04-30T17:43:59Z Krystof: my self-build did not go so well 2014-04-30T17:43:59Z Krystof: debugger invoked on a SB-INT:BUG: 2014-04-30T17:43:59Z Krystof: failed AVER: (REGISTER-P SRC) 2014-04-30T17:44:29Z Krystof: in WRITE-CONSTANTS-H 2014-04-30T17:46:59Z Krystof: 2: ((FLET #:G135800 :IN SB-VM::CMP-INST-EMITTER) 2014-04-30T17:46:59Z Krystof: :AL 2014-04-30T17:46:59Z Krystof: # 2014-04-30T17:46:59Z Krystof: #) 2014-04-30T17:47:35Z stassats: didn't i get that fixed? 2014-04-30T17:47:58Z stassats: for if-eq 2014-04-30T17:52:11Z michael_lee quit (Ping timeout: 255 seconds) 2014-04-30T17:58:29Z drmeister quit (Remote host closed the connection) 2014-04-30T18:01:38Z ASau joined #sbcl 2014-04-30T18:06:23Z stassats: for variables, linkage-table doesn't seem to be doing enough indirections 2014-04-30T18:07:45Z stassats: which seems to be my fault 2014-04-30T18:12:18Z stassats: pushed :linkage-table changes 2014-04-30T18:13:54Z drmeister joined #sbcl 2014-04-30T18:16:38Z stassats: thought about trying sb-dynamic-core, but it's too new 2014-04-30T18:16:43Z drmeister quit (Read error: Connection reset by peer) 2014-04-30T18:27:16Z rpg joined #sbcl 2014-04-30T18:39:27Z nyef: Actualy, try sb-dynamic-core with a -mthumb runtime. 2014-04-30T18:39:54Z stassats: sb-dynamic-core is not yet merged 2014-04-30T18:40:01Z nyef: Oh, right. 2014-04-30T18:40:28Z stassats: or rather, the arm port is not yet merged 2014-04-30T18:40:40Z cracauer: what is sb-dynamic-core? 2014-04-30T18:40:55Z stassats: linkage-table for symbols in sbcl 2014-04-30T18:41:28Z stassats: meaning, recompile the runtime, and you're good to go, no need to rebuild the core 2014-04-30T18:42:15Z cracauer: Thanks. 2014-04-30T18:42:26Z oleo quit (Ping timeout: 246 seconds) 2014-04-30T18:45:08Z oleo joined #sbcl 2014-04-30T18:50:45Z foom: cracauer: we use that so we can relink the runtime binary as part of the build process. 2014-04-30T18:51:38Z nyef: And on ARM, it might allow us to run a -mthumb runtime with our current core. 2014-04-30T18:52:25Z stassats: how does it help? 2014-04-30T18:52:32Z Krystof: stassats: yeah, sorry, you probably did, I pulled back to nyef's HEAD while finding out about thumb 2014-04-30T18:52:38Z Krystof: an extra layer of indirection always helps 2014-04-30T18:52:50Z stassats: minion: advice on indirection? 2014-04-30T18:52:51Z minion: You can't expect automated advice for everything. 2014-04-30T18:52:57Z stassats: minion: advice on level? 2014-04-30T18:52:57Z minion: You can't expect automated advice for everything. 2014-04-30T18:53:02Z stassats: minion: bad bot 2014-04-30T18:53:05Z minion: i'm not a bot. i prefer the term ``electronically composed''. 2014-04-30T18:53:32Z nyef: stassats: arm-assm.S, lines 152 and 231, approximately. 2014-04-30T18:55:17Z stassats: i'm adding things into PC, would that work? 2014-04-30T18:55:41Z stassats: here's the sequence: https://github.com/stassats/sbcl-arm/blob/758f3e0fed880c7d07db6aca75ea0ed97b7b90e4/src/runtime/arm-arch.c#L354 2014-04-30T18:56:41Z stassats: and the second // orr reg, reg, address[8:16] should have [16:24] 2014-04-30T18:58:43Z nyef: No, we need a BX instruction at the end. 2014-04-30T18:59:56Z stassats: that would require 6 bytes per entry 2014-04-30T18:59:58Z nyef: Why are you faffing about with segments and not doing an ldr reg, [pc+0] / bx reg /
? 2014-04-30T19:00:22Z stassats: prefetching? 2014-04-30T19:00:39Z slyrus joined #sbcl 2014-04-30T19:00:47Z nyef: Reading PC always biases by 8. 2014-04-30T19:01:13Z nyef: ... At least, I think it was 8. 2014-04-30T19:02:04Z nyef: Something to do with the pipeline on the first ARMs, and then they enshrined it in the spec because having it vary on a per-CPU basis would suck horribly. 2014-04-30T19:03:19Z stassats: i meant prefetching coupled with speculative execution 2014-04-30T19:03:31Z stassats: non-instructions may throw it off 2014-04-30T19:11:54Z nyef: So... put a NOP or a recursive jump in before. And a BX is a jump anyway. 2014-04-30T19:12:54Z stassats: how can i test this? 2014-04-30T19:13:09Z stassats: maybe compile a .so with thumb 2014-04-30T19:15:28Z stassats: well, gcc inserts a dummy function which does BX 2014-04-30T19:28:17Z rpg_ joined #sbcl 2014-04-30T19:28:26Z rpg_ quit (Client Quit) 2014-04-30T19:29:31Z rpg_ joined #sbcl 2014-04-30T19:29:45Z stassats: the indication that it is thumb is on the first bit, so, would dlsym return the address with the first bit set? 2014-04-30T19:32:18Z rpg quit (Ping timeout: 240 seconds) 2014-04-30T19:33:47Z drmeister joined #sbcl 2014-04-30T19:34:56Z nyef: I hope it's either that, or a dummy function. The issue is that we pass the address in a boxed register. 2014-04-30T19:35:58Z stassats: i can see the first set bit being a problem for a boxed register 2014-04-30T19:36:16Z stassats: what is it boxed to? 2014-04-30T19:36:31Z stassats: a shifted fixnum? 2014-04-30T19:38:00Z nyef: If it's an ARM function, as opposed to a THUMB function, the low two address bits are clear. 2014-04-30T19:38:25Z slyrus quit (Ping timeout: 252 seconds) 2014-04-30T19:42:52Z stassats: i can't convince LD not to generate those stubs 2014-04-30T19:45:34Z stassats: nyef: is the undefined function error broken for you too? 2014-04-30T19:47:15Z nyef: Err... I don't believe so. Broken, how? 2014-04-30T19:47:19Z stassats: i'll make it use BX and just hope that it works 2014-04-30T19:47:24Z stassats: nyef: not working 2014-04-30T19:47:51Z stassats: The value -1361760300 is not of type (MOD 536870909). 2014-04-30T19:48:12Z stassats: during a call to make-compiled-code-location from compute-calling-frame 2014-04-30T19:48:38Z stassats: i'm already used for this to mean a call to undefined function 2014-04-30T19:48:54Z nyef: Ahh, yeah, that happens, but only when you do the backtrace. 2014-04-30T19:49:17Z stassats: which slime does 2014-04-30T19:49:31Z nyef: If you do it from a bare REPL, you still see the effects, though. 2014-04-30T19:49:41Z nyef: But it reports the initial error correctly. 2014-04-30T19:50:02Z stassats: so, what's causing this? 2014-04-30T19:50:53Z nyef: No idea. 2014-04-30T20:01:28Z stassats: it's coming from find-escaped-frame 2014-04-30T20:05:31Z nyef: Is that one of the ones with arm-specific damage? 2014-04-30T20:05:33Z erikvarga joined #sbcl 2014-04-30T20:06:00Z nyef: Ah. Yes, it is. 2014-04-30T20:06:18Z stassats: the subtraction produces a negative number 2014-04-30T20:06:28Z stassats: PC is smaller than whatever it subtracts from it 2014-04-30T20:06:44Z nyef: Yup, because we're in the runtime when it triggers. 2014-04-30T20:07:02Z stassats: what does "escaped" mean? 2014-04-30T20:08:18Z nyef: In this case, I think it means more "interrupted" than anything else. 2014-04-30T20:15:22Z stassats: what makes a frame interrupted? 2014-04-30T20:16:41Z ltbarcly_ joined #sbcl 2014-04-30T20:19:20Z ltbarcly quit (Ping timeout: 255 seconds) 2014-04-30T20:22:44Z nyef: The kernel reaching down like the hand of god, smacking the CPU state into a sigcontext, and passing it off to a handler function? 2014-04-30T20:23:13Z stassats: so, a top of what would be considered a backtrace 2014-04-30T20:23:41Z nyef: It's often set as a stack-top-hint, yes. 2014-04-30T20:23:41Z stassats: can there be more than one? 2014-04-30T20:23:58Z nyef: Yup! When you're in the debugger, cause another error. 2014-04-30T20:24:01Z Krystof: ... 2014-04-30T20:24:04Z Krystof: there can be only one 2014-04-30T20:24:08Z Krystof: no, sorry, ignore me 2014-04-30T20:24:13Z nyef: Heh! Thank you, Krystof! (-: 2014-04-30T20:25:01Z stassats: Interrupt MacLeod? 2014-04-30T20:25:53Z nyef: There can only be one... if you're running SBCL on AWS in Scotland? 2014-04-30T20:29:14Z stassats: LDR PC, [PC, #-4] #xADDR could be used as an entry, only two words 2014-04-30T20:29:29Z stassats: bot, no interworking 2014-04-30T20:29:41Z stassats: or maybe? 2014-04-30T20:31:52Z nyef: No interworking. 2014-04-30T20:35:44Z rpg_ quit (Quit: rpg_) 2014-04-30T20:37:00Z fikusz quit (Quit: Leaving) 2014-04-30T20:38:01Z fikusz joined #sbcl 2014-04-30T20:40:49Z stassats: we can have two jump tables, one for thumb, one for pinky 2014-04-30T20:42:23Z nyef: Okay, on my short-list are more stepping bits, and possibly this find-escaped-frame damage. 2014-04-30T20:42:52Z stassats: and i'll bxify linkage-table 2014-04-30T20:44:06Z stassats: and arm instructions are fun 2014-04-30T20:44:19Z nyef: Yeah, they can be delightfully twisted at times. 2014-04-30T20:44:20Z stassats: though x86 presents more layers "ooh, now i get it" of fun 2014-04-30T20:44:34Z stassats: with all the backwards compatible prefixes 2014-04-30T20:45:27Z nyef: There is that. 2014-04-30T20:48:10Z stassats: can *elsewhere* be 32MB from not-*elsewhere*? 2014-04-30T20:50:29Z nyef: I really, really hope not. 2014-04-30T20:50:44Z nyef: They're constrained to a single CODE-OBJECT. 2014-04-30T20:53:35Z stassats: the whole linkage table could be prefilled with ldr/bx 2014-04-30T20:53:50Z stassats: only the address changes 2014-04-30T20:57:14Z nyef: You could store the addresses in non-contiguous space, too. 2014-04-30T20:57:47Z nyef: Or even... What if you moved the ldr/bx to call_into_c ? 2014-04-30T20:57:55Z nyef: ... That might be a bit much. 2014-04-30T20:58:07Z nyef: But you'd end up having a fixnum index into the table as your function pointer. 2014-04-30T21:01:06Z nyef: Okay, I need to run. I'll probably not be back until tomorrow. 2014-04-30T21:01:10Z nyef quit (Quit: G'night all.) 2014-04-30T21:25:16Z rpg joined #sbcl 2014-04-30T21:26:54Z erikvarga left #sbcl 2014-04-30T21:57:59Z drmeister quit (Remote host closed the connection) 2014-04-30T22:09:18Z rpg quit (Ping timeout: 240 seconds) 2014-04-30T22:10:12Z rpg joined #sbcl 2014-04-30T22:11:43Z drmeister joined #sbcl 2014-04-30T22:21:07Z zRecursive joined #sbcl 2014-04-30T22:38:33Z eudoxia joined #sbcl 2014-04-30T22:39:28Z ltbarcly_ quit (Quit: Computer has gone to sleep.) 2014-04-30T22:43:47Z drmeister quit (Read error: Connection reset by peer) 2014-04-30T23:18:30Z cmack`` joined #sbcl 2014-04-30T23:19:57Z segv- quit (Remote host closed the connection) 2014-04-30T23:19:59Z cmack` quit (Ping timeout: 240 seconds) 2014-04-30T23:22:07Z cmack``` joined #sbcl 2014-04-30T23:23:56Z zRecursive quit (Remote host closed the connection) 2014-04-30T23:24:05Z cmack`` quit (Ping timeout: 276 seconds) 2014-04-30T23:35:23Z ltbarcly joined #sbcl