2015-02-04T10:55:55Z ccl-logbot joined #sbcl 2015-02-04T10:55:55Z 2015-02-04T10:55:55Z names: ccl-logbot nikki93 pacon psy_ angavrilov fridim_ Posterdati Shinmera Cymew pranavrc christoph_debian ASau cojy_ irsol selat foom mega1 jdz adlai loke msmith psilord1 soggybread fitzsim hlavaty milosn weissschloss brucem specbot fikusz_ dead__ pkhuong_ mood Perlboy ivan4th` yauz_2 luis` dustinm`` salv00 reb` jrm @Krystof scymtym kanru snafuchs jsnell PuercoPop pchrist carvite drmeister akkad p_l flip214 Bike minion redline6561 wasabiz__ |3b| antoszka nicdev 2015-02-04T10:55:55Z names: fe[nl]ix Blkt joshe ferada les 2015-02-04T11:08:59Z attila_lendvai joined #sbcl 2015-02-04T11:26:51Z nikki93 quit (Remote host closed the connection) 2015-02-04T11:32:39Z scymtym_ joined #sbcl 2015-02-04T11:36:00Z scymtym_: this could be a regression: https://ci.cor-lab.org/job/sbcl-master/845/featureset=5,label=ubuntu_trusty_32bit/consoleFull (search for "UNEXPECTED-FAILURE :KILL-NON-LISP-THREAD") 2015-02-04T11:36:20Z scymtym_: featureset=5 is with sb-safepoint, sb-thruption, sb-wtimer 2015-02-04T11:59:28Z pacon quit (Quit: Leaving) 2015-02-04T12:11:52Z specbot quit (Disconnected by services) 2015-02-04T12:11:55Z specbot joined #sbcl 2015-02-04T12:15:00Z luis- joined #sbcl 2015-02-04T12:16:58Z brucem_ joined #sbcl 2015-02-04T12:20:55Z brucem quit (*.net *.split) 2015-02-04T12:20:55Z dead__ quit (*.net *.split) 2015-02-04T12:20:55Z luis` quit (*.net *.split) 2015-02-04T12:23:05Z brucem_ is now known as brucem 2015-02-04T12:23:59Z brucem quit (Changing host) 2015-02-04T12:23:59Z brucem joined #sbcl 2015-02-04T12:24:44Z eudoxia joined #sbcl 2015-02-04T12:27:20Z dead__ joined #sbcl 2015-02-04T12:27:23Z nikki93 joined #sbcl 2015-02-04T12:31:57Z nikki93 quit (Ping timeout: 252 seconds) 2015-02-04T12:44:46Z msmith left #sbcl 2015-02-04T12:44:55Z mega1 quit (Ping timeout: 265 seconds) 2015-02-04T12:52:36Z attila_lendvai: if someone feels like playing with backtraces and error reporting, I think this is an important improvement when dealing with nested errors: https://github.com/sbcl/sbcl/pull/3 2015-02-04T13:15:50Z pranavrc quit 2015-02-04T14:20:10Z nyef joined #sbcl 2015-02-04T14:26:30Z nikki93 joined #sbcl 2015-02-04T14:31:03Z nikki93 quit (Ping timeout: 244 seconds) 2015-02-04T14:35:02Z nikki93 joined #sbcl 2015-02-04T14:39:33Z nikki93 quit (Ping timeout: 252 seconds) 2015-02-04T14:44:22Z nikki93 joined #sbcl 2015-02-04T14:46:36Z salv00 quit (Remote host closed the connection) 2015-02-04T14:51:54Z salva joined #sbcl 2015-02-04T14:56:52Z nikki93 quit (Ping timeout: 265 seconds) 2015-02-04T14:58:11Z nikki93 joined #sbcl 2015-02-04T15:16:29Z nikki93 quit (Ping timeout: 245 seconds) 2015-02-04T15:19:15Z nikki93 joined #sbcl 2015-02-04T15:19:49Z oleo joined #sbcl 2015-02-04T15:21:53Z fitzsim quit (Read error: Connection reset by peer) 2015-02-04T15:22:09Z fitzsim joined #sbcl 2015-02-04T15:28:53Z nikki93 quit (Ping timeout: 240 seconds) 2015-02-04T15:31:13Z nikki93 joined #sbcl 2015-02-04T15:54:07Z nikki93 quit (Ping timeout: 255 seconds) 2015-02-04T16:02:34Z nikki93 joined #sbcl 2015-02-04T16:09:13Z nikki93 quit (Ping timeout: 244 seconds) 2015-02-04T16:17:18Z davazp joined #sbcl 2015-02-04T16:30:05Z nikki93 joined #sbcl 2015-02-04T16:35:13Z nikki93 quit (Ping timeout: 264 seconds) 2015-02-04T16:36:58Z nikki93 joined #sbcl 2015-02-06T09:52:00Z ccl-logbot joined #sbcl 2015-02-06T09:52:00Z 2015-02-06T09:52:01Z names: ccl-logbot specbot minion psy_ scymtym_ Shinmera angavrilov mega1 gingerale dustinm` pranavrc christoph_debian psilord1 karswell edgar-rft irsol scymtym salva ASau adlai Cymew Intensity alchemis7 brucem luis fridim_ Posterdati cojy_ foom jdz loke soggybread hlavaty milosn weissschloss fe[nl]ix Blkt joshe ferada les nicdev antoszka |3b| wasabiz__ redline6561 Bike flip214 p_l akkad drmeister carvite pchrist PuercoPop jsnell snafuchs kanru @Krystof jrm reb` 2015-02-06T09:52:01Z names: yauz_2 ivan4th` mood pkhuong_ fikusz_ 2015-02-06T10:02:13Z dustinm` quit (Ping timeout: 264 seconds) 2015-02-06T10:24:42Z psy_ quit (Read error: Connection reset by peer) 2015-02-06T10:25:05Z psy_ joined #sbcl 2015-02-06T10:50:48Z scymtym_ quit (Ping timeout: 245 seconds) 2015-02-06T11:12:00Z Shinmera quit (Quit: しつれいしなければならないんです。) 2015-02-06T11:30:36Z scymtym: is anybody already working on the cyclic superclass bug? otherwise i would give it a shot 2015-02-06T11:47:08Z attila_lendvai joined #sbcl 2015-02-06T11:56:56Z mega1` joined #sbcl 2015-02-06T11:59:40Z Intensity quit (Ping timeout: 265 seconds) 2015-02-06T12:00:22Z Intensity joined #sbcl 2015-02-06T12:00:56Z mega1 quit (Ping timeout: 245 seconds) 2015-02-06T12:07:12Z ivan4th` quit (Ping timeout: 276 seconds) 2015-02-06T12:07:37Z scymtym quit (Ping timeout: 250 seconds) 2015-02-06T12:08:11Z attila_lendvai quit (Quit: Leaving.) 2015-02-06T12:08:11Z ivan4th joined #sbcl 2015-02-06T12:13:20Z yauz joined #sbcl 2015-02-06T12:14:09Z carvite_ joined #sbcl 2015-02-06T12:16:55Z scymtym joined #sbcl 2015-02-06T12:19:10Z Intensity quit (*.net *.split) 2015-02-06T12:19:10Z psy_ quit (*.net *.split) 2015-02-06T12:19:11Z pkhuong_ quit (*.net *.split) 2015-02-06T12:19:11Z yauz_2 quit (*.net *.split) 2015-02-06T12:19:12Z pchrist quit (*.net *.split) 2015-02-06T12:19:12Z carvite quit (*.net *.split) 2015-02-06T12:20:19Z Shinmera joined #sbcl 2015-02-06T12:20:27Z carvite_ quit (Changing host) 2015-02-06T12:20:27Z carvite_ joined #sbcl 2015-02-06T12:20:42Z carvite_ is now known as carvite 2015-02-06T12:20:58Z Intensity joined #sbcl 2015-02-06T12:20:58Z psy_ joined #sbcl 2015-02-06T12:20:58Z pkhuong_ joined #sbcl 2015-02-06T12:20:58Z pchrist joined #sbcl 2015-02-06T12:31:39Z eudoxia joined #sbcl 2015-02-06T12:50:08Z psy_ quit (Remote host closed the connection) 2015-02-06T13:02:58Z dustinm` joined #sbcl 2015-02-06T13:09:09Z dustinm` quit (Ping timeout: 252 seconds) 2015-02-06T13:11:41Z stassats joined #sbcl 2015-02-06T13:14:04Z pranavrc quit 2015-02-06T13:23:52Z DeadTrickster joined #sbcl 2015-02-06T13:26:27Z oleo joined #sbcl 2015-02-06T13:41:19Z attila_lendvai joined #sbcl 2015-02-06T13:41:19Z attila_lendvai quit (Changing host) 2015-02-06T13:41:19Z attila_lendvai joined #sbcl 2015-02-06T14:30:09Z ASau quit (Remote host closed the connection) 2015-02-06T14:47:15Z psy_ joined #sbcl 2015-02-06T15:28:27Z oleo is now known as Guest51060 2015-02-06T15:29:19Z oleo__ joined #sbcl 2015-02-06T15:31:37Z Guest51060 quit (Ping timeout: 264 seconds) 2015-02-06T15:32:21Z oleo__ quit (Read error: Connection reset by peer) 2015-02-06T15:34:45Z oleo__ joined #sbcl 2015-02-06T15:35:01Z oleo__ is now known as oleo 2015-02-06T15:48:36Z Adlai` joined #sbcl 2015-02-06T15:48:45Z adlai quit (Ping timeout: 250 seconds) 2015-02-06T16:51:46Z fridim_ quit (Ping timeout: 264 seconds) 2015-02-06T16:52:29Z fridim_ joined #sbcl 2015-02-06T16:57:50Z flip214: shouldn't (reduce #'+) with properly declared types result in an inline expansion with (speed 3) (space 0)? I still see an external call to REDUCE. 2015-02-06T16:58:15Z stassats: no 2015-02-06T16:58:35Z flip214: okay, thanks 2015-02-06T16:59:49Z stassats: how to figure that out: 1) (describe 'reduce) => it's not declared inline or maybe-inline 2) reduce M-., are there any transforms? 2015-02-06T17:00:02Z stassats: no to both questions, meaning it will never be inline expanded 2015-02-06T17:01:12Z flip214: yeah, I saw that in the inspector. 2015-02-06T17:01:24Z flip214: I just wondered, because I thought it would be a prime example for inlining. 2015-02-06T17:10:17Z |3b| quit (Read error: Connection reset by peer) 2015-02-06T17:11:49Z |3b| joined #sbcl 2015-02-06T17:19:52Z oleo quit (Ping timeout: 245 seconds) 2015-02-06T17:20:42Z soggybread quit (Ping timeout: 245 seconds) 2015-02-06T17:21:33Z soggybread joined #sbcl 2015-02-06T17:24:47Z oleo joined #sbcl 2015-02-06T17:25:10Z dustinm` joined #sbcl 2015-02-06T17:45:33Z hlavaty quit (Remote host closed the connection) 2015-02-06T17:48:29Z milanj joined #sbcl 2015-02-06T17:52:00Z karswell quit (Ping timeout: 252 seconds) 2015-02-06T17:58:27Z dustinm` quit (Ping timeout: 252 seconds) 2015-02-06T18:13:47Z eudoxia quit (Quit: Leaving) 2015-02-06T18:18:06Z ASau joined #sbcl 2015-02-06T18:30:30Z attila_lendvai1 joined #sbcl 2015-02-06T18:30:30Z attila_lendvai quit (Disconnected by services) 2015-02-06T18:30:30Z attila_lendvai1 quit (Changing host) 2015-02-06T18:30:30Z attila_lendvai1 joined #sbcl 2015-02-06T18:36:52Z selat joined #sbcl 2015-02-06T18:50:59Z karswell joined #sbcl 2015-02-06T18:55:24Z selat quit (Quit: Lost terminal) 2015-02-06T19:02:47Z attila_lendvai1 quit (Quit: Leaving.) 2015-02-06T19:06:15Z Bicyclidine joined #sbcl 2015-02-06T19:14:23Z eudoxia joined #sbcl 2015-02-06T19:15:10Z ehaliewicz joined #sbcl 2015-02-06T19:19:21Z fitzsim joined #sbcl 2015-02-06T19:37:14Z oleo quit (Quit: Verlassend) 2015-02-06T19:38:40Z oleo joined #sbcl 2015-02-06T19:39:52Z Shinmera quit (Quit: しつれいしなければならないんです。) 2015-02-06T19:48:43Z ivan4th quit (Ping timeout: 245 seconds) 2015-02-06T19:53:03Z dustinm` joined #sbcl 2015-02-06T19:56:19Z ivan4th joined #sbcl 2015-02-06T20:04:35Z dustinm` quit (Ping timeout: 252 seconds) 2015-02-06T20:05:05Z fitzsim quit (Remote host closed the connection) 2015-02-06T20:15:09Z dustinm` joined #sbcl 2015-02-06T20:24:42Z eudoxia quit (Quit: Leaving) 2015-02-06T20:41:37Z dustinm` quit (Ping timeout: 252 seconds) 2015-02-06T20:57:32Z eudoxia joined #sbcl 2015-02-06T21:04:02Z milanj quit (Quit: Leaving) 2015-02-06T21:08:50Z eudoxia quit (Quit: Leaving) 2015-02-06T21:29:28Z angavrilov quit (Remote host closed the connection) 2015-02-06T21:43:53Z Bicyclidine quit (Ping timeout: 246 seconds) 2015-02-06T21:44:07Z Bicyclidine joined #sbcl 2015-02-06T22:16:03Z scymtym_ joined #sbcl 2015-02-06T22:54:04Z attila_lendvai joined #sbcl 2015-02-06T22:54:04Z attila_lendvai quit (Changing host) 2015-02-06T22:54:04Z attila_lendvai joined #sbcl 2015-02-06T22:56:23Z csziacobus joined #sbcl 2015-02-06T23:05:55Z Quadrescence joined #sbcl 2015-02-06T23:06:49Z Bicyclidine quit (Ping timeout: 252 seconds) 2015-02-06T23:08:42Z Bicyclidine joined #sbcl 2015-02-06T23:14:24Z Bicyclidine quit (Ping timeout: 265 seconds) 2015-02-06T23:15:59Z Bicyclidine joined #sbcl 2015-02-06T23:18:20Z gingerale quit (Ping timeout: 252 seconds) 2015-02-06T23:21:27Z Bicyclidine quit (Ping timeout: 250 seconds) 2015-02-06T23:38:17Z Bicyclidine joined #sbcl 2015-02-06T23:50:14Z attila_lendvai quit (Quit: Leaving.) 2015-02-07T00:04:52Z Bicyclidine quit (Ping timeout: 240 seconds) 2015-02-07T00:30:03Z dead_ joined #sbcl 2015-02-07T00:35:53Z dead_ quit (Remote host closed the connection) 2015-02-07T00:57:30Z edgar-rft quit (Quit: lifetime disconnected because no action exists) 2015-02-07T01:48:41Z edgar-rft joined #sbcl 2015-02-07T02:38:47Z nyef joined #sbcl 2015-02-07T02:40:53Z nyef: stassats: I just ran into a situation where I had to use TRACE :ENCAPSULATE NIL to get the trace to kick in at all, and it's # all over the place. 2015-02-07T02:41:36Z nyef: It was a case of "oh, right, it's taking the function reference at load time, not at runtime, so the fdefn doesn't even get looked at". 2015-02-07T02:42:51Z nyef: This was on 1.1.5.23 under an uncontrolled debug policy, and I KNOW that :encapsulate nil is semantically broken, but it's a scenario to consider. 2015-02-07T03:38:34Z christoph_debian quit (Ping timeout: 245 seconds) 2015-02-07T03:41:11Z dustinm` joined #sbcl 2015-02-07T03:44:02Z scymtym_ quit (Ping timeout: 246 seconds) 2015-02-07T03:47:03Z dustinm` quit (Ping timeout: 245 seconds) 2015-02-07T03:52:13Z christoph_debian joined #sbcl 2015-02-07T03:58:12Z nyef quit (Quit: Why am I still awake?) 2015-02-07T04:12:01Z oleo is now known as Guest72520 2015-02-07T04:12:53Z oleo__ joined #sbcl 2015-02-07T04:15:31Z Guest72520 quit (Ping timeout: 265 seconds) 2015-02-07T04:29:52Z dustinm` joined #sbcl 2015-02-07T04:36:32Z stassats quit (Ping timeout: 245 seconds) 2015-02-07T04:45:36Z csziacobus quit (Quit: csziacobus) 2015-02-07T05:28:03Z fridim_ quit (Ping timeout: 250 seconds) 2015-02-07T06:05:36Z gabriel_laddel joined #sbcl 2015-02-07T06:36:22Z psy_ quit (Ping timeout: 240 seconds) 2015-02-07T06:56:40Z Shinmera joined #sbcl 2015-02-07T07:11:16Z psy_ joined #sbcl 2015-02-14T12:12:55Z ccl-logbot joined #sbcl 2015-02-14T12:12:55Z 2015-02-14T12:12:55Z names: ccl-logbot fridim_ oleo stassats haplo` attila_lendvai psy_ angavrilov scymtym_ gingerale Shinmera zacts edgar-rft pranavrc christoph_debian |3b| minion specbot mega1 DeadTrickster @Krystof selat karswell` foom irsol ASau cojy_ milosn pkhuong dustinm` fikusz jsnell weissschloss salv0 ams Bike Adlai drmeister nightshade427 jrm jdz psilord1 White_Flame akkad nicdev alchemis7 soggybread Intensity p_l luis ivan4th pchrist scymtym carvite yauz brucem Posterdati 2015-02-14T12:12:55Z names: loke mood reb` kanru snafuchs PuercoPop flip214 redline6561 wasabiz__ antoszka fe[nl]ix Blkt ferada les 2015-02-14T12:33:55Z attila_lendvai quit (Quit: Leaving.) 2015-02-14T12:45:34Z attila_lendvai joined #sbcl 2015-02-14T12:46:05Z attila_lendvai quit (Client Quit) 2015-02-14T12:49:34Z pranavrc quit 2015-02-14T12:56:50Z nyef joined #sbcl 2015-02-14T13:38:16Z nyef: Hrm. Dominator information is only computed if we're doing loop analysis... 2015-02-14T13:54:48Z stassats: made a test case for the indirect closure values where the saved FP has both a register location and a save location 2015-02-14T13:54:57Z stassats: and i store only the save location in the debug function 2015-02-14T13:55:38Z stassats: (not to mention that it's just broken on gengc) 2015-02-14T13:56:35Z nyef: Ooh. Structural data-flow analysis is in Aho, Sethi, and Ullman. I wonder if it's also in Aho and Ullman? 2015-02-14T13:58:35Z stassats` joined #sbcl 2015-02-14T14:01:11Z stassats: fetching the old FP on arm doesn't seem to give me the right thing 2015-02-14T14:01:38Z stassats: ok, it has wrong sc 2015-02-14T14:02:33Z nyef: ... Looks like it's not in Aho and Ullman. 2015-02-14T14:03:16Z stassats: on x86oids i used sap scs, but sap scs live on a different stack on gengc, and not translating scs on arm makes it to a any-reg, when in fact it's unsigned-reg 2015-02-14T14:04:03Z stassats: or descriptor-reg 2015-02-14T14:06:47Z stassats: alright, using descriptor-sap works 2015-02-14T14:07:28Z stassats: now, i'll have to save more information about indirect fp, reusing the old format doesn't work, bummer 2015-02-14T14:10:36Z nyef: I'm still at a loss on how to fix STACK analysis. /-: 2015-02-14T14:15:28Z stassats: didn't think about stack analysis this week 2015-02-14T14:15:44Z nyef: If I can get the analysis right, I have an angle on the cleanups. 2015-02-14T14:15:51Z nyef: The trick is getting the analysis right. 2015-02-14T14:25:42Z stassats: sc-offset is 27 bits, and it's stored in a T vector, on x86-64 we can bundle them together saving space 2015-02-14T14:36:29Z stassats`: found out that connecting to my arm development phone is faster through ethernet over USB instead of wifi 2015-02-14T14:36:38Z stassats`: better for syncinc cross compilation results 2015-02-14T14:39:00Z stassats`: found a new instance of :INVALID-VALUE-FOR-UNESCAPED-REGISTER-STORAGE 2015-02-14T14:39:14Z LiamH joined #sbcl 2015-02-14T14:42:50Z stassats`: when in fact the frame is actually escaped, through a sigsegv 2015-02-14T14:51:24Z stassats`: (defun segv (vector) (declare (optimize (safety 0))) (svref vector 0)) 2015-02-14T14:51:34Z stassats`: (segv 0) => (SEGV :INVALID-VALUE-FOR-UNESCAPED-REGISTER-STORAGE) 2015-02-14T14:53:41Z stassats`: with SPEED it even things that it's in tl,external 2015-02-14T14:56:30Z Krystof quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2015-02-14T14:56:36Z Xof joined #sbcl 2015-02-14T14:56:36Z ChanServ has set mode +o Xof 2015-02-14T14:58:52Z nyef: If I have a block B1 starting with an "entry DX" for some set of variables V, and a block B2 which is the nearest dominator for B1 and all blocks B[2-n] that push any of the variables V, am I guaranteed that a backwards flow analysis of V from B1 limited to those blocks dominated by B2 will do what I want? 2015-02-14T14:59:35Z nyef: ... It feels like there's a massive hole in this idea. 2015-02-14T15:03:49Z stassats`: i added fake_foreign_function_callto lisp_memory_fault_error and it still can't find the escapee 2015-02-14T15:05:46Z nyef: Oh, god. Not l_m_f_e() again! 2015-02-14T15:07:23Z stassats`: it doesn't call MEMORY_FAULT_ERROR, it returns to it 2015-02-14T15:07:33Z nyef: Exactly. 2015-02-14T15:08:11Z nyef: If there's an aspect of the runtime that I dislike the most, that'd be it. 2015-02-14T15:14:09Z stassats`: so how do i solve that? 2015-02-14T15:14:50Z nyef: The first question is why we do the return thing instead of leaving the exception frame on the stack. 2015-02-14T15:15:13Z stassats`: i just tried calling and something goes wrong 2015-02-14T15:17:10Z stassats`: The value NIL is not of type SB-DI:FRAME. 2015-02-14T15:17:26Z nyef: Ahh, backtrace! 2015-02-14T15:18:56Z nyef: The backtrace through the call-in mistakes the signal context for a return address or something like that. 2015-02-14T15:28:14Z stassats`: don't see an obvious solution for that 2015-02-14T15:31:06Z nyef: It's all about the calling convention: If it's a lisp frame then the return-address scheme has to be lispy, and otherwise it has to be alien, and the boundary between the two is call_into_lisp / call_into_c. 2015-02-14T15:31:12Z nyef: More or less. 2015-02-14T15:34:17Z stassats`: if we store the context we'll have to copy it or something, since it can be stack allocated as well 2015-02-14T15:34:52Z nyef: We can't portably copy contexts. 2015-02-14T15:35:06Z stassats`: the info we need 2015-02-14T15:35:06Z nyef: And we must presume that they ARE stack allocated. 2015-02-14T15:38:41Z soggybread quit (Ping timeout: 256 seconds) 2015-02-14T15:39:46Z nyef: It'd be better to fix the backtrace logic. Don't we need reliable backtrace anyway? 2015-02-14T15:42:20Z stassats`: it gets pretty confused by not fp saving foreign code already, how to fix _that_? 2015-02-14T15:43:00Z stassats`: we do save frame pointers with some declarations 2015-02-14T15:43:59Z nyef: Either DWARF, or arrange for something with frame pointers to find the preceeding lisp frame on a stack. 2015-02-14T15:44:34Z stassats`: SB-C:ALIEN-FUNCALL-SAVES-FP-AND-PC 2015-02-14T15:45:19Z soggybread joined #sbcl 2015-02-14T15:45:21Z stassats`: so, by the time ltn happens, can i delete lvars (or refs to lvars or whatever i need to delete)? 2015-02-14T15:47:08Z stassats`: i can try... 2015-02-14T15:53:08Z stassats`: ... if i knew how to do that 2015-02-14T15:54:18Z stassats`: probably propagating constants is a better way 2015-02-14T15:54:24Z stassats`: just need to handle casts properly 2015-02-14T16:04:50Z Xof quit (Remote host closed the connection) 2015-02-14T16:05:25Z stassats`: forgot about https://bugs.launchpad.net/sbcl/+bug/1389433 2015-02-14T16:15:06Z stassats`: i can just stop calling lognot on non-integers in the transform, but the result is unreachable code, not a type conflict 2015-02-14T16:15:30Z stassats`: oh, that's what it should be 2015-02-14T16:17:53Z stassats`: nyef: do you still have your DCE-branch sbcl handy? 2015-02-14T16:18:24Z nyef: Should do, yes. 2015-02-14T16:18:46Z stassats`: try (lambda (c) (block nil (flet ((f (x) (unless x (return)) (the integer x))) (logior c (f nil))))) 2015-02-14T16:19:09Z nyef: SB-KERNEL:CASE-FAILURE. 2015-02-14T16:19:14Z stassats`: ok 2015-02-14T16:19:24Z nyef: Hang on, let me try something. 2015-02-14T16:20:28Z nyef: No, still case-failure. 2015-02-14T16:20:40Z nyef: Tried adding an invocation of DCE just before IR1-PHASES. 2015-02-14T16:21:22Z stassats`: ok, then it really shouldn't expect an integer when it asks for T 2015-02-14T16:22:13Z nyef: Maybe the DCE needs to be done DURING ir1-phases? 2015-02-14T16:22:27Z stassats`: it does the right thing when i just restrict it to integers 2015-02-14T16:23:30Z stassats`: and that seems like a good enough solution to me 2015-02-14T16:28:58Z nyef: ... Okay, the backwards flow analysis to the closest dominator seems like a workable idea. There can be no other entries, the propagation is backwards-only, and the closest dominator is at worst the start of the enclosing LET value forms. 2015-02-14T16:30:46Z nyef: Now the problem is that dominance information is only computed if the loop analyzer is enabled. 2015-02-14T16:31:16Z stassats`: how are CIFs related to loops? 2015-02-14T16:32:38Z nyef: They're the only real control-flow split operation required for a loop exit test? 2015-02-14T16:32:47Z nyef: Or are you trying to ask something else? 2015-02-14T16:33:02Z stassats`: the original problem didn't have loops, or did it? 2015-02-14T16:33:28Z nyef: It didn't, but careful arrangement of loops makes things worse. 2015-02-14T16:35:03Z nyef: I don't want to just drive the bug to be more subtle, I want to kill it. 2015-02-14T16:35:13Z stassats`: decided to push undefined warnings from fopcompile, since I don't think i'll approach fopcompiling from IR1 anytime soon 2015-02-14T16:35:53Z nyef: Also, I'm looking at a solution which requires the output of an analysis performed only when the loop analyzer is enabled. 2015-02-14T16:36:15Z stassats`: I think it's enough bug fixing for today 2015-02-14T16:36:39Z stassats`: maybe i'll try the constant propagation thing tomorrow 2015-02-14T16:37:03Z stassats`: and if you'll fix stack analysis, we'll no longer have serious compiler bugs 2015-02-14T16:37:13Z nyef: Type propagation is dataflow analysis, isn't it? 2015-02-14T16:38:04Z stassats`: the problem is that ltn looks at the type and sees that it's a singleton and uses its value 2015-02-14T16:38:19Z stassats`: leaving the cast, the allocated closure hanging 2015-02-14T16:38:54Z nyef: Can you snap the cast out earlier? 2015-02-14T16:38:55Z stassats`: so i'll have to propagate such constants from the lambda bind, not just lets, and also to handle it from casts 2015-02-14T16:39:31Z nyef: Still, type inference is a dataflow analysis, isn't it? 2015-02-14T16:39:35Z stassats`: i'll have to leave casts 2015-02-14T16:39:59Z stassats`: and they'll have to be elided (or not) elsewhere 2015-02-14T16:41:27Z stassats`: ir1 does handle constants too, but i guess it happens early enough not to leave any orphans behind 2015-02-14T16:41:30Z nyef: Add something to ir1-optimize-cast or whatever it's called? 2015-02-14T16:42:23Z stassats`: that wouldn't help with external lambda binds, because it doesn't propagate anything 2015-02-14T16:42:45Z nyef: It may be that you need two separate fixes. 2015-02-14T16:42:52Z stassats`: so it would have to be a combination of that and casts handling 2015-02-14T16:43:18Z stassats`: but the only way the lambda bind can have a constant is through a singleton cast, so it's really about cast propagation 2015-02-14T16:44:47Z stassats`: sb-kernel:classoid-typep is another bug worth fixing 2015-02-14T16:45:32Z stassats`: all other bugs are just "it'd be nice if..." 2015-02-14T16:45:57Z nyef: ALL, or just the compiler bugs? 2015-02-14T16:46:14Z stassats`: all, i think 2015-02-14T16:46:25Z nyef: Hrm. 2015-02-14T16:46:25Z stassats`: as in, not showstoppers 2015-02-14T16:46:30Z nyef: Ah, right. 2015-02-14T16:47:09Z stassats`: the solaris thing is bad too, but who cares about solaris anyway 2015-02-14T16:47:46Z stassats`: there's also some sb-concurrency memory reordering badness 2015-02-14T16:47:56Z stassats`: and some bad threading things on windows 2015-02-14T16:48:04Z nyef: What about the external-format things? 2015-02-14T16:48:08Z stassats`: so, a lot of work just to get everything stable 2015-02-14T16:48:24Z stassats`: nyef: i want to prioritize bug fixing, not new features 2015-02-14T16:48:29Z nyef: Fair enough. 2015-02-14T16:48:49Z stassats`: unless features are easy enough 2015-02-14T16:49:13Z nyef: It's not like anybody uses the windows port, anyway. 2015-02-14T16:49:23Z nyef: I know that the original author certainly doesn't. d-: 2015-02-14T16:49:29Z stassats`: i think it's most downloaded on sourceforge 2015-02-14T16:50:53Z stassats`: see https://sourceforge.net/projects/sbcl/files/sbcl/1.2.7/ 2015-02-14T16:51:38Z stassats`: of course linux users tend to prefer sources, git and os packages 2015-02-14T16:51:41Z nyef: Heh. Closely followed by the x86 linux binary? 2015-02-14T16:51:54Z nyef: Exactly. Massive selection bias there. 2015-02-14T16:52:22Z stassats`: windows downloads 1,601+956 2015-02-14T16:52:33Z stassats`: source code 2,105 2015-02-14T16:53:04Z stassats`: linux x86oids 436+654 2015-02-14T16:53:48Z stassats`: granted, 1.2.7 was a first release in a while where i made a lot of binaries 2015-02-14T16:53:58Z stassats`: didn't get too excited to repeat that for 1.2.8 2015-02-14T16:54:05Z stassats`: maybe for 1.2.9, if i automate it 2015-02-14T16:55:45Z stassats`: foreign callbacks from foreign threads on non-sb-safepoint should be fairly easy to achieve 2015-02-14T16:55:50Z stassats`: maybe for tomorrow as well 2015-02-14T17:05:09Z |3b| has been using windows sbcl lately, and could see wanting to distribute binaries using it at some point in the future 2015-02-14T17:28:20Z nyef: ... wha? The DFO pass numbers in REVERSE DFO? 2015-02-14T17:38:04Z attila_lendvai joined #sbcl 2015-02-14T17:39:21Z attila_lendvai quit (Client Quit) 2015-02-14T17:48:39Z stassats: found a way to save 720K 2015-02-14T17:49:09Z stassats: not storing old-fp and return-pc in the debug function, they are always the same 2015-02-14T17:49:23Z haplo` quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2015-02-14T17:49:42Z stassats`: and nfp, which is just never used 2015-02-14T18:15:12Z nyef: This is x86oid-specific, right? 2015-02-14T18:15:50Z nyef: (Other backends use NFP, and MAY use a different old-fp or return-pc for local functions.) 2015-02-14T18:16:34Z stassats`: the may use, but they don't 2015-02-14T18:19:27Z stassats`: removing compiled-debug-info-name saves 620K, but i'm not sure about that one 2015-02-14T18:19:35Z stassats`: usually the name can be fetched from the first entry 2015-02-14T18:33:53Z nyef: A local call that doesn't get inlined doesn't get the body of the local function spliced into the CFG, does it? 2015-02-14T18:34:30Z nyef: That is, it's represented by a REF to the CLAMBDA and a COMBINATION, not by monkeying the CBLOCK next/prev relationships? 2015-02-14T18:34:42Z nyef: Err.... succ/pred relationships? 2015-02-14T18:36:16Z stassats`: SB-C::CONVERT-CALL? 2015-02-14T18:38:42Z nyef: Assuming that it doesn't monkey the flow graph, I think I need a reverse-path search to find all generators for each variable that do not include a predecessor of the closest common dominator of the generators. 2015-02-14T18:40:37Z nyef: And, because we're in reverse-DFO, that common dominator is the one with the lowest block number. 2015-02-14T18:41:12Z nyef: Still feels weak. 2015-02-14T18:41:48Z nyef: Right, the bound determination has to include the entry DX itself as a root for the dominator search. 2015-02-14T18:43:08Z nyef: I think I need to write this up. /-: 2015-02-14T18:46:00Z stassats quit (Remote host closed the connection) 2015-02-14T18:49:42Z stassats`: stupid slime, eating my repl history 2015-02-14T18:50:31Z ASau quit (Ping timeout: 255 seconds) 2015-02-14T18:57:51Z ASau joined #sbcl 2015-02-14T18:58:56Z stassats`: bah, "No space left on device" and no restart 2015-02-14T18:59:07Z stassats`: i freed the space and now i have to start anew 2015-02-14T19:02:34Z stassats`: and it bloody removes the files when i quit 2015-02-14T19:04:18Z stassats`: note to self: when writing into a log file, don't use with-open-file 2015-02-14T19:04:22Z stassats`: or any similar kind of file 2015-02-14T19:06:20Z stassats`: ok, what i wanted to do: we can't compile 4GB source files 2015-02-14T19:06:27Z stassats`: 4294967310 fell through ETYPECASE expression. Wanted one of ((UNSIGNED-BYTE 8) (UNSIGNED-BYTE 16) (UNSIGNED-BYTE 32)). 2015-02-14T19:06:40Z stassats`: for source info offsets 2015-02-14T19:07:06Z nyef: ... Wha? Four gigabyte SOURCE files? 2015-02-14T19:07:18Z stassats`: yeah, why not? 2015-02-14T19:07:28Z nyef: The mind boggles. 2015-02-14T19:08:04Z stassats`: (with-open-file (stream "/tmp/foo.lisp" :direction :output :element-type :ascii :if-exists :supersede :if-does-not-exist :create) (write-line "(defun foo ()" stream) (loop repeat size do (terpri (* 4096 1024 1024))) (write-sequence ")" stream) (write-sequence "(defun bar ())" stream)) 2015-02-14T19:09:24Z stassats`: this is actually easy to fix 2015-02-14T19:14:08Z stassats`: huh, 32 bit sbcl can compile it? 2015-02-14T19:15:26Z nyef: ... largefile? 2015-02-14T19:15:27Z stassats`: start-positions is #(0 14) 2015-02-14T19:16:01Z stassats`: (ldb (byte 32 0) 4294967310) => 14 2015-02-14T19:17:40Z nyef: It... wraps? 2015-02-14T19:17:56Z nyef: I think you may have found another bug. d-: 2015-02-14T19:19:54Z stassats`: (with-open-file (stream "/tmp/foo.lisp") (file-position stream 4294967310) (file-position stream)) => 14 2015-02-14T19:20:06Z stassats`: and you said it "Wha?", i did find a real bug with it! 2015-02-14T19:24:39Z attila_lendvai joined #sbcl 2015-02-14T19:24:39Z attila_lendvai quit (Changing host) 2015-02-14T19:24:39Z attila_lendvai joined #sbcl 2015-02-14T19:24:42Z attila_lendvai quit (Client Quit) 2015-02-14T19:28:31Z stassats`: alien-funcall for some reason derives the result to be 32-bit 2015-02-14T19:30:01Z stassats`: ha, arm doesn't have it 2015-02-14T19:32:12Z stassats`: it passes the argument alright, though, so it's only with the result 2015-02-14T19:33:10Z stassats`: mips is alright too 2015-02-14T19:33:27Z nyef: ... What are you using for a mips? 2015-02-14T19:33:32Z stassats`: qemu 2015-02-14T19:33:35Z nyef: Ah. 2015-02-14T19:33:51Z stassats`: basically that kind of things is all it's good for 2015-02-14T19:34:59Z stassats`: i did C-c C-c on unix-lseek, now i get 4294967310, but the derived type is still bad 2015-02-14T19:39:24Z nyef: Derived type information? 2015-02-14T19:40:07Z nyef: Looks like off-t is 32 bits? 2015-02-14T19:40:26Z stassats`: no, it's 64 2015-02-14T19:40:45Z nyef: Hrm. That's a little sketchy in and of itself... 2015-02-14T19:40:52Z stassats`: not really 2015-02-14T19:51:00Z stassats`: so, it actually returns two values 2015-02-14T19:51:14Z stassats`: two halves 2015-02-14T19:52:40Z stassats`: but that same ting happens on arm 2015-02-14T19:56:17Z Shinmera quit (Quit: しつれいしなければならないんです。) 2015-02-14T20:06:14Z stassats`: found it 2015-02-14T20:08:25Z stassats`: nyef: that was actually yours 2015-02-14T20:08:45Z nyef: Oh? 2015-02-14T20:08:53Z stassats`: e7476d980c0b4949c9416b59249d0d621c0f747d spot the problem 2015-02-14T20:11:55Z nyef: Not seeing it right off. 2015-02-14T20:12:10Z stassats`: in :alien-rep 2015-02-14T20:12:44Z attila_lendvai joined #sbcl 2015-02-14T20:12:44Z attila_lendvai quit (Changing host) 2015-02-14T20:12:44Z attila_lendvai joined #sbcl 2015-02-14T20:14:01Z jrm: nyef: (dotimes (i 10000) (setf *screen-mode-line-format* '(:eval (run-shell-command "cat /tmp/ml-info.txt" t)))) doesn't seem to be causing any (immediate) problems here. 2015-02-14T20:14:02Z nyef: Still not seeing it. 2015-02-14T20:14:32Z nyef: jrm: Right, because it's just a setf of literal data. Invoke the RUN-SHELL-COMMAND bit in a loop instead. 2015-02-14T20:14:44Z jrm: k 2015-02-14T20:14:59Z stassats`: nyef: you are returning the size of sb!vm:n-word-bits on x86 2015-02-14T20:15:12Z stassats`: which happens to be less than needed for signed-byte 64 2015-02-14T20:15:17Z nyef: Ahh. 2015-02-14T20:15:20Z stassats`: fixed with (max (alien-integer-type-bits type) sb!vm:n-word-bits) 2015-02-14T20:15:43Z stassats`: how come nobody noticed that for over 5 years? 2015-02-14T20:15:46Z nyef: Yes, okay, if it's larger then it gets split anyway... 2015-02-14T20:15:57Z nyef: Because we almost never USE such large values? 2015-02-14T20:16:07Z stassats`: only happened when i tried to compile 4GB file 2015-02-14T20:16:08Z nyef: Almost six years now. 2015-02-14T20:19:25Z jrm: nyef: This will likely cause problems regardless, since there is a loop in the shell script: http://ftfl.ca/paste/ml.html 2015-02-14T20:20:18Z nyef: The shell script has little to do with it. I'm concerned about the use of RUN-SHELL-COMMAND. 2015-02-14T20:20:42Z jrm: k 2015-02-14T20:21:05Z nyef: Hell, substitute /etc/hostname or something. 2015-02-14T20:23:55Z stassats`: would we be content 18.1 exabyte source files? 2015-02-14T20:24:09Z stassats`: with 2015-02-14T20:25:37Z stassats`: seems like less than what zfs or btrfs have 2015-02-14T20:25:52Z stassats`: more, rather 2015-02-14T20:26:10Z stassats`: i'm just laze to write (t t) 2015-02-14T20:26:21Z stassats`: in an etypecase form 2015-02-14T20:26:28Z nyef: Is it a source file limitation, or a file stream limitation? 2015-02-14T20:26:44Z stassats`: debug source map limitation 2015-02-14T20:26:51Z stassats`: coerce-to-smallest-eltype specifically 2015-02-14T20:27:41Z nyef: I think that a 4GB source file was already silly. 2015-02-14T20:28:43Z stassats`: handling 18 EB files at the cost of one line is easy 2015-02-14T20:31:32Z stassats`: nyef: just imagine a concatenate file with everything 2015-02-14T20:31:38Z stassats`: d 2015-02-14T20:34:30Z ASau quit (Remote host closed the connection) 2015-02-14T20:39:26Z ASau joined #sbcl 2015-02-14T20:39:38Z ASau quit (Remote host closed the connection) 2015-02-14T20:39:40Z stassats`: ppc doesn't always save old-fp and return-pc 2015-02-14T20:40:34Z stassats`: only arm and x86oids always put them into save locations 2015-02-14T20:41:22Z ASau joined #sbcl 2015-02-14T20:41:36Z nyef: Sounds about right. 2015-02-14T20:41:45Z stassats`: not clear why, though 2015-02-14T20:41:54Z nyef: On which side? 2015-02-14T20:42:20Z stassats`: ppc 2015-02-14T20:42:44Z nyef: Flexibility with nested local calls? 2015-02-14T20:42:52Z nyef: Something like that, maybe. 2015-02-14T20:43:14Z nyef: x86oids and ARM are the odd ones out, to the point of having custom hacks in PACK to cover for the damage. 2015-02-14T20:43:19Z stassats`: when they save, the save at the same place 2015-02-14T20:44:41Z stassats`: either way, not saved => in LRA and OCFP registers, saved => same register locations 2015-02-14T20:44:57Z stassats`: same stack locations 2015-02-14T20:45:11Z nyef: Something to do with leaf functions, maybe? 2015-02-14T20:45:17Z stassats`: now, how would unescaped frames deal LRA and OCFP? 2015-02-14T20:46:10Z stassats`: it just accesses them through lra-offset 2015-02-14T20:46:31Z stassats`: so the return-pc and old-fp in the debug function are pretty useless 2015-02-14T20:50:39Z stassats`: ok, no, they can be put somewhere else 2015-02-14T20:51:29Z stassats`: basically, the difference is because x86oids do not have LRA and OCFP 2015-02-14T20:51:44Z stassats`: registers 2015-02-14T20:52:29Z stassats`: and ppc does, and it can rename them, but the stack save locations are always the same 2015-02-14T20:53:03Z stassats`: well, a couple of conditionals for 700K of space savings sounds like a good price 2015-02-14T20:57:41Z pacon joined #sbcl 2015-02-14T20:59:21Z stassats`: i like the NEWS entry in particular 2015-02-14T20:59:37Z stassats`: the users will rejoice with support of 4G sources 2015-02-14T21:00:02Z Shinmera joined #sbcl 2015-02-14T21:02:15Z fridim_ quit (Ping timeout: 256 seconds) 2015-02-14T21:05:44Z stassats`: incidentally, i made my 600th commit 2015-02-14T21:07:29Z stassats`: only a thousand to go to overtake nikodemus 2015-02-14T21:08:30Z Xof joined #sbcl 2015-02-14T21:09:23Z ChanServ has set mode +o Xof 2015-02-14T21:22:23Z jrm: nyef: Other than taking more time to start up (dotimes (i 1000) (run-shell-command "hostname" nil)) doesn't cause any immediate problems. 2015-02-14T21:25:25Z nyef: I meant (run-shell-command "cat /etc/hostname" t) 2015-02-14T21:26:33Z attila_lendvai quit (Quit: Leaving.) 2015-02-14T21:46:24Z jrm: I tried (dotimes (i 1000) (run-shell-command "cat /etc/hostname" t)) as well and there were no obvious issues. 2015-02-14T21:46:42Z jrm: I'll keep messing around. 2015-02-14T21:50:51Z nyef: It could be that it's something else entirely. /-: 2015-02-14T21:52:44Z angavrilov quit (Remote host closed the connection) 2015-02-14T21:58:45Z attila_lendvai joined #sbcl 2015-02-14T21:58:45Z attila_lendvai quit (Changing host) 2015-02-14T21:58:45Z attila_lendvai joined #sbcl 2015-02-14T22:05:14Z attila_lendvai quit (Remote host closed the connection) 2015-02-14T22:07:31Z |3b|: jrm: next time it crashes might check contents of /tmp/ml-info.txt for anything odd and see if you can tell how many file handles the SBCL process has open 2015-02-14T22:13:18Z attila_lendvai joined #sbcl 2015-02-14T22:13:18Z attila_lendvai quit (Changing host) 2015-02-14T22:13:18Z attila_lendvai joined #sbcl 2015-02-14T22:14:27Z |3b|: jrm: possibly also try turning off the script that updates /tmp/ml-info.txt and the try replacing /tmp/ml-info.txt with progressively smaller prefixes of the normal contents (down to empty file) 2015-02-14T22:15:09Z nyef: ... And it could still be something else entirely, this is just the most obvious tree to bark up, as it were. 2015-02-14T22:16:42Z |3b|: yeah 2015-02-14T22:17:33Z |3b| is suspicious of a race there, but no idea if it could actually have any effect or if it would just not have as much in modeline for a few sec 2015-02-14T22:18:54Z stassats joined #sbcl 2015-02-14T22:19:35Z attila_lendvai quit (Quit: Leaving.) 2015-02-14T22:19:37Z |3b| would also try to arrange to save N hours of ml-info.txt, curious if there might be some unexpected output from any of those command right after waking from sleep or whatever, so being able to check old ones after a crash could rule that out 2015-02-14T22:21:25Z stassats: what was the error again, just a segfault? 2015-02-14T22:21:33Z stassats: this all sounds too interesting 2015-02-14T22:23:03Z |3b| doesn't remember, just that stump was catching it and making it harder to diagnose and it takes forever to reproduce 2015-02-14T22:23:54Z drmeister quit (Read error: Connection reset by peer) 2015-02-14T22:24:46Z stassats: is there a ticket or some other place? 2015-02-14T22:24:55Z stassats: like sbcl-devel 2015-02-14T22:26:01Z |3b|: no idea, i've just been following along in #sbcl 2015-02-14T22:27:50Z jrm: We might have a winner. I'm just confirming. 2015-02-14T22:34:31Z jrm: The shell script, ~/local/bin/ml loops and for each iteration of the loop in overwrites /tmp/ml-info.txt with something like ^[^B^7*gly^] ^[^8*M^] 2016^[^9*M^]/3072^[^9*M^] ^[^8*D^] 0.0^[^9*K/s^] 0.0^[^9*K/s^] ^[^8*re0^] 9.4^[^9*K/s^] 3.1^[^9*K/s^] ^[^8*V^] 64 ^[^B^7*Sat Feb 14 18:34:13^] 2015-02-14T22:36:32Z stassats: so, is that the issue? 2015-02-14T22:36:51Z attila_lendvai joined #sbcl 2015-02-14T22:36:51Z attila_lendvai quit (Changing host) 2015-02-14T22:36:51Z attila_lendvai joined #sbcl 2015-02-14T22:36:53Z stassats: jrm: if you don't solve it today, can you open a ticket with all the information you know? 2015-02-14T22:37:05Z stassats: so that it doesn't fall through the cracks 2015-02-14T22:37:06Z jrm: stassats: yup 2015-02-14T22:37:16Z nyef: ... Non-atomic update of the file in question? 2015-02-14T22:37:30Z attila_lendvai quit (Client Quit) 2015-02-14T22:37:38Z |3b|: nyef: with > 2015-02-14T22:38:05Z jrm: So, I can make stumpwm unresponsive (crash?) immediately by directly trying to capture the output of the script with (setf *screen-mode-line-format* '(:eval (run-shell-command "~/local/bin/ml" t))) 2015-02-14T22:38:10Z nyef: Ah, so the "better" thing to do would be to > into a different file and then mv it into place? 2015-02-14T22:38:50Z |3b|: jrm: how long does it take for the script to run? 2015-02-14T22:39:10Z jrm: |3b|: A second or so. 2015-02-14T22:39:26Z jrm: race condition? 2015-02-14T22:39:34Z attila_lendvai joined #sbcl 2015-02-14T22:39:52Z |3b|: if it waits for that to finish somewhere that blocks UI, a second or so could be too much 2015-02-14T22:40:31Z |3b|: aside from that, if anything running it directly should fix it not make it worse :p 2015-02-14T22:41:02Z |3b|: well, assuming you told it to write the output to stdout instead of the file 2015-02-14T22:41:31Z |3b|: if it didn't write anything to stdout, it could still point to race condition 2015-02-14T22:42:45Z |3b| was wondering if it saw partial input due to race, and some particular combination of partial input confused stump 2015-02-14T22:43:27Z |3b|: or alternately, if the script wrote bad output in rare condition due to unexpected output from one of the programs it called (for example when network or other device isn't up yet after a sleep) 2015-02-14T22:43:45Z drmeister joined #sbcl 2015-02-14T22:45:07Z stassats: so it's a program that write output into a file? 2015-02-14T22:46:48Z jrm: stassats: Yes, it's a shell script, that did write output to a file (one line for StumpWM's mode-line to pick up). That might have been related to crashes that took hours or days. Now it's writing that line directly to standard output and StumpWM is unresponsive immediately. 2015-02-14T22:47:30Z stassats: just one line? 2015-02-14T22:48:20Z stassats: that doesn't sound enough to exhaust the pipe buffers 2015-02-14T22:48:22Z |3b|: jrm: try a script that just cats a static file instead of taking 1sec 2015-02-14T22:48:35Z |3b|: or just cat the static file 2015-02-14T22:49:00Z jrm: k 2015-02-14T22:49:25Z |3b| thinks running the script directly adds too many other factors to be useful for diagnosing previous problem 2015-02-14T22:49:57Z jrm: |3b| OK, but... 2015-02-14T22:50:11Z jrm: My Valentine :-P is going to hurt me if I don't pay some attention to her, so I'll be back soon. 2015-02-14T22:50:44Z |3b| should be doing other things too, so whenever is fine 2015-02-14T22:51:28Z stassats: which seems to be 64K 2015-02-14T22:57:00Z |3b|: jrm: also, just to make sure, the version of the script you were calling directly wasn't still looping, was it? 2015-02-14T22:59:27Z attila_lendvai quit (Quit: Leaving.) 2015-02-14T23:27:16Z gingerale quit (Ping timeout: 255 seconds) 2015-02-14T23:28:27Z Shinmera quit (Quit: しつれいしなければならないんです。) 2015-02-14T23:30:28Z oleo is now known as Guest78328 2015-02-14T23:31:23Z oleo__ joined #sbcl 2015-02-14T23:32:35Z stassats quit (Ping timeout: 246 seconds) 2015-02-14T23:33:07Z stassats` quit (Ping timeout: 255 seconds) 2015-02-14T23:33:34Z Guest78328 quit (Ping timeout: 245 seconds) 2015-02-14T23:53:05Z LiamH quit (Quit: Leaving.) 2015-02-14T23:53:40Z zacts quit (Ping timeout: 250 seconds) 2015-02-15T00:07:06Z Adlai quit (Ping timeout: 250 seconds) 2015-02-15T00:08:21Z pacon quit (Read error: No route to host) 2015-02-15T00:12:05Z Adlai joined #sbcl 2015-02-15T00:28:09Z ASau quit (Ping timeout: 252 seconds) 2015-02-15T00:34:12Z ASau joined #sbcl 2015-02-15T01:09:52Z irsol quit (Ping timeout: 255 seconds) 2015-02-15T01:10:10Z pacon joined #sbcl 2015-02-15T01:13:01Z irsol joined #sbcl 2015-02-15T01:26:16Z csziacobus joined #sbcl 2015-02-15T01:29:42Z Quadrescence joined #sbcl 2015-02-15T01:50:04Z pacon quit (Read error: Connection reset by peer) 2015-02-15T02:02:17Z nyef: ... With this STACK thing, given that the DFO pass numbers in reverse order, I think I'm going to be able to use the phrase "lowest common dominator" somewhere. (-: 2015-02-15T02:05:50Z jrm: |3b|: I did take the loop out yes. 2015-02-15T02:06:51Z jrm: I did get a crash immediately this time: http://ftfl.ca/paste/sbcl.log.html 2015-02-15T02:10:13Z nyef: ... About half of that backtrace is crap, and smack-bang in the middle is an XLIB:DRAWABLE-ERROR. I somehow doubt it's the main issue. 2015-02-15T02:12:08Z nyef: And I'm still amused by "lowest common dominator". (-: 2015-02-15T02:12:11Z |3b|: weren't we looking at a problem when it reconnected after an error though? 2015-02-15T02:12:37Z |3b| doesn't really remember, and is afk again anyway 2015-02-15T02:13:02Z jrm: :/ 2015-02-15T02:14:12Z nyef: IIRC, something happened, there was a memory-fault-error (which trashes the context), the catch-all thing that stumpy has kicked in and trashed the context, and then somewhere a little further down the line things went even further off the rails. 2015-02-15T02:16:23Z jrm: My lack of understanding of the backtraces is weakening my resolve to get to the bottom of this. 2015-02-15T02:19:03Z nyef: The backtrace is SWANK crap until frame 14, internals guts in frame 14 and 15, stumpy getting its say in frame 16, SIGNAL in frame 17, the error in frame 18, and then relatively important stuff starting at frame 19. 2015-02-15T02:19:11Z nyef: There are 34 frames. 2015-02-15T02:19:24Z nyef: (/ 34 2) => 17. 2015-02-15T02:19:33Z nyef: So, approximately half crap. 2015-02-15T02:20:19Z jrm: I also have this: http://ftfl.ca/paste/sbcl-2.log.html 2015-02-15T02:20:43Z nyef: Drop down to frame... 16 and get the details on the error, hopefully it'll indicate the drawable, the sequence number, and some other details. 2015-02-15T02:20:51Z nyef: Yeah, that. 2015-02-15T02:21:23Z nyef: Hrm. DRAWABLE-ERROR on #xC00024. 2015-02-15T02:22:01Z nyef: Ah! Have a look at frames 5 and 6 in that backtrace: The windows have the same ID. 2015-02-15T02:23:12Z nyef: IIRC, DRAWABLE-ERROR is from an invalid value passed as a DRAWABLE (a window or a pixmap), suggesting that the window has already been destroyed. 2015-02-15T02:23:19Z nyef: Or was never created in the first place. 2015-02-15T02:25:11Z nyef: I'm fairly sure there's a utility that should be able to get more information about an arbitrary window, but the name of it is currently escaping me. 2015-02-15T02:25:32Z nyef: Ah! xwininfo. 2015-02-15T02:26:00Z nyef: M-! xwininfo -id 0xc00024 2015-02-15T02:26:24Z nyef: But I rather expect that it'll say that there's no such window. 2015-02-15T02:27:32Z jrm: It did. 2015-02-15T02:29:19Z nyef: Okay, so your mode-line window got destroyed somehow. 2015-02-15T02:29:24Z nyef: ... is this post-reinit? 2015-02-15T02:30:17Z jrm: Yeah, I got kicked out... back to the xdm prompt. 2015-02-15T02:30:19Z nyef: I can see something happening like a large blow-up, the error handler kicking in and reconnecting to the X server, and the old mode-line window object getting retained for whatever reason. 2015-02-15T02:32:52Z nyef: Yeah, same thing with the first backtrace you posted. REDRAW-MODE-LINE on #xc0001b -> RESIZE-MODE-LINE ditto -> XLIB -> XLIB:DRAWABLE-ERROR on #xc0001b. 2015-02-15T02:33:14Z nyef: Followed by 18 frames of noise. 2015-02-15T02:33:32Z jrm: Here is more noise probably: ftfl.ca/paste/sldb.html 2015-02-15T02:34:26Z nyef: Pure signal, same error scenario. 2015-02-15T02:34:42Z nyef: On the other hand, the available restarts suck. 2015-02-15T02:35:49Z nyef: Odd, the mode-line window ID seems to be decreasing over time. 2015-02-15T02:36:19Z nyef: This one is #xe0000a. 2015-02-15T02:37:16Z nyef: Assuming 15 or 16 bit client bits and the rest of the card29 (it was 29, wasn't it?) as the server-assigned client id... 2015-02-15T02:41:12Z nyef: ... There's a timer involved... 2015-02-15T02:42:50Z nyef: Hrm. Can you get me a DESCRIBE on one of those MODE-LINE structures? Or even fire up the inspector on one? 2015-02-15T02:44:38Z jrm: Sorry to be difficult. I know very little about lisp, so I'll either need to take some time to figure out what that means or I'll need more specific instructions. 2015-02-15T02:46:56Z nyef: Damn. And you're using SLIME, quite obviously, and I've stopped bothering with SLIME because it's seemed more trouble than it's worth for me. /-: 2015-02-15T02:47:24Z jrm: I'm trying to learn. 2015-02-15T02:47:35Z jrm: These are my first baby step. 2015-02-15T02:47:38Z jrm: steps 2015-02-15T02:48:20Z jrm: I'm beginning to discover the same thing about SLIME the hard way. When it works it's pretty darn cool to change your WM on the fly. 2015-02-15T02:49:25Z jrm: Can you elaborate on a "DESCRIBE on one of those MODE-LINE structures"? 2015-02-15T02:49:34Z Quadrescence quit (Quit: This computer has gone to sleep) 2015-02-15T02:50:35Z nyef: Okay, let's try this. I found the manual via google. 2015-02-15T02:50:59Z nyef: In sldb, put the cursor on a RESIZE-MODE-LINE or RESIZE-MODE-LINE frame in the backtrace. 2015-02-15T02:51:20Z nyef: Then hit t which should show some local-variableish things. 2015-02-15T02:53:31Z jrm: http://ftfl.ca/paste/slime-inspector.html 2015-02-15T02:53:48Z nyef: Ah! Good. 2015-02-15T02:54:45Z nyef: Hrm. 2015-02-15T02:55:37Z nyef: This is odd. 2015-02-15T02:56:28Z nyef: If you look at the definition of RESIZE-MODE-LINE in mode-line.lisp, you'll see that it starts out by doing a conditional SETF of DRAWABLE-HEIGHT if the MODE is :STUMP (which it is). 2015-02-15T02:57:34Z nyef: Then it does a couple of things that involve DRAWABLE-BORDER-WIDTH, but only AFTER it's set the DRAWABLE-HEIGHT. 2015-02-15T02:59:44Z nyef: So either the (SETF DRAWABLE-HEIGHT) is getting deferred for a bit (that is, it doesn't get sent to the server before the geometry query does) or the window is getting killed between the two. 2015-02-15T03:00:40Z jrm: I see, said the blind man. 2015-02-15T03:01:18Z nyef: ... to his dead dog? 2015-02-15T03:01:19Z nyef: Err. 2015-02-15T03:01:22Z nyef: deaf dog, sorry. 2015-02-15T03:01:42Z nyef: To his DEAF dog. 2015-02-15T03:01:48Z jrm: :) 2015-02-15T03:02:08Z nyef: The other two variants I know of involve woodworking tools and urination. 2015-02-15T03:03:04Z jrm: ..as he picked up his hammer and saw.. and something, something.. 2015-02-15T03:03:35Z nyef: You're able to crash stumpwm this way whenever you want, right? 2015-02-15T03:04:08Z jrm: Yes 2015-02-15T03:04:23Z jrm: let me elaborate a little 2015-02-15T03:05:55Z scymtym_ quit (Ping timeout: 244 seconds) 2015-02-15T03:06:27Z jrm: I can no longer get the rat (does that term still apply to StumpWM or just RatPoison? Does it make sense to you?), so I can't switch windows. I can still evaluate things in a send them to swank and they work as expected. E.G. toggle the mode line, or (message "blah"). 2015-02-15T03:06:30Z nyef: Hrm. I'm not seeing any obvious way to kick CLX into a "synchronous" mode like could be done with XLib. 2015-02-15T03:06:58Z jrm: s/in a/and/ 2015-02-15T03:07:47Z nyef: Can you easily get the system to this point from a clean start, or haven't you tried that yet? 2015-02-15T03:08:07Z jrm: Err, so I could probably evaluate things like (run-or-raise "urxvtc") to switch windows. 2015-02-15T03:08:33Z jrm: Oh shit, that was a bad idea. 2015-02-15T03:08:45Z jrm: Now, how to get back to emacs. 2015-02-15T03:09:01Z nyef: You have a shell prompt? 2015-02-15T03:09:12Z jrm: Yeah, I'm in urxvtc now. 2015-02-15T03:09:26Z nyef: Can you at least see the emacs window, or do we need to do this the hard way? 2015-02-15T03:09:48Z jrm: I cannot see any window other than urxvtc. 2015-02-15T03:10:20Z jrm: I'm at a vt. 2015-02-15T03:11:00Z jrm: If I had run an emacsclient/server I could reconnect, but I think I just ran emacs. 2015-02-15T03:11:36Z nyef: Hrm. Given an x window id, how can I rasise it from a shell prompt? 2015-02-15T03:12:10Z jrm: https://stackoverflow.com/questions/1028972/x11-raise-an-existing-window-via-command-line 2015-02-15T03:12:56Z nyef: A good start. 2015-02-15T03:13:23Z nyef: So, you should be able to use "xlsclients -root -childrent" to find the window id of your emacs... 2015-02-15T03:14:31Z nyef: Err... "xlsclients -root -children". I don't know where that #\t came from. 2015-02-15T03:15:56Z jrm: Hmm.. only shows other apps running clipit and urxvt 2015-02-15T03:16:22Z nyef: That's odd... Is the emacs still showing up in ps ? 2015-02-15T03:16:43Z jrm: yup 2015-02-15T03:16:53Z nyef: I've got no idea, then. 2015-02-15T03:17:14Z jrm: I had to run xlsclients -display :0.0 -l -a 2015-02-15T03:17:23Z jrm: jyes? 2015-02-15T03:19:34Z nyef: Wait, why the -display argument? 2015-02-15T03:19:58Z nyef: I thought you were at a urxvt? 2015-02-15T03:20:10Z nyef: Or did the keyboard focus get stuck? 2015-02-15T03:20:47Z jrm: -root and -children syntax gave usage: xlsclients [-display dpy]... 2015-02-15T03:21:00Z jrm: I am at urxvt 2015-02-15T03:21:43Z jrm: The valentine went to bed, so I snagged her laptop, so I'm talking to you from that. 2015-02-15T03:22:02Z nyef: Oh, drat. Did I say xlsclients? I meant xwininfo. 2015-02-15T03:22:24Z nyef: I so rarely interact with X at this level. 2015-02-15T03:22:51Z nyef: Still, the display information should be in the environment, not on the command line. 2015-02-15T03:24:38Z jrm: interesting: with no args: xwininfo: error can't grab the mouse. I also can't select any text. xwininfo -root -children shows 13 children, most with no name. 2015-02-15T03:25:08Z jrm: Since I can cause this issue right away, would it be easier to just C-M-f1 pkill stumpwm and start over? 2015-02-15T03:25:24Z nyef: May as well. 2015-02-15T03:26:28Z jrm: Ok, I'm back with the default, empty mode-line. 2015-02-15T03:26:55Z nyef: How do you trigger this, btw? 2015-02-15T03:27:44Z jrm: evaluate (setf *screen-mode-line-format* '(:eval (run-shell-command "~/local/bin/ml" t))) 2015-02-15T03:28:17Z nyef: So it's something odd about your script? 2015-02-15T03:29:19Z jrm: I thin |3b| suggested it might be because it takes a moment to run and there is some race condition before the mode-line updates. 2015-02-15T03:29:21Z zacts joined #sbcl 2015-02-15T03:30:09Z jrm: |3b|: also suggested that evaluating the command directly might introduce new, independent problems 2015-02-15T03:30:22Z jrm: if I understood correctly 2015-02-15T03:30:40Z nyef: Ah. Fun and games. 2015-02-15T03:31:14Z jrm: This is what the *slime-events* buffer looks like before any problems start. 2015-02-15T03:35:23Z jrm: Oh, the URL would help: http://ftfl.ca/paste/slime-events.html 2015-02-15T03:37:31Z nyef: ... When does the mode-line kick in? 2015-02-15T03:39:22Z christoph_debian quit (Ping timeout: 240 seconds) 2015-02-15T03:40:05Z jrm: I start it at the very end of ~/.stumpwmrc. This slime-events buffer gets created when I do M-x slime-connect, so the mode-line is already displayed, but it's a very simple, default mode-line now: http://ftfl.ca/paste/stumpwmrc.html 2015-02-15T03:40:59Z jrm: It's not reading anything from a shell script or the output from a shell script. When I evaluate (setf *screen-mode-line-format* '(:eval (run-shell-command "~/local/bin/ml" t))) problems will start. 2015-02-15T03:42:56Z nyef: I just did a setf of *screen-mode-line-format* similarly, but I'm not seeing a visible effect. 2015-02-15T03:43:11Z nyef: Preliminary conclusion: The mode-line isn't getting shown for me for some reason. 2015-02-15T03:43:28Z jrm: evaluate (toggle-mode-line (current-screen) (current-head)) 2015-02-15T03:43:34Z nyef: Yeah, just saw that. 2015-02-15T03:44:26Z jrm: Hmm... we must have added this: (setf *top-level-error-action* :break) 2015-02-15T03:46:05Z nyef: What does *mode-line-timeout* control? 2015-02-15T03:46:21Z jrm: Interesting: http://www.emacswiki.org/emacs/Comments_on_StumpWM 2015-02-15T03:48:39Z jrm: I might be seeing this: "This can be problematic because if the user hit's a mapped key the ENTIRE keyboard will be frozen and you will have to login remotely to regain control. :abort quits stumpmwm." 2015-02-15T03:49:08Z jrm: Err, the entire keyboard isn't frozen. 2015-02-15T03:50:50Z nyef: Depends on what's going on. If it dumps to a debugger and your focus is somewhere like a web browser, and there's nothing that takes control and forces the debugger to be visible and have keyboard focus, then you might have a problem or two. 2015-02-15T03:52:34Z jrm: I'll initiate the problem now and if there is nothing new in any of the debugging messages, or you have no new ideas, I'll try taking out the mode-line, but keep the (swank:create-server) and see if I have problems. 2015-02-15T03:52:35Z christoph_debian joined #sbcl 2015-02-15T03:53:01Z nyef: We might be at that point. /-: 2015-02-15T03:55:40Z jrm: Not sure if there is anything new here: http://ftfl.ca/paste/sldb.html 2015-02-15T03:56:11Z nyef: Nope, that's the same thing. 2015-02-15T03:56:56Z nyef: There's the resize-mode-line and redraw-mode-line in frames 5 and 6, the xlib:drawable-border-width in frame 4, and the xlib:drawable-error as the condition. 2015-02-15T03:57:19Z nyef: And it's from update-all-mode-lines in frame 8, and run-expired-timers in frame 9. 2015-02-15T03:58:28Z jrm: So this is what happened. I evaluated (setf *screen-mode-line-format* '(:eval (run-shell-command "~/local/bin/ml" t))), but it actually worked. The mode-line was updating and I didn't see any errors. Then I toggled the mode-line and then the sldb buffer came up. 2015-02-15T04:00:52Z nyef: Ahh... 2015-02-15T04:02:41Z jrm: I should say more about, but it actually worked... The mode line worked, but no rat. 2015-02-15T04:02:57Z nyef: "rat"? 2015-02-15T04:03:28Z jrm: In stumpwm, when you hit the prefix key, the mouse cursor changes to let you know it's ready to accept stumpwm commands. 2015-02-15T04:03:34Z nyef: Ah. 2015-02-15T04:03:46Z jrm: The square mouse cursor 2015-02-15T04:04:05Z jrm: That's what it was called in rat poison. I think that's what it's referred to in StumpWM. 2015-02-15T04:10:01Z Quadrescence joined #sbcl 2015-02-15T04:12:02Z nyef: I'm still failing to crash stumpwm, even with a hideous delay in the mode-line script and toggling the mode-line a couple of times. 2015-02-15T04:13:06Z jrm: And you still get the rat? You asked me whether sbcl was built with asynch at one point... 2015-02-15T04:13:16Z nyef: The mode-line is interesting, though I'm not at all convinced about the rest of the interaction style. 2015-02-15T04:13:30Z nyef: Yeah, neither of us are running :sb-thread. 2015-02-15T04:15:11Z nyef: Yeah, with the mode-line disabled the rat shows up fairly close to instantly. With it enabled the lag is so bad that I keep losing track of what's going on, as does stumpwm (I can type ahead so fast that the input goes to the terminal window instead of to stumpwm). 2015-02-15T04:15:26Z nyef: Actually, that strongly suggests that there's a bug in the keyboard grab logic. 2015-02-15T04:20:51Z jrm: hm.. uni is having connection problems 2015-02-15T04:21:54Z nyef: minion: remove-if 2015-02-15T04:21:54Z minion: Sorry, I couldn't find anything in the database for ``remove-if''. 2015-02-15T04:21:57Z nyef: clhs remove-if 2015-02-15T04:21:57Z specbot: http://www.lispworks.com/reference/HyperSpec/Body/f_rm_rm.htm 2015-02-15T04:29:42Z |3b| quit (Read error: Connection reset by peer) 2015-02-15T04:30:56Z |3b| joined #sbcl 2015-02-15T04:43:37Z edgar-rft quit (Quit: memory access abandoned by timeout after 899588584273388496226 milliseconds) 2015-02-15T04:55:23Z PuercoPop quit (Ping timeout: 245 seconds) 2015-02-15T04:55:37Z PuercoPop joined #sbcl 2015-02-15T04:56:13Z foom quit (Ping timeout: 245 seconds) 2015-02-15T04:56:40Z foom joined #sbcl 2015-02-15T05:08:50Z nyef quit (Quit: G'night all) 2015-02-15T05:21:42Z Quadrescence quit (Quit: This computer has gone to sleep) 2015-02-15T05:40:24Z nightshade427 quit (Ping timeout: 245 seconds) 2015-02-15T05:40:35Z nightshade427 joined #sbcl 2015-02-15T05:40:42Z |3b| quit (Read error: Connection reset by peer) 2015-02-15T05:41:04Z nicdev quit (Read error: Connection reset by peer) 2015-02-15T05:41:28Z nicdev joined #sbcl 2015-02-15T05:41:57Z |3b| joined #sbcl 2015-02-15T06:30:23Z csziacobus quit (Quit: csziacobus) 2015-02-15T07:13:09Z gingerale joined #sbcl 2015-02-15T07:42:30Z Shinmera joined #sbcl 2015-02-15T08:18:32Z angavrilov joined #sbcl 2015-02-15T08:35:23Z wasabiz__ quit (Ping timeout: 245 seconds) 2015-02-15T08:49:29Z scymtym_ joined #sbcl 2015-02-15T09:16:38Z zuq joined #sbcl 2015-02-15T09:19:02Z yauz_2 joined #sbcl 2015-02-15T09:22:14Z zuq quit (Remote host closed the connection) 2015-02-15T09:31:42Z yauz quit (*.net *.split) 2015-02-15T09:39:50Z fridim_ joined #sbcl 2015-02-15T09:47:28Z PuercoPop quit (Ping timeout: 245 seconds) 2015-02-15T09:48:03Z PuercoPop joined #sbcl 2015-02-15T10:00:55Z fridim_ quit (Ping timeout: 250 seconds) 2015-02-15T10:11:32Z fridim_ joined #sbcl 2015-02-15T10:40:48Z foom quit (Ping timeout: 245 seconds) 2015-02-15T10:42:35Z foom joined #sbcl 2015-02-15T10:43:16Z edgar-rft joined #sbcl 2015-02-15T11:19:19Z stassats joined #sbcl 2015-02-15T11:59:59Z oleo joined #sbcl 2015-02-15T12:02:22Z oleo__ quit (Ping timeout: 255 seconds) 2015-02-15T12:35:44Z attila_lendvai joined #sbcl 2015-02-15T12:35:57Z stassats` joined #sbcl 2015-02-15T13:00:54Z stassats`: i think i got foreign thread callbacks on non-sb-safepoint 2015-02-15T13:02:20Z attila_lendvai quit (Quit: Leaving.) 2015-02-15T13:06:54Z attila_lendvai joined #sbcl 2015-02-15T13:25:13Z attila_lendvai quit (Quit: Leaving.) 2015-02-15T13:25:44Z stassats`: that was really easy, i should have done that a couple of years ago 2015-02-15T13:27:00Z attila_lendvai joined #sbcl 2015-02-15T13:30:09Z stassats`: and it's in 2015-02-15T13:31:01Z stassats`: this release seems to be good, lots of good fixes 2015-02-15T13:36:19Z Xof: famous last words 2015-02-15T13:36:30Z Xof: now it will also include several brown paper bag bugs 2015-02-15T13:37:55Z stassats`: segfaults are unrecoverable on ppc, looks like the same issue as with stack exhaustion previously 2015-02-15T13:38:08Z stassats`: make that heap exhaustion 2015-02-15T14:19:48Z LiamH joined #sbcl 2015-02-15T14:21:09Z nyef joined #sbcl 2015-02-15T14:29:19Z attila_lendvai quit (Quit: Leaving.) 2015-02-15T14:38:26Z leuler joined #sbcl 2015-02-15T14:53:55Z eudoxia joined #sbcl 2015-02-15T14:58:37Z mega1 quit (Ping timeout: 252 seconds) 2015-02-15T15:35:39Z stassats` quit (Remote host closed the connection) 2015-02-15T15:36:42Z stassats` joined #sbcl 2015-02-15T15:40:10Z eudoxia quit (Quit: Leaving) 2015-02-15T15:43:44Z attila_lendvai joined #sbcl 2015-02-15T15:45:06Z attila_lendvai quit (Client Quit) 2015-02-15T16:12:31Z attila_lendvai joined #sbcl 2015-02-15T16:12:31Z attila_lendvai quit (Changing host) 2015-02-15T16:12:31Z attila_lendvai joined #sbcl 2015-02-15T16:15:25Z attila_lendvai quit (Client Quit) 2015-02-15T16:16:41Z stassats`: ok, i think i'm done with combing through old tickets 2015-02-15T16:21:14Z attila_lendvai joined #sbcl 2015-02-15T16:25:38Z nyef: I got the backprop part of STACK to work right... I think. It still blows up in the same place, but that's another matter. 2015-02-15T16:25:56Z nyef: Well, it's because the ordering logic can't handle what's going on, really. 2015-02-15T16:26:34Z nyef: stassats`: Do you happen to understand the USE-GOOD-FOR-DX-P noise? 2015-02-15T16:28:41Z nyef: Ah. Heh. "because all UVLs being live at the BLOCK start are live in PRED" wrong! 2015-02-15T16:29:16Z stassats`: which part of use-good-for-dx-p? 2015-02-15T16:29:29Z nyef: The overall semantic. 2015-02-15T16:30:45Z nyef: Basically, I'm thinking that there are three possibilities for something being "good for dx". Either it IS "good" and would like to be DXed, it doesn't care and can go either way, or it ISN'T "good" and the DX must not happen. 2015-02-15T16:31:22Z nyef: I don't know if that third case actually exists or not. 2015-02-15T16:32:08Z stassats`: we have truly-dynamic-extent and just dynamic-extent, but when that's decided, not sure 2015-02-15T16:32:19Z nyef: But something like a constant would fall into the "doesn't care" case. It shouldn't forbid DX but it can't license DX either. 2015-02-15T16:32:40Z nyef: Whereas an invocation of CONS where the result is declared DX does license DX. 2015-02-15T16:32:43Z attila_lendvai quit (Quit: Leaving.) 2015-02-15T16:33:31Z nyef: Basically (if (cons nil nil) 42) should be able to DX the CONS, and I don't believe that it can at this point. 2015-02-15T16:33:40Z attila_lendvai joined #sbcl 2015-02-15T16:33:42Z attila_lendvai quit (Client Quit) 2015-02-15T16:33:57Z stassats`: i think the difference between always-dynamic vs maybe-dynamic is when allocating vectors with unknown size 2015-02-15T16:34:05Z nyef: Admittedly, right now such a thing would blow up horribly in STACK, but I'm working on that. 2015-02-15T16:39:15Z stassats`: we actually can be smart about things like returning dx vars 2015-02-15T16:40:17Z nyef: Ah, well. More of an IWBNI than an outright bug. 2015-02-15T16:51:02Z nyef: ... Lovely. There's a partial-ordering issue in this mess as well. 2015-02-15T16:59:30Z stassats`: building with ccl no longer works again 2015-02-15T16:59:42Z stassats`: now it's actually a memory corruption 2015-02-15T17:02:01Z Xof: theirs or ours? 2015-02-15T17:02:28Z stassats`: who knows 2015-02-15T17:02:38Z stassats`: theirs but caused by us? 2015-02-15T17:04:39Z stassats`: because it's in xc 2015-02-15T17:08:33Z Xof: so ccl executing our cross-compiler code? 2015-02-15T17:08:40Z stassats`: right 2015-02-15T17:12:45Z stassats`: managed to get a backtrace, coming from MAKE-VALUES-TYPE-CACHED 2015-02-15T17:17:46Z Xof: that has been changing recently 2015-02-15T17:18:45Z stassats`: 2305843009214459243 doesn't appear to be fixnum 2015-02-15T17:20:46Z Xof: some fixnum vs sb!xc:fixnum declaration? 2015-02-15T17:20:58Z Xof: I need to bake the reproducible build stuff into the build process 2015-02-15T17:21:21Z stassats`: (hash-value ... :type (signed-byte #.sb!vm:n-fixnum-bits)) 2015-02-15T17:21:29Z stassats`: and then (defun type-list-cache-hash (types) (loop with res fixnum = 0 for type in types do (setq res (logxor (ash res -1) (type-hash-value type))) finally (return res))) 2015-02-15T17:21:58Z stassats`: and it has fixnum all over it 2015-02-15T17:26:26Z stassats`: mailed the problem to Douglas 2015-02-15T17:26:54Z stassats`: and there i just wanted to close an old "CCL hosted build broken" 2015-02-15T17:27:13Z stassats`: but there's usually a good chance that it gets broken again each time you want to close it 2015-02-15T17:27:20Z stassats`: so maybe it's a good idea to keep it always open 2015-02-15T18:20:45Z flip214: stassats`: regarding https://github.com/edicl/hunchentoot/blob/b32cdee0154b81aa754246d113dbf551f414dc22/util.lisp#L186 ... 2015-02-15T18:21:16Z flip214: why not "(defun ensure-parse-integer (string &key (start 0) (end (length string)) (radix 10))" 2015-02-15T18:21:38Z flip214: instead of the "(LET ((end (or end ..."? readability? 2015-02-15T18:22:00Z nyef: flip214: Presumably, the semantic is that an explicitly-supplied NIL should still mean the length of the string. 2015-02-15T18:22:35Z flip214: nyef: ah, yes. right. 2015-02-15T18:22:40Z flip214: thanks for explaining. 2015-02-15T18:22:57Z flip214: but why not do the same thing for start, then?# 2015-02-15T18:23:22Z flip214: and/or radix? NIL wouldn't work for all of these... 2015-02-15T18:27:07Z nyef: There's a difference between a default value and an explicit semantic for NIL. 2015-02-15T18:33:08Z stassats`: it just has the same semantics as parse-integer 2015-02-15T18:36:49Z Quadrescence joined #sbcl 2015-02-15T18:37:33Z Quadrescence quit (Remote host closed the connection) 2015-02-15T18:41:03Z nyef: Ugh. So, lifetime is a backwards flow analysis, but ordering is a forwards flow analysis that can turn around and declare things that had been declared as live as now being dead. 2015-02-15T18:41:27Z stassats`: the flowing dead? 2015-02-15T18:41:32Z nyef: Something like that. 2015-02-15T18:42:33Z stassats` didn't reach the constant propagation problem today, maybe next weekend 2015-02-15T18:43:13Z nyef: Mmm. And this is one of those cases where I thought I had it, only to find out that there's an entire second level to deal with. 2015-02-15T18:45:44Z nyef: Ooh. If you turn on *check-consistency* the blown AVER changes from (subsetp end end-stack) to (subsetp start start-stack). 2015-02-15T18:49:11Z leuler: I'm currently overhauling the way NOPs are emitted on x86[-64] to fill the space between the code and the unboxed constants. There is a bare 64 in emit-constant-segment-header which is presumably the cache line size. Where would I put a constant defining this? src/compiler/x86[-64]/parms.lisp? 2015-02-15T18:50:25Z stassats`: wherever it works 2015-02-15T18:51:09Z stassats`: but surely different x86-64s have different cache lines 2015-02-15T18:51:56Z nyef: Stupid misleading (and currently-wrong) comment: It's not that we need to delete dead UVLs, it's that we need to delete KILLED ones. 2015-02-15T18:53:14Z leuler: stassats`: Not as far as I know. 2015-02-15T18:59:54Z Quadrescence joined #sbcl 2015-02-15T19:00:49Z stassats`: then x86 2015-02-15T19:10:32Z stassats`: leuler: but can you measure the effects of your nopes? 2015-02-15T19:10:52Z stassats`: what about trap data? 2015-02-15T19:10:58Z stassats`: we can actually store them out of line 2015-02-15T19:12:38Z nyef: clhs every 2015-02-15T19:12:38Z specbot: http://www.lispworks.com/reference/HyperSpec/Body/f_everyc.htm 2015-02-15T19:14:33Z leuler: stassats`: I intend just to change the code in emit-constant-segment-header to emit NOPs for all the space before the constants. Currently it leaves part of that empty (zeroes) which makes the disassembly ugly. The idea to put code and data into different cache lines when optimising for speed has been there already and I don't want to change that. 2015-02-15T19:15:19Z stassats`: but can you measure it? 2015-02-15T19:15:34Z leuler: You want to rip that out? 2015-02-15T19:16:08Z stassats`: i want to measure it! 2015-02-15T19:24:27Z akkad quit (Ping timeout: 245 seconds) 2015-02-15T19:27:59Z leuler: I don't want to measure it. My intended change is just to make the disassembly more readable. You can measure that by looking at the output. The remaining code stays as it is; it follows general optimization guidelines; that's fine with me. 2015-02-15T19:29:23Z akkad joined #sbcl 2015-02-15T20:52:26Z Bicyclidine joined #sbcl 2015-02-15T20:53:19Z prxq joined #sbcl 2015-02-15T21:24:08Z fridim_ quit (Ping timeout: 250 seconds) 2015-02-15T21:41:02Z leuler quit (Quit: ERC Version 5.1.2 $Revision: 1.796.2.6 $ (IRC client for Emacs)) 2015-02-15T21:46:02Z nyef: stassats`: I managed to get the conditional-DX thing to fail with a (newly added) AVER in DISCARD-UNUSED-VALUES. 2015-02-15T21:46:59Z stassats: what about not an aver but just disallowing dx? 2015-02-15T21:47:10Z stassats: in use-good-for-dx-p or something 2015-02-15T21:48:09Z nyef: What about this being the point where I need to add cleanup logic to make conditional DX actually work? 2015-02-15T21:48:38Z stassats: actually working is actually better 2015-02-15T21:49:13Z nyef: Also means we might be able to relax the constraints on USE-GOOD-FOR-DX-P and DX-allocate more things. 2015-02-15T21:52:13Z stassats: (declaim (dynamic-extent x)) => WARNING: unrecognized declaration (DYNAMIC-EXTENT X) 2015-02-15T21:52:16Z stassats: is it really unrecognized? 2015-02-15T21:55:43Z nyef: I have no idea? 2015-02-15T21:55:55Z nyef: It probably shouldn't be unrecognized, even if it's meaningless at toplevel. 2015-02-15T21:56:29Z stassats: that's what i'm saying 2015-02-15T21:56:42Z Xof: stassats: did you get my mail about trunacte? 2015-02-15T21:56:52Z stassats: Xof: right, will sneak into some other commit 2015-02-15T21:56:54Z Xof: ok, good 2015-02-15T21:57:13Z Xof: the answer to "what did I forget when switching off my old server" was "exim configuration" 2015-02-15T21:57:25Z Xof: so you were my first successful correspondent :-) 2015-02-15T21:57:29Z stassats: or maybe i should just do it and boost my commit count 2015-02-15T21:58:05Z stassats: ok, i'll do it now or i'll forget it 2015-02-15T21:59:22Z stassats: ok, done, thanks for the notice 2015-02-15T22:00:05Z Shinmera quit (Quit: しつれいしなければならないんです。) 2015-02-15T22:06:40Z Adlai quit (Ping timeout: 252 seconds) 2015-02-15T22:08:24Z angavrilov quit (Remote host closed the connection) 2015-02-15T22:10:30Z Adlai joined #sbcl 2015-02-15T22:10:33Z ejbs joined #sbcl 2015-02-15T22:12:20Z ejbs: stassats: there's a typo in the code here (trunacte): 2015-02-15T22:12:20Z ejbs: https://github.com/sbcl/sbcl/commit/fe29e71cf0b86d7b3dbbd6e37262bafd7d3c707a 2015-02-15T22:12:25Z ejbs: Oops, sorry 2015-02-15T22:12:55Z ejbs: stassats` 2015-02-15T22:13:07Z stassats: oops, too late 2015-02-15T22:13:08Z minion: stassats, memo from ejbs: there's a typo in the code here (trunacte): https://github.com/sbcl/sbcl/commit/fe29e71cf0b86d7b3dbbd6e37262bafd7d3c707a - I've also mentioned this in #sbcl 2015-02-15T22:13:35Z stassats: ejbs: thanks anyway 2015-02-15T22:14:15Z ejbs: stassats: No problem 2015-02-15T22:21:55Z csziacobus joined #sbcl 2015-02-15T22:36:14Z edgar-rft quit (Quit: connection lost into permanent deadlock) 2015-02-15T22:43:20Z dmitigr joined #sbcl 2015-02-15T22:57:16Z gingerale quit (Ping timeout: 252 seconds) 2015-02-15T23:00:44Z ejbs` joined #sbcl 2015-02-15T23:03:05Z ejbs quit (Ping timeout: 246 seconds) 2015-02-15T23:04:04Z nyef: stassats: Does http://paste.lisp.org/display/145772 look like a win to you? 2015-02-15T23:04:37Z Quadrescence quit (Ping timeout: 264 seconds) 2015-02-15T23:07:48Z Quadrescence joined #sbcl 2015-02-15T23:09:23Z prxq quit (Remote host closed the connection) 2015-02-15T23:15:43Z stassats: too late for me 2015-02-15T23:16:06Z stassats: time to read some cold war nuclear stories before bed, not sbcl traces 2015-02-15T23:17:34Z nyef: Enjoy. 2015-02-15T23:17:49Z nyef: I'll finish packaging up commits for this and get it pushed somewhere. 2015-02-15T23:23:01Z stassats` quit (Ping timeout: 250 seconds) 2015-02-15T23:23:13Z stassats quit (Ping timeout: 255 seconds) 2015-02-15T23:29:43Z ejbs` quit (Ping timeout: 264 seconds) 2015-02-16T00:34:34Z oleo is now known as Guest75895 2015-02-16T00:35:29Z oleo__ joined #sbcl 2015-02-16T00:38:13Z Guest75895 quit (Ping timeout: 264 seconds) 2015-02-16T00:56:10Z Posterdati quit (Quit: KVIrc 4.1.3 Equilibrium http://www.kvirc.net/) 2015-02-16T01:15:19Z dmitigr quit (Ping timeout: 264 seconds) 2015-02-16T01:24:42Z Eschatologist joined #sbcl 2015-02-16T01:26:32Z Eschatologist quit (Client Quit) 2015-02-16T01:48:05Z echo-area joined #sbcl 2015-02-16T02:15:55Z Adlai quit (Ping timeout: 264 seconds) 2015-02-16T02:19:33Z Adlai joined #sbcl 2015-02-16T02:23:55Z Bicyclidine quit (Quit: leaving) 2015-02-16T02:50:48Z zacts quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2015-02-16T03:13:16Z scymtym_ quit (Ping timeout: 265 seconds) 2015-02-16T03:38:57Z christoph_debian quit (Ping timeout: 246 seconds) 2015-02-16T03:52:15Z christoph_debian joined #sbcl 2015-02-16T05:33:21Z LiamH quit (Quit: Leaving.) 2015-02-16T05:35:49Z pranavrc joined #sbcl 2015-02-16T05:49:06Z nyef quit (Quit: G'night all) 2015-02-16T06:03:47Z csziacobus quit (Quit: csziacobus) 2015-02-16T06:48:11Z gingerale joined #sbcl 2015-02-16T07:28:50Z pranavrc quit 2015-02-16T07:29:19Z pranavrc joined #sbcl 2015-02-16T07:29:19Z pranavrc quit (Changing host) 2015-02-16T07:29:19Z pranavrc joined #sbcl 2015-02-16T08:01:04Z angavrilov joined #sbcl 2015-02-16T08:07:29Z pranavrc quit (Read error: Connection reset by peer) 2015-02-16T08:07:54Z pranavrc joined #sbcl 2015-02-16T08:07:54Z pranavrc quit (Changing host) 2015-02-16T08:07:54Z pranavrc joined #sbcl 2015-02-16T08:18:34Z fridim_ joined #sbcl 2015-02-16T09:06:48Z Shinmera joined #sbcl 2015-02-16T09:22:33Z echo-are` joined #sbcl 2015-02-16T09:23:43Z echo-area quit (Ping timeout: 264 seconds) 2015-02-16T09:36:18Z psy_ quit (Ping timeout: 246 seconds) 2015-02-16T09:52:46Z Shinmera quit (Quit: しつれいしなければならないんです。) 2015-02-16T09:59:22Z Shinmera joined #sbcl 2015-02-16T10:20:42Z dmitigr joined #sbcl 2015-02-16T10:45:22Z edgar-rft joined #sbcl 2015-02-16T10:57:33Z stassats joined #sbcl 2015-02-16T11:16:02Z Shinmera quit (Quit: しつれいしなければならないんです。) 2015-02-16T11:19:02Z Shinmera joined #sbcl 2015-02-16T11:26:55Z dmitigr` joined #sbcl 2015-02-16T11:30:20Z dmitigr quit (Ping timeout: 246 seconds) 2015-02-16T11:33:42Z echo-are` quit (Remote host closed the connection) 2015-02-16T11:47:32Z Posterdati joined #sbcl 2015-02-16T12:18:32Z Posterdati quit (Quit: KVIrc 4.1.3 Equilibrium http://www.kvirc.net/) 2015-02-16T12:22:42Z Posterdati joined #sbcl 2015-02-16T12:39:47Z christoph_debian: ahoi! the sbcl.core seems to expose a pretty patterned delte between two builds (https://reproducible.debian.net/rb-pkg/sbcl.html scroll down to sbcl.core) -- can anyone explain what this is? can I get it away somehow? (people here started to care about binary-reproducible builds and are asking) 2015-02-16T12:42:26Z Xof: I did some reproducible build work a couple of months ago 2015-02-16T12:42:33Z Xof: it doesn't fix everything but it does fix some things 2015-02-16T12:42:47Z Xof: one leak from the host that still persists is the build directory 2015-02-16T12:43:47Z Xof: but I think 1.2.4 is before most of that work 2015-02-16T12:43:48Z flip214: Xof: here the issue is that the binary still has different bytes in it, after compilation on the same host in the same directory. IIUC. 2015-02-16T12:44:12Z Xof: see http://christophe.rhodes.io/notes/blog/posts/2014/reproducible_builds_-_a_month_ahead_of_schedule/ for gory details 2015-02-16T12:47:21Z christoph_debian: Xof: flip214: thanks for pointers already! 2015-02-16T12:48:28Z Xof: christoph_debian: with a bit more work (maybe a day or two) I think I can probably finish things up 2015-02-16T12:51:23Z flip214: christoph_debian: thanks for doing the debian packages! 2015-02-16T12:53:09Z christoph_debian: :-) 2015-02-16T12:58:52Z Shinmera quit (Quit: しつれいしなければならないんです。) 2015-02-16T13:01:13Z pranavrc quit 2015-02-16T13:06:39Z pranavrc joined #sbcl 2015-02-16T13:06:39Z pranavrc quit (Changing host) 2015-02-16T13:06:39Z pranavrc joined #sbcl 2015-02-16T13:06:45Z pranavrc quit (Remote host closed the connection) 2015-02-16T13:08:27Z flip214: with current git head I get "Failed to find the TRUENAME of output/stuff-groveled-from-headers.lisp" when doing an NFS-based crosscompile ... 2015-02-16T13:11:44Z flip214: hmmm, disregard that ... seems to have been some unclean situation in the working copy. (?) 2015-02-16T13:11:59Z stassats` joined #sbcl 2015-02-16T13:17:30Z mega1 joined #sbcl 2015-02-16T13:20:05Z oleo joined #sbcl 2015-02-16T13:22:20Z oleo__ quit (Ping timeout: 246 seconds) 2015-02-16T13:24:16Z Shinmera joined #sbcl 2015-02-16T13:39:32Z nyef joined #sbcl 2015-02-16T13:55:01Z flip214: Is "//checking for leftover cold-init symbols" ... "Found 47:" anything to worry about? 2015-02-16T13:58:01Z dmitigr` quit (Ping timeout: 264 seconds) 2015-02-16T13:58:18Z |3b| quit (Read error: Connection reset by peer) 2015-02-16T13:59:33Z |3b|` joined #sbcl 2015-02-16T14:32:45Z dmitigr joined #sbcl 2015-02-16T15:10:23Z LiamH joined #sbcl 2015-02-16T15:19:18Z nyef: stassats: Did you say that lp#1044465 and lp#1255782 were the last of the known outright bugs in the compiler? 2015-02-16T15:28:00Z stassats`: there are some more 2015-02-16T15:28:30Z stassats`: lp 1234919 2015-02-16T15:28:30Z specbot: https://bugs.launchpad.net/bugs/1234919 2015-02-16T15:30:37Z stassats`: seems easy enough, though 2015-02-16T15:30:58Z nyef: Hrm. 2015-02-16T15:38:54Z stassats`: but first for some day job 2015-02-16T16:03:35Z joshe joined #sbcl 2015-02-16T16:05:45Z nyef: Trying to build git HEAD on x86-64/linux: The value -2305843009204879977 is not of type FIXNUM. 2015-02-16T16:06:26Z stassats: 32-bit host? 2015-02-16T16:06:27Z nyef: Trying to load from-host/src/code/class-init.fasl in host-1. 2015-02-16T16:06:53Z nyef: 64-bit host. 2015-02-16T16:07:08Z nyef: Three fixnum tag bits. 2015-02-16T16:07:08Z stassats: from which era? 2015-02-16T16:07:11Z stassats: right 2015-02-16T16:07:23Z stassats: but i get the same thing with a 32-bit target 2015-02-16T16:07:27Z stassats: host 2015-02-16T16:07:29Z flip214: Hmmm, changing 3 DEFCLASS to 3 DEFSTRUCT makes the TIME output write "46,660 forms interpreted" and double the processor cycle count ... 2015-02-16T16:09:36Z stassats: nyef: reported to Douglas 2015-02-16T16:10:52Z nyef: So much for getting the conditional DX thing ready for review this morning. 2015-02-16T16:11:12Z stassats: just build a 32-bit sbcl then 2015-02-16T16:11:31Z stassats: if your gcc can stomach it 2015-02-16T16:12:57Z nyef: Let's see if bumping n-fixnum-tag-bits to 3 clears it up... (-:< 2015-02-16T16:13:23Z stassats: we should have no fixnums declrations in the cross compiler, really 2015-02-16T16:13:42Z stassats: or any declarations which do not include size, like base-char 2015-02-16T16:14:22Z stassats: or they should at least be consistent 2015-02-16T16:14:40Z stassats: because i think this problem is because of fixnum and (signed-byte n-fixnum-bits) 2015-02-16T16:15:21Z Xof: it's hard to catch 2015-02-16T16:15:29Z Xof: apart from the full reproducible build test 2015-02-16T16:17:11Z flip214: (defstruct struct-a () (int 1 :type integer)) 2015-02-16T16:17:13Z nyef: ... Blows up writing from-xc/src/code/numbers.lisp-obj. 2015-02-16T16:17:15Z flip214: WARNING: slot name of NIL indicates probable syntax error in DEFSTRUCT 2015-02-16T16:17:39Z flip214: says sbcl 1.2.4 on debian. 2015-02-16T16:18:19Z stassats: that is true 2015-02-16T16:19:15Z flip214: stassats: ahhhh, right.... the () is left-over from DEFCLASS. 2015-02-16T16:20:04Z nyef: flip214: Got a bit of CLOStrophobia going on? 2015-02-16T16:20:38Z flip214: nyef: trying to squeeze performance. 2015-02-16T16:20:48Z flip214: will be a PWM controller... 2015-02-16T16:20:55Z nyef: Fair enough. 2015-02-16T16:21:03Z flip214: still get "37,328 forms interpreted" and nearly double the CPU cycles. hmmm. 2015-02-16T16:21:13Z stassats: for what? 2015-02-16T16:21:51Z flip214: when changing the 3 DEFCLASS to 3 DEFSTRUCTs, the (TIME (... 1000 runs)) nearly doubles. 2015-02-16T16:22:37Z stassats: no, that's because of something else 2015-02-16T16:22:47Z flip214: what does the "N forms interpreted" mean? Why is that printed now? 2015-02-16T16:22:56Z stassats: show us the code 2015-02-16T16:22:59Z flip214: nothing else is different... I'm switching git versions around. 2015-02-16T16:23:33Z flip214: hmmm, not yet public... and I'm not sure whether I want that now, it's still work in progress. 2015-02-16T16:23:42Z stassats: then you are on your own 2015-02-16T16:24:11Z flip214: yeah, gimme a bit of time. 2015-02-16T16:26:02Z flip214: stassats: how about a private email with a tar.gz? 2015-02-16T16:26:17Z stassats: let's settle on a tar.bz2 2015-02-16T16:27:43Z flip214: fair enough 2015-02-16T16:28:12Z flip214: @gmail.com? 2015-02-16T16:28:19Z stassats: yep 2015-02-16T16:29:54Z flip214: thanks 2015-02-16T16:31:40Z flip214: you got it 2015-02-16T16:34:00Z flip214: the attached patch is the old one, with () in the defstruct still. sorry. 2015-02-16T16:35:01Z psy_ joined #sbcl 2015-02-16T16:35:18Z stassats: (make-instance 'time-range-forward) 2015-02-16T16:35:26Z stassats: it should be make-time-range-forward 2015-02-16T16:55:18Z flip214: okay ... and that makes the form interpreted? 2015-02-16T16:56:34Z flip214: thanks, down to <60M cycles... 2015-02-16T16:56:49Z stassats: it makes your code non portable 2015-02-16T16:57:04Z stassats: being slow is not the most worrying consequence 2015-02-16T16:57:33Z flip214: I don't really care about portable... 2015-02-16T16:57:47Z stassats: not standard 2015-02-16T16:57:49Z flip214: I need to run that on my Raspberry, with at least 100Hz. 2015-02-16T16:58:00Z nyef: Heh. I'm reminded of a recent discussion with pjb about portability libraries and dependencies. 2015-02-16T16:58:17Z stassats: non standard means it may stop working in the next version of sbcl 2015-02-16T16:58:38Z flip214: hmmm, what's not standarf here? 2015-02-16T16:58:41Z flip214: *standard 2015-02-16T16:58:45Z stassats: make-instance 2015-02-16T16:59:27Z flip214: well, I don't have that anymore.... changing to make-time-range-forward makes it work fast again, so I changed the few calls. 2015-02-16T16:59:55Z flip214: thank you very much! 2015-02-16T17:00:20Z flip214: any other feedback for me? style, algorithm, ...? 2015-02-16T17:00:35Z stassats: don't have the time 2015-02-16T17:00:40Z flip214: thanks all the same! 2015-02-16T17:00:51Z flip214: just asked because something might have caught your eye. 2015-02-16T17:01:52Z psy_ quit (Ping timeout: 252 seconds) 2015-02-16T17:02:11Z stassats: no, it went pretty quickly, i grepped for time-range-forward, saw make-instance and that's it 2015-02-16T17:05:39Z psy_ joined #sbcl 2015-02-16T17:25:10Z dmitigr quit (*.net *.split) 2015-02-16T17:25:11Z Quadrescence quit (*.net *.split) 2015-02-16T17:25:12Z milosn quit (*.net *.split) 2015-02-16T17:25:13Z luis quit (*.net *.split) 2015-02-16T17:25:13Z ivan4th quit (*.net *.split) 2015-02-16T17:25:16Z antoszka quit (*.net *.split) 2015-02-16T17:25:55Z psy_ quit (Ping timeout: 252 seconds) 2015-02-16T17:29:12Z sdemarre joined #sbcl 2015-02-16T17:31:32Z dmitigr joined #sbcl 2015-02-16T17:31:32Z Quadrescence joined #sbcl 2015-02-16T17:31:32Z milosn joined #sbcl 2015-02-16T17:31:32Z luis joined #sbcl 2015-02-16T17:31:32Z ivan4th joined #sbcl 2015-02-16T17:31:32Z antoszka joined #sbcl 2015-02-16T17:34:23Z psy_ joined #sbcl 2015-02-16T17:42:49Z psy_ quit (Ping timeout: 250 seconds) 2015-02-16T17:49:46Z dmitigr quit (Ping timeout: 255 seconds) 2015-02-16T18:00:14Z irsol quit (Ping timeout: 245 seconds) 2015-02-16T18:04:36Z irsol joined #sbcl 2015-02-16T18:05:53Z sdemarre quit (Ping timeout: 240 seconds) 2015-02-16T18:07:11Z psy_ joined #sbcl 2015-02-16T18:08:23Z jrm: nyef: This looks more like the original problem: https://picasaweb.google.com/lh/photo/Ety2e8Dt5sjCxf2ZLtRGbxEnogN27f9PzFp5hb57oGA?feat=directlink 2015-02-16T18:08:24Z davazp joined #sbcl 2015-02-16T18:10:19Z nyef: jrm: Mmm. But that's post-reinit, the original cause is long-gone and now something got corrupted and blows up when stumpwm tries to reconnect to the server. 2015-02-16T18:12:05Z jrm: Ah, so back to our original predicament. 2015-02-16T18:13:18Z nyef: stassats: 90decaf4b97ba547394a6a4dc49bf6f5e6025189 breaks building with 61-bit fixnums, possibly also with 62-bit fixnums. 2015-02-16T18:14:41Z jrm: I will submit the bug with as much info as I can provide and stop my noise. 2015-02-16T18:15:14Z stassats: nyef: how? 2015-02-16T18:16:07Z nyef: stassats: It blows up trying to cross-compile code/numbers.lisp. 2015-02-16T18:16:48Z nyef: As far as why, I figured I'd start by bisecting, then worry about trying to diagnose what's going on. 2015-02-16T18:18:18Z stassats: in what fashion does it blow up? 2015-02-16T18:18:44Z stassats: one thing i can see is that bit-index is not a fixnum 2015-02-16T18:19:04Z stassats: but that doesn't seem likely 2015-02-16T18:19:53Z nyef: Logcount, recursion in known function definition. 2015-02-16T18:20:07Z nyef: arg types = ((unsigned-byte 60)) 2015-02-16T18:21:10Z psy__ joined #sbcl 2015-02-16T18:21:31Z psy__ quit (Max SendQ exceeded) 2015-02-16T18:21:51Z psy__ joined #sbcl 2015-02-16T18:23:12Z stassats: i don't see it 2015-02-16T18:25:30Z nyef: I don't understand it yet myself, but it should be easy to replicate: Edit compiler/generic/early-vm.lisp and change n-fixnum-tag-bits from 1 to 3. 2015-02-16T18:26:03Z nyef: If it doesn't blow up for you then we're looking at another host environment contamination. 2015-02-16T18:26:37Z stassats: but that diff doesn't even touch fixnums 2015-02-16T18:27:11Z davazp quit (Ping timeout: 246 seconds) 2015-02-16T18:27:23Z nyef: Feel free to repeat my bisect, then. 2015-02-16T18:28:00Z nyef: Your build environment is faster than mine is, anyway. 2015-02-16T18:28:11Z nyef: I could easily have made a mistake with the bisection process. 2015-02-16T18:29:56Z stassats: i will eventually, still dealing with day job stuff 2015-02-16T18:30:03Z nyef: That's fair. 2015-02-16T18:37:36Z nyef: It changes the definition of type BIT-INDEX. 2015-02-16T18:38:21Z stassats: ok, and logcount is (defknown (logcount integer-length) (integer) bit-index (movable foldable flushable explicit-check)) 2015-02-16T18:38:45Z nyef: Right. 2015-02-16T18:38:56Z stassats: but it's unsigned-byte 56 2015-02-16T18:39:30Z stassats: no, that's bignum-length 2015-02-16T18:39:50Z stassats: right, there are more bits 2015-02-16T18:40:21Z stassats: so it just needs to truncate to fixnum 2015-02-16T18:40:39Z stassats: but that's not right 2015-02-16T18:41:12Z nyef: (integer 0 #x3FFFFFFFFFFFFFC0) ? 2015-02-16T18:41:15Z stassats: just add truly-the fixnum around logcount for fixnums? 2015-02-16T18:41:42Z stassats: actually, the problem is that i forgot to include my derive-type for logcount 2015-02-16T18:41:58Z stassats: despite the fact that it's stated in the log 2015-02-16T18:43:16Z stassats: nyef: that's (* (1- (expt 2 56)) 64) 2015-02-16T18:43:27Z stassats: 56 being 64-widetag-bits 2015-02-16T18:43:42Z stassats: i need to find my logcount deriver 2015-02-16T18:44:01Z stassats: it'll derive the fixnum case of logcount properly and it'll get transformed 2015-02-16T18:44:01Z nyef: Oh! Widetag bits. Right. 2015-02-16T18:48:22Z nyef: So... This will probably build with two fixnum tag bits? 2015-02-16T18:48:48Z nyef: I'll go with the band-aid approach of another TRULY-THE. Thanks. 2015-02-16T18:49:20Z stassats: i gues i forgot to add it into git or something 2015-02-16T18:50:16Z stassats: would have to write it anew, but it's trivial, i just used integer-length 2015-02-16T18:50:40Z nyef: (integer 0 #.sb!xc:n-positive-fixnum-bits) ? 2015-02-16T18:51:15Z stassats: what's that? 2015-02-16T18:52:16Z nyef: An integer between 0 and the number of bits in a positive-fixnum. 2015-02-16T18:52:35Z stassats: that's what you want for the truly-the? 2015-02-16T18:52:37Z nyef: By definition, the range of LOGCOUNT for a FIXNUM. 2015-02-16T18:53:17Z stassats: i don't that tight of a type will help the logcount stub 2015-02-16T18:53:20Z stassats: don't think 2015-02-16T18:53:39Z stassats: as a band aid, truly-the fixnum should do 2015-02-16T18:55:10Z nyef: Did I say sb!xc? It's sb!vm. 2015-02-16T18:55:48Z nyef: It doesn't help the logcount stub, but it's the truth, and it should at least let the stub get selected. 2015-02-16T18:58:00Z stassats: jrm: so, it segfaults in poll? 2015-02-16T19:02:40Z stassats: jrm: (member :os-provides-poll *features*)? 2015-02-16T19:05:29Z nyef: Hrm. Now it blows up in !cold-init. 2015-02-16T19:11:15Z sdemarre joined #sbcl 2015-02-16T19:17:48Z prxq joined #sbcl 2015-02-16T19:18:54Z nyef: Okay, looks like there's another bug for me to find. 2015-02-16T19:19:30Z nyef: Oh, wait, forgot to do a stash pop. 2015-02-16T19:38:43Z nyef: Hrm. So, still fails. Looks like I have something to dig further into. 2015-02-16T20:36:00Z leuler joined #sbcl 2015-02-16T20:43:12Z prxq quit (Remote host closed the connection) 2015-02-16T20:44:31Z Bicyclidine joined #sbcl 2015-02-16T20:52:02Z sdemarre quit (Ping timeout: 250 seconds) 2015-02-16T21:13:05Z nyef: Okay, something else is definitely going on, beyond just the LOGCOUNT thing. 2015-02-16T21:13:23Z nyef: I can build at 63-bit fixnums, but 62 and 61 both blow up in cold init. 2015-02-16T21:13:42Z nyef: And this is WITH a TRULY-THE slapped on the fixnum case of logcount. 2015-02-16T21:13:43Z stassats: at what point? 2015-02-16T21:14:07Z nyef: Backtrace is useless, so I'm about to trigger another build and use sb-show this time. 2015-02-16T21:14:31Z stassats: 32-bit host still fails with The value -2305843009204879977 is not of type FIXNUM. 2015-02-16T21:14:45Z stassats: that one is in douglas' teritorry 2015-02-16T21:14:59Z nyef: Yeah, I'm sitting at the bisected commit, not HEAD. 2015-02-16T21:15:17Z stassats: i added a logcount deriver 2015-02-16T21:15:49Z angavrilov quit (Remote host closed the connection) 2015-02-16T21:25:55Z nyef: http://paste.lisp.org/display/145783 2015-02-16T21:26:08Z nyef: Backtrace is useless... 2015-02-16T21:26:13Z nyef: No interrupt contexts. 2015-02-16T21:30:07Z Adlai quit (Ping timeout: 252 seconds) 2015-02-16T21:31:58Z selat quit (Quit: Lost terminal) 2015-02-16T21:33:15Z stassats: confusingly enough, SB-C::CAST is SB-ALIEN:CAST 2015-02-16T21:37:39Z nyef: Doesn't SB-C :USE SB-ALIEN? 2015-02-16T21:37:46Z stassats: it does 2015-02-16T21:43:36Z scymtym_ joined #sbcl 2015-02-16T21:46:13Z prxq joined #sbcl 2015-02-16T21:54:54Z Adlai joined #sbcl 2015-02-16T22:08:15Z stassats: not sure about the best way for 1234919 2015-02-16T22:08:57Z stassats: i can just patch up the symptom, but then a lot of things are careless about lvar-uses 2015-02-16T22:08:59Z stassats: which may be a cast 2015-02-16T22:17:44Z stassats: weirdly, (funcall (lambda ()) 10) => a warning, (funcall (the function (lambda ())) 10) => no warning 2015-02-16T22:20:23Z gingerale quit (Ping timeout: 240 seconds) 2015-02-16T22:24:47Z prxq quit (Read error: Connection reset by peer) 2015-02-16T22:26:06Z leuler quit (Quit: ERC Version 5.1.2 $Revision: 1.796.2.6 $ (IRC client for Emacs)) 2015-02-25T06:57:16Z ccl-logbot joined #sbcl 2015-02-25T06:57:16Z 2015-02-25T06:57:16Z names: ccl-logbot Shinmera krzysz00 echo-area pranavrc milosn christoph_debian oleo irsol karswell edgar-rft redline6561 loke p_l Bike fikusz yauz_2 pkhuong psy_ mood White_Flame flip214 snafuchs ASau Adlai` |3b| Posterdati jdz loke_ @Xof kanru specbot minion chris2 pchrist foom PuercoPop ferada Intensity fe[nl]ix Blkt nicdev joshe cojy_ psilord ams reb` les antoszka ivan4th luis nightshade427 drmeister soggybread dustinm` jsnell salv0 jrm alchemis7 carvite 2015-02-25T06:57:16Z names: brucem 2015-02-25T06:59:17Z psy_ quit (Ping timeout: 246 seconds) 2015-02-25T07:11:05Z gingerale joined #sbcl 2015-02-25T07:24:49Z edgar-rft quit (Remote host closed the connection) 2015-02-25T07:25:24Z edgar-rft joined #sbcl 2015-02-25T07:26:51Z Shinmera quit (Ping timeout: 256 seconds) 2015-02-25T08:01:44Z Cymew joined #sbcl 2015-02-25T08:02:14Z Cymew quit (Client Quit) 2015-02-25T08:02:35Z Cymew joined #sbcl 2015-02-25T08:12:59Z selat joined #sbcl 2015-02-25T08:54:57Z psy_ joined #sbcl 2015-02-25T08:55:37Z angavrilov joined #sbcl 2015-02-25T09:04:56Z gingerale quit (Remote host closed the connection) 2015-02-25T09:05:30Z gingerale joined #sbcl 2015-02-25T09:09:33Z Shinmera joined #sbcl 2015-02-25T09:19:15Z Shinmera quit (Ping timeout: 265 seconds) 2015-02-25T09:27:40Z attila_lendvai joined #sbcl 2015-02-25T09:27:40Z attila_lendvai quit (Changing host) 2015-02-25T09:27:40Z attila_lendvai joined #sbcl 2015-02-25T09:32:37Z selat quit (Quit: Lost terminal) 2015-02-25T09:41:30Z Shinmera joined #sbcl 2015-02-25T09:49:52Z scymtym joined #sbcl 2015-02-25T09:50:28Z mega1 joined #sbcl 2015-02-25T10:20:20Z selat joined #sbcl 2015-02-25T10:48:12Z edgar-rft quit (Remote host closed the connection) 2015-02-25T10:48:36Z edgar-rft joined #sbcl 2015-02-25T10:54:59Z Shinmera quit (Quit: しつれいしなければならないんです。) 2015-02-25T10:57:22Z loke_ quit (Ping timeout: 265 seconds) 2015-02-25T11:14:06Z loke_ joined #sbcl 2015-02-25T11:20:42Z attila_lendvai quit (Ping timeout: 245 seconds) 2015-02-25T11:28:14Z Adlai` is now known as adlai 2015-02-25T11:46:35Z scymtym_ joined #sbcl 2015-02-25T11:48:36Z scymtym quit (Ping timeout: 272 seconds) 2015-02-25T11:51:13Z Shinmera joined #sbcl 2015-02-25T12:00:30Z eudoxia_ joined #sbcl 2015-02-25T12:01:21Z eudoxia_ quit (Client Quit) 2015-02-25T12:02:47Z eudoxia joined #sbcl 2015-02-25T12:05:32Z selat quit (Read error: Connection reset by peer) 2015-02-25T12:05:50Z selat joined #sbcl 2015-02-25T12:06:10Z selat quit (Read error: Connection reset by peer) 2015-02-25T12:11:25Z selat joined #sbcl 2015-02-25T12:11:26Z selat quit (Read error: Connection reset by peer) 2015-02-25T12:16:42Z selat joined #sbcl 2015-02-25T12:17:30Z selat quit (Read error: Connection reset by peer) 2015-02-25T12:32:46Z stassats joined #sbcl 2015-02-25T12:36:12Z stassats: loke: there seems to be two failure modes for classoid-typep, one where you do redefine classes and another where there not yet fully defined 2015-02-25T12:36:55Z stassats: so, the latter probably could be smarter how it invalidates classoids 2015-02-25T12:37:49Z stassats: and then classoid-typep could either ignore the second mismatch or spin 2015-02-25T12:39:00Z stassats: but spinning would not necessary give you anything, since if you protect something with a typep, and it's redefined just between (typep x 'class) and the code that operates on X, it'd be wrong or (wouldn't matter) 2015-02-25T13:07:27Z psy_ quit (Remote host closed the connection) 2015-02-25T13:13:37Z pranavrc quit 2015-02-25T13:17:03Z stassats quit (Ping timeout: 265 seconds) 2015-02-25T13:31:38Z Shinmera quit (Remote host closed the connection) 2015-02-25T13:32:02Z Shinmera joined #sbcl 2015-02-25T14:02:58Z ivan4th quit (Ping timeout: 255 seconds) 2015-02-25T14:05:36Z ivan4th joined #sbcl 2015-02-25T14:38:50Z oleo is now known as Guest26242 2015-02-25T14:39:47Z oleo__ joined #sbcl 2015-02-25T14:41:49Z Guest26242 quit (Ping timeout: 250 seconds) 2015-02-25T15:08:48Z psy_ joined #sbcl 2015-02-25T15:11:33Z prxq joined #sbcl 2015-02-25T16:04:03Z Shinmera quit (Quit: しつれいしなければならないんです。) 2015-02-25T16:05:28Z stassats joined #sbcl 2015-02-25T16:06:38Z edgar-rft quit (Remote host closed the connection) 2015-02-25T16:07:57Z attila_lendvai joined #sbcl 2015-02-25T16:07:57Z attila_lendvai quit (Changing host) 2015-02-25T16:07:57Z attila_lendvai joined #sbcl 2015-02-25T16:22:11Z loke_ quit (Ping timeout: 250 seconds) 2015-02-25T16:31:35Z eudoxia quit (Quit: Leaving) 2015-02-25T16:33:16Z loke_ joined #sbcl 2015-02-25T16:35:31Z attila_lendvai quit (Ping timeout: 255 seconds) 2015-02-25T16:36:56Z Cymew quit (Ping timeout: 256 seconds) 2015-02-25T16:41:51Z gabriel_laddel joined #sbcl 2015-02-25T16:42:36Z Shinmera joined #sbcl 2015-02-25T16:55:41Z jrm quit (Ping timeout: 252 seconds) 2015-02-25T16:55:44Z gabriel_laddel quit (Remote host closed the connection) 2015-02-25T16:59:58Z nyef joined #sbcl 2015-02-25T17:32:47Z selat joined #sbcl 2015-02-25T17:46:04Z macmathis joined #sbcl 2015-02-25T17:59:37Z macmathis quit (Quit: leaving) 2015-02-25T18:12:36Z Bicyclidine joined #sbcl 2015-02-25T18:23:30Z jrm joined #sbcl 2015-02-25T18:39:14Z krzysz00 quit (Ping timeout: 246 seconds) 2015-02-25T18:40:41Z csziacobus joined #sbcl 2015-02-25T18:44:30Z eudoxia joined #sbcl 2015-02-25T18:50:27Z Bicyclidine quit (Ping timeout: 252 seconds) 2015-02-25T18:51:53Z Bicyclidine joined #sbcl 2015-02-25T18:57:16Z Bicyclidine quit (Ping timeout: 255 seconds) 2015-02-25T19:02:38Z Bicyclidine joined #sbcl 2015-02-25T19:06:19Z krzysz00 joined #sbcl 2015-02-25T19:24:37Z Quadrescence joined #sbcl 2015-02-25T19:32:18Z Quadrescence quit (Quit: This computer has gone to sleep) 2015-02-25T19:35:20Z krzysz00 quit (Ping timeout: 252 seconds) 2015-02-25T20:15:09Z prxq quit (Ping timeout: 252 seconds) 2015-02-25T20:23:59Z stassats` joined #sbcl 2015-02-25T20:47:54Z nikki93_ joined #sbcl 2015-02-25T20:52:29Z edgar-rft joined #sbcl 2015-02-25T21:03:02Z nikki93_ quit (Remote host closed the connection) 2015-02-25T21:29:46Z selat quit (Read error: Connection reset by peer) 2015-02-25T21:35:40Z selat joined #sbcl 2015-02-25T21:38:36Z selat quit (Read error: Connection reset by peer) 2015-02-25T21:42:28Z ehaliewicz joined #sbcl 2015-02-25T21:44:38Z selat joined #sbcl 2015-02-25T21:47:10Z selat quit (Read error: Connection reset by peer) 2015-02-25T21:52:28Z selat joined #sbcl 2015-02-25T21:52:48Z selat quit (Read error: Connection reset by peer) 2015-02-25T21:58:03Z selat joined #sbcl 2015-02-25T22:00:07Z rpg joined #sbcl 2015-02-25T22:03:36Z prxq joined #sbcl 2015-02-25T22:18:24Z csziacobus quit (Quit: csziacobus) 2015-02-25T22:18:37Z csziacobus joined #sbcl 2015-02-25T22:22:22Z gingerale quit (Ping timeout: 245 seconds) 2015-02-25T22:27:42Z angavrilov quit (Remote host closed the connection) 2015-02-25T22:54:01Z eudoxia quit (Quit: Leaving) 2015-02-25T23:16:29Z scymtym_: on x86, the new EQL/INTEGER-FOLDING test fails: https://ci.cor-lab.org/job/sbcl-master/featureset=1,label=ubuntu_trusty_32bit/870/consoleFull (search for "UNEXPECTED-FAILURE :EQL/INTEGER-FOLDING") 2015-02-25T23:26:01Z oleo__ quit (Quit: Verlassend) 2015-02-25T23:27:05Z oleo joined #sbcl 2015-02-25T23:45:57Z alchemis7 quit (Quit: @) 2015-02-25T23:56:10Z rpg_ joined #sbcl 2015-02-25T23:56:22Z rpg_ quit (Client Quit) 2015-02-25T23:57:11Z White_Flame quit (Remote host closed the connection) 2015-02-25T23:58:41Z attila_lendvai joined #sbcl 2015-02-26T00:00:35Z mega1 quit (Ping timeout: 246 seconds) 2015-02-26T00:00:38Z rpg quit (Ping timeout: 256 seconds) 2015-02-26T00:05:10Z Bicyclidine quit (Ping timeout: 272 seconds) 2015-02-26T00:06:50Z Bicyclidine joined #sbcl 2015-02-26T00:11:03Z Bicyclidine quit (Ping timeout: 250 seconds) 2015-02-26T00:12:07Z Bicyclidine joined #sbcl 2015-02-26T00:17:33Z ehaliewicz quit (Remote host closed the connection) 2015-02-26T00:33:45Z rpg joined #sbcl 2015-02-26T00:37:22Z stassats`: x86 boxes single floats 2015-02-26T00:37:29Z stassats`: i don't know what that test tests 2015-02-26T00:37:58Z stassats`: scymtym_: can you report it on the sbcl-devel? 2015-02-26T00:38:12Z scymtym_: sure 2015-02-26T00:38:22Z nyef: All 32-bit backends box single-floats. 2015-02-26T00:38:49Z stassats`: all mortals are socrates 2015-02-26T00:39:27Z nyef: ... Therefore, all x86 boxes are Socrates? 2015-02-26T00:40:05Z nyef: ("And what does it mean to "single" a "float", anyway?) 2015-02-26T00:41:37Z stassats`: choosing from many floats? 2015-02-26T00:42:00Z stassats`: or buffaloes, not sure here 2015-02-26T00:42:35Z nyef: Heh. Wasn't "buffalo buffalo buffalo buffalo buffalo" a legitimate English sentence, somehow? 2015-02-26T00:42:36Z prxq quit (Ping timeout: 252 seconds) 2015-02-26T00:42:52Z nyef: The third "buffalo" being the verb. 2015-02-26T00:43:05Z Bicyclidine: Buffalo (city) buffalo (noun) buffalo (verb) Buffalo (city) buffalo (noun) 2015-02-26T00:43:30Z nyef: Something like that, yes. 2015-02-26T00:57:21Z Shinmera quit (Quit: しつれいしなければならないんです。) 2015-02-26T00:59:50Z Bicyclidine quit (Quit: get down) 2015-02-26T01:07:01Z stassats` quit (Ping timeout: 264 seconds) 2015-02-26T01:07:05Z stassats quit (Ping timeout: 256 seconds) 2015-02-26T01:20:41Z rpg quit (Quit: rpg) 2015-02-26T01:31:26Z alchemis7 joined #sbcl 2015-02-26T01:35:27Z rpg joined #sbcl 2015-02-26T01:35:28Z rpg quit (Remote host closed the connection) 2015-02-26T01:43:19Z salv0 quit (Ping timeout: 265 seconds) 2015-02-26T01:43:59Z salv0 joined #sbcl 2015-02-26T01:46:44Z nikki93_ joined #sbcl 2015-02-26T01:53:37Z salv0 quit (Ping timeout: 245 seconds) 2015-02-26T01:53:42Z eschatologist joined #sbcl 2015-02-26T01:53:59Z salv0 joined #sbcl 2015-02-26T02:13:59Z attila_lendvai quit (Ping timeout: 245 seconds) 2015-02-26T02:24:08Z loke: stassatis: I can reliably reproduce the problem every single time (it's basically two threads doing a drakma http call at the same time) 2015-02-26T02:24:16Z loke: stassats even 2015-02-26T02:56:27Z oleo is now known as Guest77090 2015-02-26T02:57:24Z oleo__ joined #sbcl 2015-02-26T02:59:26Z Guest77090 quit (Ping timeout: 246 seconds) 2015-02-26T03:24:01Z edgar-rft quit (Quit: edgar-rft) 2015-02-26T03:39:14Z christoph_debian quit (Ping timeout: 272 seconds) 2015-02-26T03:45:21Z eschatologist quit (Quit: Textual IRC Client: www.textualapp.com) 2015-02-26T03:47:01Z rszeno joined #sbcl 2015-02-26T03:52:10Z christoph_debian joined #sbcl 2015-02-26T04:10:38Z rszeno quit (Quit: Leaving.) 2015-02-26T04:12:48Z psy_ quit (Ping timeout: 256 seconds) 2015-02-26T04:17:06Z selat quit (Quit: Lost terminal) 2015-02-26T04:56:34Z nikki93_ quit (Remote host closed the connection) 2015-02-26T05:06:14Z psy_ joined #sbcl 2015-02-26T05:17:45Z nyef quit (Quit: G'night all) 2015-02-26T05:35:22Z nikki93_ joined #sbcl 2015-02-26T05:35:52Z scymtym_ quit (Ping timeout: 240 seconds) 2015-02-26T05:53:46Z pranavrc joined #sbcl 2015-02-26T05:59:47Z nikki93_ quit (Remote host closed the connection) 2015-02-26T06:01:50Z nikki93_ joined #sbcl 2015-02-26T06:01:58Z nikki93_ quit (Remote host closed the connection) 2015-02-26T06:42:19Z Quadrescence joined #sbcl 2015-02-26T06:44:53Z Shinmera joined #sbcl 2015-02-26T06:45:56Z csziacobus quit (Quit: csziacobus) 2015-02-26T06:49:24Z selat joined #sbcl 2015-02-26T06:55:05Z Quadrescence quit (Quit: This computer has gone to sleep) 2015-02-26T06:58:53Z gingerale joined #sbcl 2015-02-26T07:42:16Z psy_ quit (Quit: Leaving) 2015-02-26T08:02:29Z angavrilov joined #sbcl 2015-02-26T08:13:22Z mega1 joined #sbcl 2015-02-26T08:31:44Z mega1 quit (Quit: ERC Version 5.3 (IRC client for Emacs)) 2015-02-26T08:32:02Z mega1 joined #sbcl 2015-02-26T08:33:37Z Cymew joined #sbcl 2015-02-26T08:33:49Z Cymew quit (Read error: Connection reset by peer) 2015-02-26T08:34:04Z Cymew joined #sbcl 2015-02-26T08:36:06Z rszeno joined #sbcl 2015-02-26T08:51:44Z psy_ joined #sbcl 2015-02-26T08:52:04Z psy_ quit (Max SendQ exceeded) 2015-02-26T08:52:57Z psy_ joined #sbcl 2015-02-26T09:34:44Z selat quit (Quit: Lost terminal) 2015-02-26T10:02:55Z shrdlu68 joined #sbcl 2015-02-26T10:06:04Z shrdlu68: I am looking for the documentation for sb-unix:* . The only one I have found so far is sb-unix:unix-exit on http://www.sbcl.org/manual/index.html#Support-For-Unix. (documentation #'sb-unix:unix-open t) returns nil. 2015-02-26T10:07:53Z shrdlu68 quit (Quit: Leaving) 2015-02-26T10:09:12Z shrdlu68 joined #sbcl 2015-02-26T10:11:38Z flip214: shrdlu68: list all exported symbols, and do DESCRIBE on them. Or, if your editor allows, do auto-completion and ask for the documentation... 2015-02-26T10:18:44Z selat joined #sbcl 2015-02-26T10:21:15Z shrdlu68: flip214:thanks. 2015-02-26T10:24:44Z Xof: wait, that's a terrible answer 2015-02-26T10:24:53Z Xof: the sb-unix package is internal functionality, which can break at any time 2015-02-26T10:25:29Z Xof: you can discover that by looking at the documentation for the sb-unix package: (documentation (find-package "SB-UNIX") t) 2015-02-26T10:25:53Z Xof: what do you actually want to do? 2015-02-26T10:28:04Z shrdlu68: Someone suggested I use sb-unix:unix-open and read for reading device files. I want to investigate the possible pros and cons of that. 2015-02-26T10:30:32Z Xof: You probably want sb-posix instead. http://www.sbcl.org/manual/#sb_002dposix 2015-02-26T10:38:26Z shrdlu68: sb-posix will do. 2015-02-26T10:38:31Z shrdlu68 quit (Quit: Leaving) 2015-02-26T10:39:35Z echo-area quit (Read error: Connection reset by peer) 2015-02-26T10:40:17Z echo-area joined #sbcl 2015-02-26T10:43:53Z Shinmera quit (Quit: しつれいしなければならないんです。) 2015-02-26T11:11:01Z Shinmera joined #sbcl 2015-02-26T12:06:26Z echo-area quit (Remote host closed the connection) 2015-02-26T12:10:14Z echo-area joined #sbcl 2015-02-26T12:39:47Z HardikChoudhary joined #sbcl 2015-02-26T12:40:51Z eudoxia joined #sbcl 2015-02-26T12:51:08Z oleo__ quit (Quit: Verlassend) 2015-02-26T12:52:15Z oleo joined #sbcl 2015-02-26T12:55:18Z selat quit (Read error: Connection reset by peer) 2015-02-26T12:58:32Z scymtym joined #sbcl 2015-02-26T13:01:07Z selat joined #sbcl 2015-02-26T13:01:45Z selat quit (Read error: Connection reset by peer) 2015-02-26T13:06:22Z stassats joined #sbcl 2015-02-26T13:07:25Z selat joined #sbcl 2015-02-26T13:07:33Z selat quit (Read error: Connection reset by peer) 2015-02-26T13:12:57Z stassats` joined #sbcl 2015-02-26T13:13:37Z oleo quit (Quit: Verlassend) 2015-02-26T13:17:15Z psy_ quit (Remote host closed the connection) 2015-02-26T13:18:08Z selat joined #sbcl 2015-02-26T13:19:04Z selat quit (Read error: Connection reset by peer) 2015-02-26T13:21:07Z edgar-rft joined #sbcl 2015-02-26T13:24:18Z selat joined #sbcl 2015-02-26T13:25:41Z selat quit (Read error: Connection reset by peer) 2015-02-26T13:27:52Z pranavrc quit 2015-02-26T13:30:57Z selat joined #sbcl 2015-02-26T13:32:55Z selat quit (Read error: Connection reset by peer) 2015-02-26T13:35:48Z HardikChoudhary quit (Quit: Leaving.) 2015-02-26T13:38:49Z selat joined #sbcl 2015-02-26T13:39:15Z selat quit (Read error: Connection reset by peer) 2015-02-26T13:49:54Z selat joined #sbcl 2015-02-26T13:50:01Z selat quit (Read error: Connection reset by peer) 2015-02-26T13:55:18Z selat joined #sbcl 2015-02-26T13:56:13Z selat quit (Read error: Connection reset by peer) 2015-02-26T13:56:20Z davazp joined #sbcl 2015-02-26T13:58:10Z attila_lendvai joined #sbcl 2015-02-26T14:01:28Z selat joined #sbcl 2015-02-26T14:02:49Z selat quit (Read error: Connection reset by peer) 2015-02-26T14:08:04Z selat joined #sbcl 2015-02-26T14:08:04Z selat quit (Read error: Connection reset by peer) 2015-02-26T14:09:17Z oleo joined #sbcl 2015-02-26T14:18:19Z selat joined #sbcl 2015-02-26T14:18:35Z selat quit (Read error: Connection reset by peer) 2015-02-26T14:18:52Z stassats quit (Ping timeout: 240 seconds) 2015-02-26T14:19:24Z stassats` quit (Ping timeout: 245 seconds) 2015-02-26T14:23:52Z selat joined #sbcl 2015-02-26T14:24:05Z selat quit (Read error: Connection reset by peer) 2015-02-26T14:29:20Z selat joined #sbcl 2015-02-26T14:29:48Z selat quit (Read error: Connection reset by peer) 2015-02-26T14:35:03Z selat joined #sbcl 2015-02-26T14:35:20Z selat quit (Read error: Connection reset by peer) 2015-02-26T14:43:38Z pkhuong quit (Ping timeout: 246 seconds) 2015-02-26T14:45:54Z selat joined #sbcl 2015-02-26T14:47:59Z selat quit (Read error: Connection reset by peer) 2015-02-26T14:55:12Z rszeno quit (Quit: Leaving.) 2015-02-26T14:58:55Z oleo is now known as Guest59545 2015-02-26T14:59:41Z ASau quit (Ping timeout: 246 seconds) 2015-02-26T14:59:51Z oleo__ joined #sbcl 2015-02-26T15:02:16Z Guest59545 quit (Ping timeout: 265 seconds) 2015-02-26T15:03:35Z ASau joined #sbcl 2015-02-26T15:15:03Z selat joined #sbcl 2015-02-26T15:15:06Z selat quit (Read error: Connection reset by peer) 2015-02-26T15:29:45Z oleo__ quit (Quit: Verlassend) 2015-02-26T15:30:55Z selat joined #sbcl 2015-02-26T15:31:06Z selat quit (Read error: Connection reset by peer) 2015-02-26T15:36:21Z selat joined #sbcl 2015-02-26T15:36:23Z selat quit (Read error: Connection reset by peer) 2015-02-26T15:43:55Z ASau` joined #sbcl 2015-02-26T15:46:58Z salv0 quit (Ping timeout: 256 seconds) 2015-02-26T15:47:27Z ASau quit (Ping timeout: 244 seconds) 2015-02-26T15:49:48Z ASau` is now known as ASau 2015-02-26T15:56:39Z oleo joined #sbcl 2015-02-26T15:57:01Z flip214: Xof: another idea for gsoc: threading on ARM. 2015-02-26T15:57:53Z davazp quit (Ping timeout: 256 seconds) 2015-02-26T16:01:37Z nikki93_ joined #sbcl 2015-02-26T16:07:50Z Cymew quit (Ping timeout: 272 seconds) 2015-02-26T16:10:15Z stassats joined #sbcl 2015-02-26T16:10:39Z stassats` joined #sbcl 2015-02-26T16:12:33Z nikki93_ quit (Remote host closed the connection) 2015-02-26T16:48:56Z rpg joined #sbcl 2015-02-26T17:03:01Z eudoxia quit (Read error: Connection reset by peer) 2015-02-26T17:05:20Z luis quit (Quit: ZNC - http://znc.sourceforge.net) 2015-02-26T17:06:16Z luis joined #sbcl 2015-02-26T17:13:23Z attila_lendvai quit (Quit: Leaving.) 2015-02-26T17:13:32Z Bicyclidine joined #sbcl 2015-02-26T17:15:33Z Quadrescence joined #sbcl 2015-02-26T17:16:29Z mega1 quit (Ping timeout: 245 seconds) 2015-02-26T17:24:59Z Quadrescence quit (Quit: This computer has gone to sleep) 2015-02-26T17:29:29Z attila_lendvai joined #sbcl 2015-02-26T17:29:29Z attila_lendvai quit (Changing host) 2015-02-26T17:29:29Z attila_lendvai joined #sbcl 2015-02-26T17:30:33Z psy_ joined #sbcl 2015-02-26T17:33:33Z gabriel_laddel joined #sbcl 2015-02-26T17:45:14Z mega1 joined #sbcl 2015-02-26T17:45:33Z mega1 quit (Client Quit) 2015-02-26T17:45:47Z mega1 joined #sbcl 2015-02-26T17:54:07Z nikki93_ joined #sbcl 2015-02-26T17:58:17Z attila_lendvai quit (Read error: No route to host) 2015-02-26T17:58:59Z attila_lendvai joined #sbcl 2015-02-26T17:58:59Z attila_lendvai quit (Changing host) 2015-02-26T17:58:59Z attila_lendvai joined #sbcl 2015-02-26T18:17:45Z attila_lendvai quit (Quit: Leaving.) 2015-02-26T18:18:38Z nikki93_ quit (Remote host closed the connection) 2015-02-26T18:20:02Z attila_lendvai joined #sbcl 2015-02-26T18:39:54Z nikki93_ joined #sbcl 2015-02-26T19:03:11Z rpg quit (Ping timeout: 252 seconds) 2015-02-26T19:05:07Z davazp joined #sbcl 2015-02-26T19:32:30Z nikki93_ quit (Remote host closed the connection) 2015-02-26T19:33:03Z nikki93_ joined #sbcl 2015-02-26T19:37:09Z joshe quit (Remote host closed the connection) 2015-02-26T19:37:19Z nikki93_ quit (Ping timeout: 250 seconds) 2015-02-26T19:37:31Z joshe joined #sbcl 2015-02-26T19:40:24Z davazp quit (Read error: Connection reset by peer) 2015-02-26T19:54:34Z Bicyclidine quit (Ping timeout: 272 seconds) 2015-02-26T19:56:03Z Bicyclidine joined #sbcl 2015-02-26T20:14:18Z nikki93_ joined #sbcl 2015-02-26T20:27:53Z nikki93_ quit (Remote host closed the connection) 2015-02-26T20:28:26Z nikki93_ joined #sbcl 2015-02-26T20:33:18Z HardikChoudhary joined #sbcl 2015-02-26T20:33:25Z nikki93_ quit (Ping timeout: 264 seconds) 2015-02-26T20:36:30Z nikki93_ joined #sbcl 2015-02-26T20:37:17Z nikki93_ quit (Remote host closed the connection) 2015-02-26T20:39:05Z nikki93_ joined #sbcl 2015-02-26T20:43:22Z HardikChoudhary quit (Quit: Leaving.) 2015-02-26T20:45:05Z nikki93_ quit (Remote host closed the connection) 2015-02-26T20:49:02Z Bicyclid1ne joined #sbcl 2015-02-26T20:50:16Z Bicyclidine quit (Ping timeout: 265 seconds) 2015-02-26T20:54:06Z Bicyclid1ne quit (Ping timeout: 256 seconds) 2015-02-26T20:55:15Z nikki93_ joined #sbcl 2015-02-26T20:55:17Z Bicyclidine joined #sbcl 2015-02-26T21:11:54Z Bicyclidine quit (Ping timeout: 245 seconds) 2015-02-26T21:14:56Z gabriel_laddel quit (Remote host closed the connection) 2015-02-26T21:17:03Z nikki93_ quit (Remote host closed the connection) 2015-02-26T21:18:46Z Bicyclidine joined #sbcl 2015-02-26T21:25:32Z rpg joined #sbcl 2015-02-26T21:30:08Z rpg: Can anyone offer any guidelines for debugging a crash to LDB on SBCL? When I look at the backtrace, I see what seem to be only entries pertaining to pretty printing and PRINT-UNREADABLE-OBJECT. 2015-02-26T21:33:17Z rpg: Looks like stack exhaustion, but it's happening when I compile a system. Works fine on Allegro. 2015-02-26T21:41:55Z rpg: Actually looks like it might be an infinite recursion trying to print a style warning from inside ASDF. 2015-02-26T21:45:00Z angavrilov quit (Remote host closed the connection) 2015-02-26T21:52:12Z rpg: One follow-up: this seems to be happening in ASDF code that catches warnings and re-signals them. This call causes an infinite loop: (FORMAT NIL "caught ~S:~%~~@< ~~@;~~A~~:>" STYLE-WARNING) 2015-02-26T21:52:42Z stassats: aka the most annoying asdf feature? 2015-02-26T21:57:56Z rpg: stassats: Not my favorite, either. 2015-02-26T21:58:22Z rpg: I *may* be wrong in blaming ASDF for this -- I have only the ldb backtrace to go on. 2015-02-26T22:00:49Z rpg: I *think* the proximate cause is the call to format in PRINT-COMPILER-CONDITION in ir1report.lisp 2015-02-26T22:04:42Z rpg: Somehow, in the compilation context, this call: (WARN 'SB-INT:SIMPLE-STYLE-WARNING :FORMAT-CONTROL "undefined ~(~A~): ~S" :FORMAT-ARGUMENTS '(:TYPE ONTOLOGY::THING)) causes an infinite loop and stack exhaustion. Maybe it *is* ASDF. 2015-02-26T22:10:44Z adlai quit (Ping timeout: 250 seconds) 2015-02-26T22:11:49Z adlai joined #sbcl 2015-02-26T22:17:37Z rpg: The stack has an infinite recurrence of 12920: (SB-PRETTY:OUTPUT-PRETTY-OBJECT SYMBOL #) calls to output-pretty-object and PRINT-OBJECT (t t) SYMBOL PRETTY-STREAM. 2015-02-26T22:18:24Z rpg: So it looks like we're trying to print to a string stream, then trying to print to a string stream to get the string to print, then trying to... 2015-02-26T22:21:01Z rpg: If I'm right and it's the symbol, it looks like if (pprint-dispatch object) returns T (PRETTY) as its second value, and itself as the first, we will get into an infinite recursion. 2015-02-26T22:22:27Z rpg: s/itself/ some function that recurses on symbols 2015-02-26T22:32:36Z rpg: remote cause is a forward reference of a type name in a TYPEP form. 2015-02-26T22:33:01Z rpg: But I still have no idea why issuing that warning would cause printer recursion and this epic fail. 2015-02-26T22:35:32Z gingerale quit (Ping timeout: 256 seconds) 2015-02-26T22:44:13Z rpg: haha. Never mind. Some....kind person... crapped up the pprint dispatch table. 2015-02-26T23:47:13Z rpg quit (Ping timeout: 264 seconds) 2015-02-26T23:50:53Z Shinmera quit (Quit: しつれいしなければならないんです。) 2015-02-27T00:05:41Z stassats quit (Ping timeout: 252 seconds) 2015-02-27T00:05:49Z stassats` quit (Ping timeout: 264 seconds) 2015-02-27T00:11:19Z attila_lendvai quit (Quit: Leaving.) 2015-02-27T00:13:51Z attila_lendvai joined #sbcl 2015-02-27T00:26:45Z minion quit (Remote host closed the connection) 2015-02-27T00:26:53Z minion joined #sbcl 2015-02-27T00:27:53Z |3b| quit (Excess Flood) 2015-02-27T00:28:48Z nicdev quit (Remote host closed the connection) 2015-02-27T00:29:09Z nicdev` joined #sbcl 2015-02-27T00:30:11Z |3b| joined #sbcl 2015-02-27T01:09:09Z Bicyclidine quit (Quit: monstrous) 2015-02-27T01:25:48Z pkhuong joined #sbcl 2015-02-27T01:26:11Z pkhuong is now known as Guest17295 2015-02-27T01:26:49Z luis quit (Ping timeout: 256 seconds) 2015-02-27T01:27:51Z luis joined #sbcl 2015-02-27T01:28:31Z loke_ quit (Ping timeout: 256 seconds) 2015-02-27T01:39:35Z loke_ joined #sbcl 2015-02-27T01:54:41Z loke: Guest17295: hello :-) 2015-02-27T01:58:53Z rpg joined #sbcl 2015-02-27T02:05:03Z redline6561 quit (Quit: ZNC - http://znc.in)