2015-02-21T00:17:32Z rszeno joined #sbcl
2015-02-21T00:24:52Z stassats: managed to do away with printing and just putting things into hashtables
2015-02-21T00:27:13Z nyef: Right, I still don't understand what's going on with DX allocation at this level.
2015-02-21T00:27:34Z nyef: And, at the same time, it's not actively broken.
2015-02-21T00:28:32Z stassats: i'm in "what the hell could be going on" state
2015-02-21T00:29:05Z stassats: concurrently putting things into separate hashtables causing deadlocks? what?
2015-02-21T00:29:41Z |3b|: calculating hash of things while those things are changing maybe?
2015-02-21T00:30:58Z |3b|: though i guess that would only be for equal/equalp hash tables
2015-02-21T00:32:10Z stassats: well, nothing can be changing while it calculates
2015-02-21T00:32:34Z stassats: and it's only on sb-safepoint
2015-02-21T00:40:52Z stassats: http://paste.lisp.org/display/145847#2
2015-02-21T00:42:52Z stassats: the deadlock is GC waiting for other threads, hashtables do not make any sense
2015-02-21T00:44:17Z stassats: it waits for one thread
2015-02-21T00:45:23Z stassats: is it the join-thread thread?
2015-02-21T00:46:50Z stassats: maybe gethash hits a safepoint in a funny way
2015-02-21T00:49:26Z stassats: maybe it's just because of locking
2015-02-21T00:52:30Z karswell quit (Read error: Connection reset by peer)
2015-02-21T00:53:21Z karswell joined #sbcl
2015-02-21T00:54:33Z stassats: doesn't seem to happen with my own locks
2015-02-21T00:56:07Z stassats: long time since i've seen bugs like this
2015-02-21T00:57:44Z karswell quit (Ping timeout: 246 seconds)
2015-02-21T01:02:04Z stassats: does anybody run safepoints?
2015-02-21T01:02:47Z nyef: Those poor windows users might.
2015-02-21T01:03:35Z stassats: no, i want somebody to confirm
2015-02-21T01:03:37Z fe[nl]ix: I do
2015-02-21T01:03:54Z stassats: fe[nl]ix: can you try the paste?
2015-02-21T01:05:00Z fe[nl]ix: more specifically, I use :sb-safepoint :sb-safepoint-strictly :sb-thruption :sb-wtimer :sb-dynamic-core
2015-02-21T01:06:11Z stassats: sb-sprof is broken on sb-safepoint-strictly
2015-02-21T01:06:31Z fe[nl]ix: I don't use it anyway
2015-02-21T01:06:39Z fe[nl]ix: the test hangs
2015-02-21T01:06:51Z stassats: good
2015-02-21T01:06:53Z fe[nl]ix: no discernible CPU use
2015-02-21T01:06:53Z stassats: (or not)
2015-02-21T01:07:07Z fe[nl]ix: hmm
2015-02-21T01:07:12Z fe[nl]ix: what's the current SBCL version ?
2015-02-21T01:07:18Z stassats: 1.2.8
2015-02-21T01:07:24Z fe[nl]ix: I have 1.2.7.8-79521d9e3e66 here
2015-02-21T01:07:57Z stassats: i don't seem to have an angle for this
2015-02-21T01:08:12Z stassats: (i'm sure it'll come to me tomorrow morning)
2015-02-21T01:09:29Z fe[nl]ix: the runtime is in an untinterruptible state
2015-02-21T01:09:46Z stassats: C-z kill -9 %1 works
2015-02-21T01:10:03Z stassats: but what's funny, it doesn't work when sb-safepoint-strictly is disabled
2015-02-21T01:10:56Z stassats: and bloody erc is got broken again after i updated emacs
2015-02-21T01:11:06Z fe[nl]ix: SIGTERM didn't work
2015-02-21T01:11:10Z fe[nl]ix: SIGQUIT did
2015-02-21T01:11:40Z stassats: it deadlocks with signals disabled
2015-02-21T01:11:51Z fe[nl]ix: which is funny, because the *nix elder scrolls recommend QUIT, TERM and KILL in that order
2015-02-21T01:13:18Z fe[nl]ix: I reproduced it, SIGTERM is ignored, SIGQUIT not
2015-02-21T01:13:38Z fe[nl]ix: and I'm off to sleep
2015-02-21T01:14:21Z stassats: i never understood what safepoints are waiting for and why
2015-02-21T01:14:37Z stassats: the code is too convoluted and not clear
2015-02-21T01:15:01Z stassats: just a state machine, why are the states they way they are, no idea
2015-02-21T01:17:52Z |3b| quit (Read error: Connection reset by peer)
2015-02-21T01:18:51Z stassats: and why would thread structures stashed in a hashtable affect that at all
2015-02-21T01:19:31Z |3b| joined #sbcl
2015-02-21T01:20:07Z stassats: i guess i'll have to unravel the hashtable implementation tomorrow to see where it sticks
2015-02-21T01:25:25Z stassats: actually, it doesn't have to be threads, (mapcar #'sb-thread::thread-interruptions-lock (sb-thread:list-all-threads)) works too
2015-02-21T01:25:28Z stassats: or thread-os-thread
2015-02-21T01:29:03Z nikki93 joined #sbcl
2015-02-21T01:29:39Z nikki93 quit (Remote host closed the connection)
2015-02-21T01:32:27Z Bicyclidine quit (Quit: annoy)
2015-02-21T01:38:15Z stassats: ok, it doesn't have to be threads, just anything
2015-02-21T01:38:54Z stassats: maybe it doesn't have to be hashtables either and it's just that broken
2015-02-21T01:53:12Z nyef: Hrm. Might have an angle on that code-deletion-note thing with the DCE pass.
2015-02-21T02:08:50Z stassats: so, rehashing is done without gcing
2015-02-21T02:13:34Z stassats: ok, caught it
2015-02-21T02:13:40Z stassats: it's allocating inside without-gcing
2015-02-21T02:16:52Z stassats: which is bad enough of an idea, but manages to deadlock sb-safepoint
2015-02-21T02:18:59Z stassats: http://paste.lisp.org/display/145847#3
2015-02-21T02:22:19Z stassats: vanilla sbcl is fine with this, although i can imagine plenty of problems
2015-02-21T02:28:35Z stassats: so it should probably allocate the necessary things outside of the without-gcing and only put them into the hashtable while gcing
2015-02-21T02:28:41Z stassats: while not gcing
2015-02-21T02:33:19Z stassats: and for weak hashtables, without-gcing even comes without rehashing, so it'd be harder hoisting out allocation
2015-02-21T02:34:23Z nyef: Ahh... It's issuing a note for an &OPTIONAL-PROCESSOR when there are &KEY arguments without &OPTIONAL arguments.
2015-02-21T02:34:38Z nyef: So, either squash that note... or don't emit the block in the first place.
2015-02-21T02:35:02Z nyef: How close are we to code freeze?
2015-02-21T02:35:29Z stassats: it's a short month, probably pretty nigh
2015-02-21T02:36:01Z nyef: Ah, right. Second shortest month in the Gregorian calendar.
2015-02-21T02:36:43Z nyef: Only eight days left to the end of the month.
2015-02-21T02:37:46Z stassats: so, this thing needs a twofold solution: reduce consing under without-gcing, not deadlock when it does happen
2015-02-21T02:38:27Z scymtym quit (Ping timeout: 252 seconds)
2015-02-21T02:39:01Z stassats: thread safety is yet another area where i just want to tear everything down and build anew
2015-02-21T02:39:48Z nyef: Ugh. I'm looking at the heart of IR1 handling of argument processing. I wonder if it would be easier to generate the bogus block and then suppress the note when deleting it.
2015-02-21T02:40:29Z edgar-rft quit (Quit: bye...)
2015-02-21T02:45:38Z stassats: i guess it hits a safepoint while inside without-gcing and then waits for without-gcing to end or something
2015-02-21T02:46:34Z nyef: Maybe add an assert to the safepoint logic so that it will lose() instead of deadlocking in such cases?
2015-02-21T02:46:52Z nyef: Or is this a two-thread deadlock?
2015-02-21T02:47:31Z stassats: losing is really not an option
2015-02-21T02:47:45Z stassats: it's a multiple thread deadlock
2015-02-21T02:47:52Z nyef: Ah, damn.
2015-02-21T02:48:26Z nyef is still a single-threaded, occasionally-event-driven man at heart.
2015-02-21T02:48:59Z nyef: Probably from having done much of the work to get PPC threads up and running.
2015-02-21T02:49:17Z nyef: Even if they weren't as stable as I thought they were for the first year or so.
2015-02-21T02:50:09Z ASau quit (Ping timeout: 256 seconds)
2015-02-21T02:50:32Z stassats: i like threads for interactivity, not for throughput
2015-02-21T02:50:47Z nyef: ... fifteen arguments to GENERATE-OPTIONAL-DEFAULT-ENTRY? WTF?
2015-02-21T02:51:47Z stassats: better than 15 &optionals followed by a &key
2015-02-21T02:52:13Z stassats: (which is something out of clim)
2015-02-21T02:52:29Z nyef: Or no &OPTIONALS followed by 15 &KEYs, which would trigger a code-deletion-note on this branch.
2015-02-21T02:54:30Z stassats: this deadlock issue looks easy enough, until i remember that it involves the safepoint code
2015-02-21T03:08:31Z joshe joined #sbcl
2015-02-21T03:35:35Z oleo is now known as Guest68710
2015-02-21T03:36:33Z oleo__ joined #sbcl
2015-02-21T03:38:21Z christoph_debian quit (Ping timeout: 250 seconds)
2015-02-21T03:38:57Z Guest68710 quit (Ping timeout: 250 seconds)
2015-02-21T03:49:40Z nikki93 joined #sbcl
2015-02-21T03:52:08Z christoph_debian joined #sbcl
2015-02-21T03:53:11Z nikki93 quit (Remote host closed the connection)
2015-02-21T04:14:20Z stassats quit (Ping timeout: 244 seconds)
2015-02-21T04:16:04Z gabriel_laddel joined #sbcl
2015-02-21T05:02:21Z ASau joined #sbcl
2015-02-21T05:21:06Z nyef quit (Quit: G'night all)
2015-02-21T05:42:03Z ASau quit (Remote host closed the connection)
2015-02-21T05:44:14Z ASau joined #sbcl
2015-02-21T06:19:13Z csziacobus quit (Quit: csziacobus)
2015-02-21T06:35:49Z gabriel_laddel quit (Ping timeout: 264 seconds)
2015-02-21T06:57:22Z |3b| quit (Read error: Connection reset by peer)
2015-02-21T06:58:57Z |3b| joined #sbcl
2015-02-21T07:04:19Z gabriel_laddel joined #sbcl
2015-02-21T07:06:04Z gingerale joined #sbcl
2015-02-21T07:11:47Z gabriel_laddel quit (Remote host closed the connection)
2015-02-21T07:18:53Z Shinmera joined #sbcl
2015-02-21T07:48:09Z sdemarre joined #sbcl
2015-02-21T08:16:34Z angavrilov joined #sbcl
2015-02-21T08:33:05Z rszeno quit (Quit: Leaving.)
2015-02-21T09:19:29Z edgar-rft joined #sbcl
2015-02-21T10:18:20Z sdemarre quit (Ping timeout: 265 seconds)
2015-02-21T10:32:29Z sdemarre joined #sbcl
2015-02-21T10:48:19Z rjmacready joined #sbcl
2015-02-21T11:08:13Z sdemarre quit (Ping timeout: 264 seconds)
2015-02-21T11:15:00Z oleo__ quit (Quit: Verlassend)
2015-02-21T11:39:31Z oleo joined #sbcl
2015-02-21T11:47:58Z scymtym joined #sbcl
2015-02-21T12:45:52Z oleo quit (Read error: Connection reset by peer)
2015-02-21T12:50:07Z sdemarre joined #sbcl
2015-02-21T12:53:52Z oleo joined #sbcl
2015-02-21T13:29:15Z rszeno joined #sbcl
2015-02-21T13:49:22Z rjmacready quit (Ping timeout: 246 seconds)
2015-02-21T13:50:17Z rszeno quit (Ping timeout: 245 seconds)
2015-02-21T13:58:31Z LiamH joined #sbcl
2015-02-21T14:03:07Z LiamH quit (Client Quit)
2015-02-21T14:07:27Z rszeno joined #sbcl
2015-02-21T14:24:35Z rszeno quit (Quit: Leaving.)
2015-02-21T14:32:30Z nyef joined #sbcl
2015-02-21T14:32:46Z nyef: G'morning all.
2015-02-21T14:39:08Z oleo: morning
2015-02-21T14:54:17Z stassats joined #sbcl
2015-02-21T15:10:54Z attila_lendvai joined #sbcl
2015-02-21T15:10:54Z attila_lendvai quit (Changing host)
2015-02-21T15:10:54Z attila_lendvai joined #sbcl
2015-02-21T15:35:11Z LiamH joined #sbcl
2015-02-21T15:52:39Z stassats` joined #sbcl
2015-02-21T15:55:11Z oleo is now known as Guest63006
2015-02-21T15:56:08Z oleo__ joined #sbcl
2015-02-21T15:57:33Z stassats` quit (Remote host closed the connection)
2015-02-21T15:58:12Z Guest63006 quit (Ping timeout: 245 seconds)
2015-02-21T16:05:42Z edgar-rft quit (Quit: bye...)
2015-02-21T16:06:49Z stassats quit (Remote host closed the connection)
2015-02-21T16:08:37Z edgar-rft joined #sbcl
2015-02-21T16:15:23Z attila_lendvai quit (Ping timeout: 252 seconds)
2015-02-21T16:20:47Z stassats joined #sbcl
2015-02-21T16:21:49Z stassats` joined #sbcl
2015-02-21T16:24:16Z stassats: scymtym: more interesting would be to make M-. work on mixin VOPs
2015-02-21T16:28:05Z Xof: there's a whole freeze period coming up for refinements like that
2015-02-21T16:28:33Z stassats: WARNING! Some of the contrib modules did not build successfully or pass their self-tests. Failed contribs:" sb-introspect
2015-02-21T16:28:36Z stassats: huh
2015-02-21T16:28:57Z stassats: 1 out of 96 total tests failed: SB-INTROSPECT-TEST::DEFINITION-SOURCE.1.
2015-02-21T16:30:07Z stassats: i see
2015-02-21T16:33:54Z stassats: Xof: when will it freeze?
2015-02-21T16:34:58Z nyef: Xof: I'm curious about when we're going to be entering freeze as well.
2015-02-21T16:35:25Z nyef: ("I, too, am curious as to when we will be entering the freeze period.")
2015-02-21T16:37:10Z stassats: i have this deadlock to solve and the constant propagation thing
2015-02-21T16:37:14Z stassats: both are non critical
2015-02-21T16:37:27Z stassats: (unless you are a windows user, for the former)
2015-02-21T16:40:07Z scymtym: stassats: sorry, just noticed on the build server. did not happen when i tested locally for some reason. investigating.
2015-02-21T16:40:18Z stassats: scymtym: i already fixed it
2015-02-21T16:40:37Z scymtym: thanks
2015-02-21T16:41:25Z Xof: I just sent mail
2015-02-21T16:43:44Z Xof: uh, no I didn't
2015-02-21T16:43:48Z Xof: I will be doing that RSN
2015-02-21T16:43:58Z Xof: aiming for release on Fridayish
2015-02-21T16:45:32Z stassats: then it's time for some testing
2015-02-21T16:46:06Z nyef: Okay, so I don't have time to get the DCE pass in place this cycle.
2015-02-21T16:46:32Z nyef: (Which is fine, it's been waiting months already for me to have an angle on why it causes that regression.)
2015-02-21T16:48:53Z edgar-rft quit (Quit: bye...)
2015-02-21T16:49:28Z stassats: and maybe it's time to setup autobuilds using virtualbox
2015-02-21T16:49:33Z edgar-rft joined #sbcl
2015-02-21T16:58:24Z scymtym: stassats: are you interested in the Jenkins-based setup?
2015-02-21T16:58:44Z stassats: i don't need tests or reports, just binaries
2015-02-21T16:58:51Z scymtym: i see
2015-02-21T16:59:22Z stassats: maybe i just should with a script that builds and uploads
2015-02-21T16:59:32Z stassats: then worry about starting up vms and launching it
2015-02-21T17:01:39Z stassats: maybe then i'll want reports and stuff, instead of finding out that something doesn't build just after the release
2015-02-21T17:15:05Z edgar-rft quit (Quit: bye...)
2015-02-21T17:15:39Z edgar-rft joined #sbcl
2015-02-21T17:26:38Z edgar-rft quit (Quit: bye...)
2015-02-21T17:27:47Z edgar-rft joined #sbcl
2015-02-21T17:28:26Z karswell joined #sbcl
2015-02-21T17:41:17Z edgar-rft quit (Quit: bye...)
2015-02-21T17:43:07Z edgar-rft joined #sbcl
2015-02-21T18:24:48Z csziacobus joined #sbcl
2015-02-21T18:28:26Z foom quit (Ping timeout: 265 seconds)
2015-02-21T18:32:06Z nikki93 joined #sbcl
2015-02-21T18:33:04Z edgar-rft quit (Quit: bye...)
2015-02-21T18:34:06Z nikki93 quit (Remote host closed the connection)
2015-02-21T18:34:55Z edgar-rft joined #sbcl
2015-02-21T18:40:59Z foom joined #sbcl
2015-02-21T19:04:49Z stassats quit (Ping timeout: 245 seconds)
2015-02-21T19:04:55Z stassats` quit (Ping timeout: 255 seconds)
2015-02-21T19:08:14Z edgar-rft quit (Quit: bye...)
2015-02-21T19:10:26Z edgar-rft joined #sbcl
2015-02-21T19:23:07Z akkad quit (Ping timeout: 252 seconds)
2015-02-21T19:24:06Z Intensity quit (Ping timeout: 246 seconds)
2015-02-21T19:24:47Z psy_ quit (Ping timeout: 246 seconds)
2015-02-21T19:26:14Z psy_ joined #sbcl
2015-02-21T19:27:21Z akkad joined #sbcl
2015-02-21T19:34:46Z Intensity joined #sbcl
2015-02-21T19:41:15Z oleo__ quit (Quit: Verlassend)
2015-02-21T19:41:35Z Intensity quit (Ping timeout: 246 seconds)
2015-02-21T19:49:27Z sdemarre quit (Ping timeout: 245 seconds)
2015-02-21T19:49:42Z oleo joined #sbcl
2015-02-21T19:56:04Z stassats joined #sbcl
2015-02-21T20:34:25Z joshe: fwiw, I've started putting together a buildbot cluster for sbcl
2015-02-21T21:18:59Z Bicyclidine joined #sbcl
2015-02-21T21:32:28Z edgar-rft quit (Quit: bye...)
2015-02-21T21:34:01Z edgar-rft joined #sbcl
2015-02-21T21:40:32Z edgar-rft quit (Quit: bye...)
2015-02-21T21:42:21Z edgar-rft joined #sbcl
2015-02-21T21:44:55Z edgar-rft quit (Client Quit)
2015-02-21T21:46:52Z edgar-rft joined #sbcl
2015-02-21T21:54:29Z angavrilov quit (Remote host closed the connection)
2015-02-21T21:55:02Z edgar-rft quit (Quit: bye...)
2015-02-21T21:55:49Z edgar-rft joined #sbcl
2015-02-21T22:19:37Z karswell quit (Ping timeout: 264 seconds)
2015-02-21T22:21:48Z nikki93 joined #sbcl
2015-02-21T22:26:36Z nikki93 quit (Remote host closed the connection)
2015-02-21T22:47:04Z eudoxia joined #sbcl
2015-02-21T22:54:50Z oleo quit (Read error: Connection reset by peer)
2015-02-21T23:01:22Z oleo joined #sbcl
2015-02-21T23:04:50Z scymtym quit (Remote host closed the connection)
2015-02-21T23:10:51Z eudoxia_ joined #sbcl
2015-02-21T23:14:09Z eudoxia quit (Ping timeout: 250 seconds)
2015-02-21T23:29:48Z stassats` joined #sbcl
2015-02-21T23:32:00Z nikki93 joined #sbcl
2015-02-21T23:33:11Z gingerale quit (Ping timeout: 252 seconds)
2015-02-21T23:44:59Z nikki93 quit (Remote host closed the connection)
2015-02-21T23:57:42Z rszeno joined #sbcl