2015-12-14T00:04:16Z scymtym quit (Ping timeout: 272 seconds)
2015-12-14T00:22:20Z SamSkulls quit (Ping timeout: 250 seconds)
2015-12-14T00:24:46Z Guest13132 quit (Ping timeout: 240 seconds)
2015-12-14T00:27:29Z SamSkulls joined #sbcl
2015-12-14T00:30:08Z stassats quit (Ping timeout: 256 seconds)
2015-12-14T00:32:57Z whiteline joined #sbcl
2015-12-14T00:39:30Z scymtym joined #sbcl
2015-12-14T01:09:23Z eudoxia joined #sbcl
2015-12-14T01:55:44Z scymtym quit (Ping timeout: 272 seconds)
2015-12-14T02:34:46Z sjl quit (Ping timeout: 240 seconds)
2015-12-14T02:38:11Z sjl joined #sbcl
2015-12-14T02:54:39Z lnostdal_ joined #sbcl
2015-12-14T02:58:20Z lnostdal__ quit (Ping timeout: 250 seconds)
2015-12-14T03:15:08Z loke_ is now known as loke
2015-12-14T03:42:26Z lnostdal_ quit (Ping timeout: 240 seconds)
2015-12-14T03:42:46Z lnostdal_ joined #sbcl
2015-12-14T04:17:51Z karswell quit (Read error: Connection reset by peer)
2015-12-14T04:18:05Z karswell joined #sbcl
2015-12-14T04:28:24Z eudoxia quit (Quit: Leaving)
2015-12-14T04:34:03Z psy_ quit (Remote host closed the connection)
2015-12-14T04:37:34Z dougk_ quit (Ping timeout: 250 seconds)
2015-12-14T04:38:23Z dougk_ joined #sbcl
2015-12-14T05:53:26Z ZabaQ joined #sbcl
2015-12-14T06:32:30Z ASau quit (Ping timeout: 272 seconds)
2015-12-14T06:38:52Z scymtym__ quit (Remote host closed the connection)
2015-12-14T06:41:37Z ASau joined #sbcl
2015-12-14T06:45:11Z ZabaQ: Are there any gotchas for successfully attaching gdb to x86 sbcl?
2015-12-14T06:45:46Z ZabaQ: It seems to segfault shortly after I do.
2015-12-14T06:46:35Z karswell quit (Remote host closed the connection)
2015-12-14T06:46:37Z karswell` joined #sbcl
2015-12-14T06:55:32Z loke: ZabaQ: You have to ignore segfaults
2015-12-14T06:55:50Z loke: they are regularly sent inside SBCL as part of the GC
2015-12-14T06:55:50Z ZabaQ: Logical :-)
2015-12-14T06:55:55Z adhoc quit (Ping timeout: 240 seconds)
2015-12-14T06:56:12Z loke: there is a GDB command to ensure that they are passed along to the application
2015-12-14T06:58:05Z adhoc joined #sbcl
2015-12-14T07:01:55Z ZabaQ: ..so I guess if you are trying to find a SIGSEGV in a foreign library, things get interesting.
2015-12-14T07:03:55Z loke: ZabaQ: I believe segvs from other libraries are flagged as such, resulting in a drop into ldb
2015-12-14T07:43:29Z ehaliewicz joined #sbcl
2015-12-14T08:00:10Z brucem quit (Quit: ZNC - http://znc.sourceforge.net)
2015-12-14T08:11:53Z luis quit (Quit: ZNC - http://znc.sourceforge.net)
2015-12-14T08:21:46Z sjl quit (Ping timeout: 240 seconds)
2015-12-14T08:21:47Z p_l joined #sbcl
2015-12-14T08:24:59Z scymtym joined #sbcl
2015-12-14T08:26:59Z p_l quit (Quit: leaving)
2015-12-14T08:27:19Z p_l joined #sbcl
2015-12-14T08:28:45Z Cymew joined #sbcl
2015-12-14T08:31:30Z flip214: ZabaQ: you could tell gdb to check the address on segv before stopping, but you're right... that's cumbersome
2015-12-14T08:40:15Z scymtym_ joined #sbcl
2015-12-14T08:42:34Z lnostdal_ quit (Ping timeout: 256 seconds)
2015-12-14T08:45:54Z ZabaQ: I think It'll be ok for my purposes. The foreign library I'm working with is small, and it works fine from "C"
2015-12-14T08:46:16Z lnostdal_ joined #sbcl
2015-12-14T08:46:30Z ZabaQ: It's probably simply a misunderstanding on my part of how CFFI works, when I break into the problem function it will probably be evident from examination.
2015-12-14T08:52:31Z scymtym: fe[nl]ix: does your user initialization enable ASDF's "deferred condition" mechanism? iiuc, there is a SB-C::UNDEFINED-WARING for :DEFAULT at some point (maybe calling :DEFAULT as a function?) that ASDF tries to serialize. could you try making compilation and loading of setup.lisp verbose in your initialization? i hope that will print the warning immediately and with more context.
2015-12-14T08:59:14Z ASau quit (Read error: Connection reset by peer)
2015-12-14T08:59:53Z ASau joined #sbcl
2015-12-14T09:04:18Z fe[nl]ix: scymtym: it's "undefined type :DEFAULT"
2015-12-14T09:04:50Z scymtym_: fe[nl]ix: thanks, that's a better lead
2015-12-14T09:07:44Z fe[nl]ix: Ok, I see
2015-12-14T09:07:57Z fe[nl]ix: ASDF has a (defvar *default-stream-element-type* :default)
2015-12-14T09:08:52Z fe[nl]ix: scymtym: can you look into that ?
2015-12-14T09:08:55Z fe[nl]ix: I need to go to sleep
2015-12-14T09:10:45Z scymtym_: fe[nl]ix: will do, but dayjob first
2015-12-14T09:10:52Z scymtym_: sorry for the inconvenience
2015-12-14T09:14:41Z sjl joined #sbcl
2015-12-14T09:32:55Z dougk_ quit (Ping timeout: 240 seconds)
2015-12-14T09:36:33Z brucem joined #sbcl
2015-12-14T09:37:12Z brucem quit (Changing host)
2015-12-14T09:37:12Z brucem joined #sbcl
2015-12-14T09:41:05Z edgar-rft joined #sbcl
2015-12-14T09:50:00Z ASau quit (Ping timeout: 250 seconds)
2015-12-14T09:57:22Z whiteline quit (Ping timeout: 256 seconds)
2015-12-14T10:06:16Z minion quit (Remote host closed the connection)
2015-12-14T10:06:16Z specbot quit (Remote host closed the connection)
2015-12-14T10:06:35Z specbot joined #sbcl
2015-12-14T10:06:36Z minion joined #sbcl
2015-12-14T10:14:22Z attila_lendvai joined #sbcl
2015-12-14T10:14:22Z attila_lendvai quit (Changing host)
2015-12-14T10:14:22Z attila_lendvai joined #sbcl
2015-12-14T10:18:13Z lnostdal__ joined #sbcl
2015-12-14T10:21:44Z lnostdal_ quit (Ping timeout: 256 seconds)
2015-12-14T10:22:46Z ehaliewicz quit (Ping timeout: 240 seconds)
2015-12-14T10:28:37Z scymtym_: ok, i know what's going on
2015-12-14T10:34:20Z ASau joined #sbcl
2015-12-14T10:36:45Z DeadTrickster quit (Read error: No route to host)
2015-12-14T10:51:12Z ASau quit (Ping timeout: 256 seconds)
2015-12-14T10:51:45Z DeadTrickster joined #sbcl
2015-12-14T11:11:14Z ignotus quit (Ping timeout: 246 seconds)
2015-12-14T11:28:45Z scymtym_: fe[nl]ix: should be fixed in master
2015-12-14T11:37:28Z sjl quit (Ping timeout: 250 seconds)
2015-12-14T11:41:17Z stassats joined #sbcl
2015-12-14T11:42:13Z sjl joined #sbcl
2015-12-14T11:47:52Z stassats quit (Ping timeout: 250 seconds)
2015-12-14T11:49:44Z stassats joined #sbcl
2015-12-14T11:50:19Z dougk_ joined #sbcl
2015-12-14T12:36:57Z hajovonta joined #sbcl
2015-12-14T12:36:59Z hajovonta: hi
2015-12-14T12:37:40Z hajovonta: I encountered the following behavior: http://pastebin.com/SkMghjnK
2015-12-14T12:38:15Z hajovonta: it is not exactly the way I expected...
2015-12-14T12:38:29Z stassats: sure
2015-12-14T12:38:38Z stassats: you're changing the closure in the loop
2015-12-14T12:38:59Z hajovonta: how can I avoid this?
2015-12-14T12:39:06Z stassats: (mapcar #'funcall (loop for item in '(5 6 7 8 9 10) collect (lambda () item)))
2015-12-14T12:39:14Z stassats: without time effects of threads
2015-12-14T12:39:56Z stassats: hajovonta: do not reuse the closure
2015-12-14T12:40:24Z hajovonta: that gives all tens...
2015-12-14T12:41:12Z stassats: you won't see that at the Olympics, would you?
2015-12-14T12:41:36Z hajovonta: :)
2015-12-14T12:42:45Z flip214: (loop for item in '(5 6 7 8 9 10) collect (let ((i item)) (lambda () i))))
2015-12-14T12:44:36Z stassats: i would go with (let ((item item))
2015-12-14T12:46:14Z hajovonta: that's great, thanks
2015-12-14T12:46:56Z hajovonta: is it a compiler-optimization thing or why does the closure gets reused?
2015-12-14T12:47:55Z stassats: because it can
2015-12-14T13:47:42Z lnostdal_ joined #sbcl
2015-12-14T13:50:46Z lnostdal__ quit (Ping timeout: 272 seconds)
2015-12-14T13:54:39Z stassats: copy-propagate is actually the culprit
2015-12-14T13:54:43Z stassats: on lots branches
2015-12-14T14:00:45Z sjl quit (Read error: Connection reset by peer)
2015-12-14T14:02:06Z sjl joined #sbcl
2015-12-14T14:35:02Z adhoc quit (Ping timeout: 256 seconds)
2015-12-14T14:36:50Z adhoc joined #sbcl
2015-12-14T15:10:49Z nyef: Today's thought: A page-aligned fault address indicates a possible TLB or cache issue. There's a known TLB issue on my main MIPS system that is not present on a second MIPS system that I have. And if the fault happens while building contribs on one main build, it always happens while building contribs for that main build.
2015-12-14T15:11:52Z nyef: Therefore: Build a MIPS version that has the fault on my main system, copy the entire thing over to the secondary system, and rebuild the contribs there. If the fault still happens, it's probably not a TLB/cache issue. If it DOESN'T happen, then it's very possibly a TLB or cache issue.
2015-12-14T15:11:53Z flip214: hmmm, I'd think of an uninitialized variable on the stack that just happens to be "wrong" on *that* machine
2015-12-14T15:12:05Z flip214: good idea
2015-12-14T15:12:20Z dougk_ quit (Remote host closed the connection)
2015-12-14T15:14:41Z nyef: The problem is that this fault has been observed on two different architectures now, and attempting to get a debugger onto it doesn't work, and there's not enough information available yet to be able to try to reproduce it in a context where a debugger (even LDB!) would work.
2015-12-14T15:16:18Z stassats: what about gdb?
2015-12-14T15:17:01Z nyef: It's only been observed so far while building contribs. The processes die when the lossage is detected.
2015-12-14T15:17:35Z stassats: disable dying then
2015-12-14T15:17:55Z nyef: Attempting to rebuild whatever contrib failed by re-running that part of the build manually causes the build to succeed, rather than reproducing the fault.
2015-12-14T15:19:04Z stassats: always disable it
2015-12-14T15:19:31Z nyef: That's probably the next step, yes.
2015-12-14T15:20:08Z nyef: Which, knowing my luck, will disturb whatever delicate balance causes it.
2015-12-14T15:23:17Z stassats: sset-union-of-difference in SB-C::COPY-FLOW-ANALYSIS is responsible for most of the consing
2015-12-14T15:31:56Z stassats: reduced it somewhat, but still not good
2015-12-14T15:32:18Z stassats: why isn't it collected, though?
2015-12-14T15:34:30Z nyef: What's holding it live?
2015-12-14T15:40:33Z stassats: interestingly, block-gen is not used
2015-12-14T15:40:40Z stassats: doesn't help much, but at least something
2015-12-14T15:48:09Z stassats: it may have been holding to some garbage, though, let's see
2015-12-14T15:48:48Z stassats: pretty unlikely, though, since tns are not being destroyed
2015-12-14T15:49:21Z ignotus joined #sbcl
2015-12-14T15:49:22Z ignotus quit (Changing host)
2015-12-14T15:49:22Z ignotus joined #sbcl
2015-12-14T15:53:13Z stassats: interestingly, (declare (optimize compilation-speed)) doesn't have an effect on copy-propagate
2015-12-14T15:53:20Z stassats: (sb-ext:restrict-compiler-policy 'compilation-speed 3) does
2015-12-14T15:58:37Z mordocai joined #sbcl
2015-12-14T16:10:33Z rpg joined #sbcl
2015-12-14T16:27:06Z ZabaQ quit (Ping timeout: 240 seconds)
2015-12-14T16:30:32Z whiteline joined #sbcl
2015-12-14T16:32:05Z stassats: a better understanding of copy propagation would help
2015-12-14T16:43:28Z edgar-rft quit (Quit: edgar-rft)
2015-12-14T16:50:26Z Cymew quit (Ping timeout: 240 seconds)
2015-12-14T17:19:48Z stassats: copy-propagation doesn't actually do anything for my test case
2015-12-14T17:24:55Z core1 joined #sbcl
2015-12-14T17:28:35Z eudoxia joined #sbcl
2015-12-14T17:32:49Z psy_ joined #sbcl
2015-12-14T17:33:24Z psy_ quit (Max SendQ exceeded)
2015-12-14T17:33:45Z psy_ joined #sbcl
2015-12-14T17:38:53Z eudoxia quit (Quit: Leaving)
2015-12-14T17:50:14Z DGASAU quit (Read error: Connection reset by peer)
2015-12-14T17:50:39Z DGASAU joined #sbcl
2015-12-14T18:01:50Z SamSkulls quit (Ping timeout: 250 seconds)
2015-12-14T19:01:24Z gingerale joined #sbcl
2015-12-14T19:01:28Z rpg quit (Quit: rpg)
2015-12-14T19:09:52Z DGASAU quit (Ping timeout: 250 seconds)
2015-12-14T19:12:46Z DGASAU joined #sbcl
2015-12-14T19:15:09Z eudoxia joined #sbcl
2015-12-14T20:15:18Z scymtym_ quit (Ping timeout: 250 seconds)
2015-12-14T20:15:47Z prxq joined #sbcl
2015-12-14T20:15:51Z mordocai quit (Quit: ERC (IRC client for Emacs 24.5.1))
2015-12-14T20:16:29Z mordocai joined #sbcl
2015-12-14T20:16:57Z nzambe joined #sbcl
2015-12-14T20:27:03Z fiveop joined #sbcl
2015-12-14T20:34:55Z eudoxia quit (Quit: Leaving)
2015-12-14T20:58:01Z fiveop quit
2015-12-14T21:03:52Z scymtym_ joined #sbcl
2015-12-14T21:07:21Z gingerale quit (Remote host closed the connection)
2015-12-14T21:13:51Z csziacobus joined #sbcl
2015-12-14T21:30:36Z clique joined #sbcl
2015-12-14T21:32:59Z stassats: i see where all the consing is coming from
2015-12-14T21:33:37Z stassats: it combines tns from the current and the predecessor blocks
2015-12-14T21:33:44Z stassats: and the number grows
2015-12-14T21:36:30Z stassats: a lot of different tns, a lot of blocks due to branches, worst case scenario
2015-12-14T21:41:01Z stassats: maybe there is a way to get a lot of blocks without jumps
2015-12-14T21:51:43Z clique: Which is the meaning of TNs? Is it an acronym?
2015-12-14T21:51:56Z sjl quit (Ping timeout: 256 seconds)
2015-12-14T21:52:04Z clique: now I'm reading here: https://github.com/sbcl/sbcl/blob/master/src/compiler/tn.lisp
2015-12-14T21:52:21Z clique: but I'm not able to fully understand the meaning.
2015-12-14T21:57:10Z csziacobus: clique: TN stands for temporary name
2015-12-14T21:57:23Z csziacobus: think of them like infinite register
2015-12-14T21:57:25Z csziacobus: s
2015-12-14T21:57:44Z clique: oh... thank you
2015-12-14T21:59:07Z clique: but think of them as register is not very helpful for me
2015-12-14T21:59:51Z clique: infinite registers...
2015-12-14T22:00:47Z clique: data register, address register, instruction register... I don't know what you're referring to
2015-12-14T22:04:03Z ASau joined #sbcl
2015-12-14T22:04:59Z clique: anyway reading here http://www.sbcl.org/gsoc2013/ideas/ is a little bit more clear
2015-12-14T22:05:25Z clique: sorry, but it's the first time I heard tthe term "TNs"
2015-12-14T22:06:59Z prxq quit (Remote host closed the connection)
2015-12-14T22:36:14Z edgar-rft joined #sbcl
2015-12-14T22:48:02Z sjl joined #sbcl
2015-12-14T22:48:56Z PuercoPop: clique: the acronyms are defined in the top-level file TLA
2015-12-14T22:50:08Z PuercoPop: (which is itself a TLA, glossary more 'discoverable' name imho)
2015-12-14T22:51:45Z clique quit (Ping timeout: 252 seconds)
2015-12-14T23:04:46Z attila_lendvai quit (Ping timeout: 240 seconds)
2015-12-14T23:10:42Z stassats quit (Ping timeout: 256 seconds)
2015-12-14T23:25:03Z whiteline quit (Quit: Leaving)
2015-12-14T23:33:46Z whiteline joined #sbcl
2015-12-14T23:34:56Z whiteline quit (Client Quit)
2015-12-14T23:40:11Z whiteline joined #sbcl
2015-12-14T23:41:23Z whiteline quit (Client Quit)
2015-12-14T23:55:23Z nyef: Hrm. Okay, fault on MIPS while running SB-GMP tests, fault address is 0x30fc0000, ALLOC is 0x30fc42d0, TRIGGER is 0x30fc0000, no interrupt contexts. Backtrace says SB-GMP::RANDOM-BITCOUNT is the most recent two frames on the stack.
2015-12-14T23:57:35Z nyef: Well, at least this explains why it's always a page-aligned address. (-:
2015-12-14T23:57:58Z mordocai quit (Remote host closed the connection)