Index

Show enters and exits. Hide enters and exits.

04:20:49anierorunning the tests for main 4.2.0 in rbx pegs my cpu and takes forever. any idea what i'd look at to track it down?
04:24:02brixenaniero: can you ps ux | grep rbx to get the PID, then gdb -p <pid> to attach to process
04:24:07brixenthen bt and paste a gist
04:24:20brixenaniero: also, is this repeatable or sporadic?
04:24:39brixenaniero: bt and also info threads
04:25:12brixenaniero: btw, what is main 4.2.0?
04:25:54anierobrixen: ruby gem, by ara howard... lots of magic
04:26:00brixenhaha
04:26:02brixenok :)
04:26:16anierook weird, now it's fast :)
04:26:17brixenhas a short list of "lots of magic folks" :)
04:27:00anierowell, never mind, it's fast. slightly broken too, seem
04:27:15brixenbtw, we should punch ppl that name gems 'main' for the pain of googling :)
04:27:30brixenthankfully rubygems.org is now gemcutter.org
04:27:58brixenaniero: pls file a ticket with steps to repro, we'll fix it! :)
04:29:08anierono problem!
04:29:14brixenthanks!
04:29:15anierothanks. i don't think rbx is at fault here
04:29:44brixenwe're never at fault, but we often fix issues where folks fail to conform to the rbx API :D
04:30:29anieroit's Just Ruby, but the error is deep in magical code somewhere. it'll take me a bit
04:30:37brixenok
04:30:38anieroalso trying to check a few libs against 1.9.2-preview3
04:30:45brixenahh fun
04:40:09anierobrixen: happened again! https://gist.github.com/240d382dc037d7334552
04:49:52brixensorry, nursing a cold here :(
04:50:02anierono problem. i'll tackle this later
04:50:03brixendo you still have that gdb session open?
04:50:09brixenone sec...
04:50:17brixenif it's open, can you hit c
04:50:23brixenlet it run a bit, then hit ^C
04:50:32brixenlooks like it's thrashing in the GC maybe
04:50:48anieroah shoot! just closed it
04:50:54brixenok, no worries
04:50:59brixenbut if it hangs again, try that
04:51:09anierook
04:51:20brixencool, thanks
04:51:36brixenalso, if you think you have conditions that make it likely to happen, file a ticket
04:51:41anierook, i hit ^C
04:52:03brixenactually, file a ticket anyway and just give the best info you have
04:52:10anieroalright!
04:52:14brixensteps to run, what to expect, bt gist, etc
04:52:26brixenlet it run a bit, ^C and see if it's often in the GC
04:52:39brixenhm, we could try attaching with the agent too
04:54:18brixenaniero: if you start your main process with -Xagent.start, then in another terminal, run bin/rbx console
04:54:24brixenthen get system.memory
04:54:32brixenthen you could print those attributes
04:55:19brixeneg system.memory.young, system.memory.mature, system.memory.large, system.memory.counter
04:55:34brixenthe counter one would be good between ^C's so you can see the allocation rate
04:55:57brixensorry, the command is 'get system.memory' etc
04:56:05brixenand I'm about to hit the bed :-/
04:56:18anierobrixen: http://github.com/evanphx/rubinius/issues/issue/381
04:56:33anierobut me too, i'll try that tomorrow
04:56:34anierothanks!
04:56:34brixenhmm, ok
04:56:41brixenso, this happens when compiling yes?
04:56:47brixenie, after removing all the .rbc?
04:57:03anierowell, presumably after compiling
04:57:09anieroit's running, printing .'s from the tests
04:57:39anierook, nm, don't have to delete the rbc's
04:57:47brixenok
04:58:01brixenwell, thanks for all the info!
04:58:08brixentry with the agent tomorrow, pls
17:21:00evanis the person working on msgpack in here?
17:35:26dbussinkevan: i guess that means there isn't ;)
17:35:30brixenI think his irc nick is bcg too
17:35:31evanyep :)
17:35:34evank
17:35:44brixenevan: did you see my response to his pull request?
17:35:52evanumm
17:35:52brixenI don't know how github delivers those
17:35:53evanno
17:35:57evanit's not sent to everyone when you respond
17:35:59evanjust him.
17:36:25brixenhttp://gist.github.com/448796
17:36:54brixenthey should have a reply all in that if they're going to send it to multiple people
17:37:27evancool.
17:37:34evanhe also didn't need to change msgpack at all
17:37:42evanhe could have done 100% of the work in rubinius
17:37:44evanie
17:37:54evanexposing rb_big_sign() in msgpack is wrong.
17:38:02brixenyep
17:38:30brixenthat was the point, but I should have explicitly said so
19:14:04evanrad
19:14:13evanyou can ask for a heap dump over query agent
19:20:31boyscoutAdd heapdump capability for memory debugging - 0d53d39 - Evan Phoenix
19:24:43evanboyscout: http://github.com/evanphx/Gauge
19:27:03boyscoutCI: Commit 0d53d39 failed. http://github.com/evanphx/rubinius/commit/0d53d39f9cb110949a97d3db36280a141535b0b6
19:27:09evan:/
19:29:14boyscoutAdd missing includes - 0dee73d - Evan Phoenix
19:30:13dbussinkevan: interested in a weird bug? :P
19:30:23evanafter lunch, sure.
19:30:43dbussinkit's a pretty complex one though, haven't been able to disect it more :S
19:31:58evanlet me see it
19:32:00evani'll muse over lunch.
19:33:53dbussinkevan: i'll write up an issue for it
19:34:37dbussinkevan: basically comes down to weird spec failures in dkubb's veritas library: http://github.com/dkubb/veritas
19:34:38evank
19:34:50evanalrighty.
19:36:07evani'm going to grab lunch then.
19:39:49dbussinkevan: http://github.com/evanphx/rubinius/issues/issue/383
19:43:57boyscoutCI: rubinius: 0dee73d successful: 3456 files, 13603 examples, 41167 expectations, 0 failures, 0 errors
20:20:10dbussinkevan: does the issue contain enough detail?
20:29:40evandbussink: i'd like to see your output of running that spec
20:30:04evanso that I can verify i'm seeing the same thing when I go to fix it.
20:30:16evandbussink: why is backport used?
20:30:20evani suspect that might be the issue.
20:30:30dbussinkevan: i've tried without, still the same issue
20:30:48dbussinkevan: for running specs on pre 1.8.7 versions
20:30:57evanok
20:31:07evancan you include a gist url to your output running that spec?
20:31:52dbussinkevan: i've added the output to the issue
20:31:56evank
20:32:24evanis there no <=> method?
20:32:31evanif you could indicate what behavior you expect to see
20:32:33evanthat would also help
20:35:12dbussinkevan: well, if you look at that limit_spec file that is run, you see that it defines some relations that are projected
20:35:19evanok
20:35:33evani'm going to rework how we load extensions
20:35:37evanthen i'll work on that bug
20:35:42dbussinkit basically defines a relation containing two elements, each containing one entry
20:35:48dbussinkthat relation is then ordered
20:35:56dbussinkand then a limit of 1 element is taken
20:36:13dbussinkevan: ok, np :)
20:36:40dbussinkevan: if dkubb is around i'll ask him if he could help you out further later today
20:37:03evanwill do.
20:39:38dbussinkevan: btw, should i add an issue for adding packing for ivars used in modules?
20:39:46dbussinkevan: or is that on some other todo list?
20:39:56dbussinkjust want to make it's not forgotten
20:40:09evango ahead and add it.
20:40:12evantag it as feature
20:40:16evanso I don't forget.
20:42:05dbussinkevan: ok, created
20:42:29evandanku!
20:43:39dbussinkyour dutch is improving :P
20:43:51evanhah
20:44:08dbussinkthere's a space missing though ;)
20:44:10dbussinkdank u
20:44:40evanhah
20:44:54evani don't even know where I picked that up
20:44:58evanbut i didn't even know it was dutch.
20:45:10evancould have been in amsterdam so many years ago
20:45:15evanand i've forgotten.
20:48:27dbussinki wonder how you pronounce it in your head though :p
20:48:41dbussinkor irl :P
20:48:50evandon koo
20:48:59evanDon, like the man's name
21:07:51dbussinkevan: hehe
21:07:58dbussinkevan: i can hear it sound funny already :P
21:10:27evanhttp://www.youtube.com/watch?v=6UOxFthTWRY
21:10:31evanthe sync is off for some reason
21:10:37evanbut the audio is there.
21:16:56dbussinkevan: https://dl.dropbox.com/u/8364158/danku.wav
21:17:28evansounds pretty much the same to me. :)
21:17:54dbussinkevan: nah, the u is pronounced pretty different :P
21:18:02dbussinkit's like the french tu
21:18:08evanmy ear is untrained :)
21:18:30dbussinkenglish doesn't really have that sound
21:18:41dbussinkyou people are pretty sounds poor :p
21:20:14dbussinkme meeting people at railsconf pretty much went like: "Hi, what's your name", me showing the batch, "ah how do you pronounce that?", me saying my name and getting a "eehm, that's impossible to pronounce look"
21:20:26dbussinkjust make it dj :P
21:21:51evani saw your full first name
21:21:57evanthough who knows how well I'm doing :)
21:22:06evans/saw/say/
21:23:00dbussinkif i say it in english style pronounciation people usually can reproduce it :P
21:23:33evan:P
21:23:51evanwell, for what it's worth, you're not the only one with that problem.
21:23:59evanpeople, americans, in the US, pronounce my name "even"
21:25:28dbussinkthat's just dumb :P
21:25:33dbussinklets call everyone john :P
21:26:21dbussinkbut i'm going to get some sleep
21:26:43dbussinkevan: haven't seen dkubb around btw, if you need more info later on just leave a message or comment on the issue
21:28:00evanwill do.
21:39:05evanbrixen: poke
22:40:57anieroevan: running with -Xprofile, do i need to do anything else but wait for it to finish?
22:41:19evanwait.
22:41:32evanit will print out the standand flat profiling table data
22:41:35evanon exit
22:41:38anierook cool
22:55:23anieroevan: bingo. 97.06 207.04 207.01 1869 110.76 110.78 ObjectSpace.find_object
22:55:30anierohttps://gist.github.com/3034962ba533d0fe19e3
23:01:27evanbingo bango bongo.
23:01:35anieroouch.
23:01:48anieroclose the ticket out, i suppose? it's a problem with fattr...
23:02:00anierohttp://github.com/evanphx/rubinius/issues/issue/381
23:02:16evanyeah, it is.
23:02:20evanwe could fork fattr and fix it
23:02:35anieroi could probably get commit access to fattr if i asked nicely
23:02:41evanok
23:02:55anieroi'll check with ara. any suggestions on what to do instead of _id2ref ?
23:03:03evansave it to a constant
23:03:07evanand read it back out
23:03:09evanor a class variable
23:03:11anieroor class var. yeah
23:03:13evananything thats a real reference
23:03:18anierook, cool. i'll look into it
23:03:23anierothanks for the help!
23:03:45evanno prob!
23:29:33kstephensvm/builtin/heap_dump.cpp:60: error: ignoring return value of 'ssize_t write(int, const void*, size_t)', declared with attribute warn_unused_result
23:30:40evanfucking a
23:30:54evanwarn_unused_result can suck it.
23:31:02evani'll fix it.
23:31:17evananything else? or just write?
23:31:56kstephensLemme see
23:31:58kstephens...
23:36:37kstephensdang you cant use (void) cast to stop -Werror
23:40:05kstephensfugly patch: http://gist.github.com/449270
23:41:21evani've fixed it
23:41:24evanabout to commit.
23:42:19boyscoutShow errors if write(2) has errors - 447e5cb - Evan Phoenix
23:50:31boyscoutCI: rubinius: 447e5cb successful: 3456 files, 13603 examples, 41167 expectations, 0 failures, 0 errors