Index

Show enters and exits. Hide enters and exits.

01:04:13boyscoutPut melbourne ext for rbx in a subdir. - 02c813d - Brian Ford
01:04:13boyscoutBuild external libs for parser universal on darwin. - 47cfec0 - Brian Ford
01:04:13boyscoutBuild melbourne for MRI. - ef46d5d - Brian Ford
01:06:46boyscoutCI: Build ef46d5d failed. http://ci.rubini.us/rubinius/builds/ef46d5d2db9019884a240d3d7df066c8a16e4026
01:08:30brixenthat's about as ugly a bt as you could get
01:26:10evanbut hey!
01:26:12evana backtrace!
01:26:17evanbetter than before
01:28:03brixenyep!
01:28:09slavaevan: did you sort out the issue where brixen's new compiler was slowing down in GC?
01:28:33evanno.
01:28:48evanfixed one thing that might help
01:28:58evanbut haven't focused strictly on that issue
01:33:48imajesstesla: ping
01:42:52brixenhrm, c-ng is partially working with melbourne under MRI
01:43:09brixenexcept that I'm getting Fixnums instead of Symbols
01:44:01slavaare you porting it to MRI for bootstrap purposes?
01:44:11brixenyeah
01:44:16slavacool
01:44:21slavaso its almost ready then
01:44:37brixenalthough, you know what I'm thinking
01:44:47brixenrbc files are platform independent
01:45:09brixenso, we should add a configure option for prebuilts just like we do for llvm
01:45:57evanbrixen: we've been blindly calling unshare everytime we use as_rstring()
01:46:05evanwhich blindly duplicates data_
01:46:09evaneven if it's not shared.
01:46:09evan:/
01:46:49brixen:(
01:47:02evanwhich is where the crash was (still not sure about it)
01:47:11evanbut it's SUCKING down performance too.
02:01:23brixenhm, why would I get a Fixnum back from rb_intern on MRI...
02:04:54evanyou shouldn't...
02:04:57evanyou get an ID back
02:04:59evannot a Symbol
02:05:04evanyou need to use SYM2ID() on it
02:05:07evanto get a Symbol
02:05:12brixenID2SYM
02:05:15evanID != Symbol on MRI
02:05:18evaner.
02:05:19evanyes.
02:05:21evanID2SYM
02:05:30brixenI just realized that
02:05:35brixenMRI is so confusing
02:05:37evanyeah, thats silly
02:05:43evanit's a legacy thing
02:05:49evanbecause the parser uses ID's internally
02:05:51evanas you know.
02:06:01brixenyeah
02:06:03Zoxcwhy is the ID mess there?
02:06:12evanZoxc: purely legacy
02:06:15evanit's never been cleaned up
02:06:20Zoxcin the first place? =P
02:06:26evanoh, in the first place
02:06:40evanit was purely the token the symbol table gave out
02:06:46evanruby didn't have symbols in the beginning
02:07:00evanat one point, matz added a quick parser hack
02:07:06evanwhich created
02:07:11evan:foo.class # => Fixnum
02:07:18evanin time, he changed that to be Symbol
02:07:26evanbut didn't go back and clean up the ID usage
02:07:43Zoxcah
02:07:47slavamatz has inflicted much pain on the world
02:08:11evanand much joy.
02:09:29brixenyeah, I forgive the pain
02:09:42brixenjust wish we'd started rewriting MRI in like 2000
02:10:57imajesthat would have been interesting
02:11:03imajesbut
02:11:05imajesit's OK
02:11:09imajesi'm just glad when we'll be done :)
02:11:30brixenheh
02:11:37boyscoutUnshare intellegently, try and fix a strange bug - 10e0af7 - Evan Phoenix
02:11:37boyscoutFix Regexp management bug - ee4d78c - Evan Phoenix
02:11:37boyscoutFix operand name so instruction decode is nicer - e8187c5 - Evan Phoenix
02:11:37boyscoutFix describe showing things multiple times - ab3f496 - Evan Phoenix
02:11:37boyscoutFix Dir["*/foo"] from picking up incorrect files - b1ca90b - Evan Phoenix
02:11:37boyscoutUpdate Rubygems to 1.3.5 - d44fed4 - Evan Phoenix
02:11:42brixenimajes: you'll be at RC?
02:12:07evanok, bbiab.
02:12:07brixenwoot!
02:12:13brixenevan: you updated rubygems!
02:12:15brixenyay
02:12:33imajesbrixen: not this time :(
02:12:38brixenimajes: :(
02:12:39imajesbrixen: stuck in argentina ;)
02:12:46brixenohh so stuck :P
02:12:49imajesheh
02:12:56brixenI'd like to be stuck in argentina
02:13:05imajesyeah, more time for me in the US this year is .. problematic
02:13:29brixendid you use up all your time chips?
02:14:58imajessomething like that
02:15:13imajesbrixen: trips via US INS usually end up with lots of questions
02:15:15imajesand bag searches
02:15:39brixenbastards
02:15:58brixenimajes: if you weren't such a kingpin and all :)
02:16:49imajesbrixen: riight
02:19:12boyscoutCI: Build d44fed4 failed. http://ci.rubini.us/rubinius/builds/d44fed4c2d47ede57ffbe95436a1622ae1af5be2
07:03:08boyscoutRevert of 47cfec0d61, Snow Leopard builds fail once you specify the arch options for just 32-bit platforms. - e269a86 - David Waite
07:03:35ddub(brixen tricked me into testing my commit bit)
07:04:42boyscoutCI: Build e269a86 failed. http://ci.rubini.us/rubinius/builds/e269a86f92e6bd5dddbb8517d0c17849633699a9
07:08:45dduboh no you didn't
07:09:36ddub:( couldn't be me
07:09:43ddubahh ok, wasn't
07:09:44ddubwhew
07:09:51brianmariohaha
07:09:51brixenyeah, that's not you
07:46:42ice7998388608
07:46:44ice799Sigh.
07:46:48ice799default stack size on my amd64 box.
07:46:51ice799gonna be here for everrr
07:56:05ddubwhats up ice799?
07:56:20ddub8MB isn't that off for a default stack size
07:56:43ddub(the kernel may only allocate/map that into the process a page at a time)
08:05:35ice799wrong channel, my bad.
08:05:43ice799i'm talking about thread stack size, though.
08:08:21ice799but, otherwise, ddub you are correct. the kernel maps memory into the process to extend the "main thread" stack when a page fault happens.
08:18:22boyscoutImported iseq and compiled_file to c-ng dir. - 2f02f9d - Brian Ford
08:18:22boyscoutFixed getting Symbols in melbourne under MRI. - 0cc2831 - Brian Ford
08:18:22boyscoutMore MRI bridging for c-ng. - 575dfad - Brian Ford
08:18:22boyscoutDon't use uninitialize line variable (thanks ddub). - 430eaf7 - Brian Ford
08:20:56boyscoutCI: Build 430eaf7 failed. http://ci.rubini.us/rubinius/builds/430eaf74d3dda168d0a941311a55bd43736bbb56
15:20:48ddubmorning
16:59:00brixenmorning
17:07:41steslaimajes: pong (you pinged last night)
17:08:19imajesstesla: yeah, about hoptoad/toadhopper i think
17:10:44evanmorning!
17:10:50evantries to start out the day upbeat
17:11:06imajesGOOD MORNING BIG BEAR!
17:13:22evanhehehe
17:15:32dbussinkevan: howdy how!
17:19:29rueEvening
17:20:05evanhi!
17:23:56Zoxcdoes rubinius still use a spaghetti stack?
17:24:00evanno
17:27:20steslaimajes: Ahh, still have questions?
17:27:29boyscoutCI: 430eaf7 success. 3003 files, 11767 examples, 35776 expectations, 0 failures, 0 errors
17:28:07rueWaait, how did it succeed now?
17:28:10imajesstesla: ah, no. i was starting to make an adapter for ramaze, but managed to get rack_hoptoad working instead
17:28:45steslaAha. Cool.
17:31:31imajesstesla: yeah, went with the simplest-possible-thing-that-works rule
17:33:01steslaimajes: Good way to do it. I ended up with Toadhopper becuase some of the other devs at EY said to check it out, and I was able to pretty quickly patch it to do what I wanted.
17:34:17imajesyeah, i got close with ramaze but really, it was more effort than i wanted already
18:04:28evanwow
18:04:30evanfor some reason
18:04:41evanthe password reset for the rubinius twitter account finally come through
18:04:42evanlike
18:04:47evan4 weeks after I requested it.
18:04:53brixenheh, crazy
18:04:56brixenbut yay!
18:04:58brixenfinally
18:05:24brixenwe all know twitter's got queue issues ;)
18:05:53evanheh
18:06:13brixendon't go posting all your tweets at once :)
18:06:26evanheh
18:07:08imajesah twitter.
18:07:15imajeswhere not everything is eventually consistent
18:09:29brixenimajes: pioneering a new tech: eventually asymptotically consistent with a constant factor
18:10:36brixenk-EAC (in honor of k-CFA)
18:12:49imajesheh
18:17:04evani'm rewriting most of our class variable support
18:17:07evanand i'm adding
18:17:16evanclass_variable_get_or_set(sym, val)
18:17:22evanto use in the
18:17:25evan@@blah ||= 3
18:17:26evancase.
18:20:34evannicely, now that the JIT is humming along
18:20:51evana monomorphic primitive is quite efficient
18:26:25evanwow, my C++ is getting... interesting.
18:27:02evanclass thing : public Parent { virtual bool ok_p(...) { ... } } match;
18:27:51evanC++ could really have used anonymous subclasses
18:30:41evansomething wierd I found out about yesterday
18:30:44evanhaskell has no backtraces
18:35:40brixenit does try pretty hard to tell you exactly why you fucked up though
19:31:19evanI found a nice syntax hole
19:31:22evanthat we can play with
19:31:28evan+@name = whatever
19:32:48brixenhmm
19:33:18evanI was actually thinking about
19:33:23evan+@@whatever
19:33:29evanfor more sane class variables
19:34:31brixensane how?
19:34:50evannot searching upwards
19:35:03evanso strictly shared between instances of that specific class
19:35:40brixenso like class instance vars with easy syntax from an instance?
19:36:10evanright
19:36:20brixenyeah, that'd be cool
20:42:44boyscoutRename rb_intern used in melbourne to avoid confusion. - ad69def - Brian Ford
20:42:44boyscoutAlways build melbourne with debug symbols. - c888a85 - Brian Ford
20:42:44boyscoutDon't segfault MRI when quarks go wild. - 233fc9b - Brian Ford
20:42:44boyscoutMelbourne must NOT use any MRI symbol machinery (any, ever, ever). - 2b0a4e9 - Brian Ford
20:42:44boyscoutName BlockPass body consistent with Iter, For. - 2a3ebd6 - Brian Ford
20:43:04brixenjj``: super is a keyword
20:43:10brixenit's handled in the compiler
20:43:45brixenheads to lunch
20:43:53brixenboyscout: be good while I'm away
20:45:44boyscoutCI: Build 2a3ebd6 failed. http://ci.rubini.us/rubinius/builds/2a3ebd6d42f1eab83323bcab5f447ac34971e988
20:55:12dbussinkboyscout is good at generating failures lately
20:55:31dbussinkevan: does it already help to identify issues?
20:58:24evanyep
21:05:01dbussinkevan: seems like capi is most often to blame
21:05:13evanin those cases, yeah.
21:05:24evani'm planning on trying to debug it more today
22:25:34evanARG
22:25:47evanspent 20 minutes debugging a "bug" which is really just a bug in rubygems
22:25:58evanbecause it transports Marshal'd objects
22:26:03evanand they changed the name of an ivar
22:31:21imajesevan: i beat you.
22:31:36imajesevan: i just spent an hour debugging why capistrano wasn't working, because textmate had me editing the WRONG deploy.rb
22:31:46evanha.
22:32:08imajesthat app needs to be fixed, and badly