Index

Show enters and exits. Hide enters and exits.

00:22:25brixen4:22!
00:22:48brixenI'm always a few minutes late
00:26:56evanheh
00:30:35brixenevan: bytecode.rb at 150 and 300: why does one push_const and one find_const?
00:30:45evanlets see...
00:30:47brixenfor :Proc
00:31:09evanbug.
00:31:13brixenk
00:31:14evanin the one at 150
00:31:22evanneed to be sure we get ::Proc
00:31:30brixenyeah, that's what I was thinkin
00:31:30evannot any Proc
00:31:35brixenok, thanks!
00:31:57evantrying to boot rails 3.0 is shaking out some nice bugs
00:32:41brixensweet!
00:32:53brixenhow far along are you?
00:34:46evangetting there.
00:34:56brixencool beans
00:35:03evan75% of the time is making sense of whait rails means to be doing
00:36:10brixenyeah
00:36:40brixenare you running the rails tests or just booting an app?
00:36:49evantrying to boot.
00:36:55brixencool
00:37:03evanand finding out that someone on the rails team just found out about delegation!!
00:37:08evanso they need to use it solve every problem!
00:38:58brixenwhen all you have is delegation, everything is a minion
00:40:33brixenit's the latent desire to be aristocracy from our collective unconscious
00:43:20brixenwhy is BakerGC::free_objects called in ~ObjectMemory?
00:43:36brixenI thought free_objects was obsoleted?
01:01:18evanbrixen: it is, thats just legacy.
01:03:00brixenok, it's legacy that's causing me segfaults
01:03:13brixenhttp://img.skitch.com/20091105-r93fer15udfc8jctpgdywqrene.jpg
01:09:16boyscoutRename some insn parameters so CompiledMethod#decode works. - 38909cd - Brian Ford
01:09:16boyscoutMore loop compiler transform specs. - 1e04cbf - Brian Ford
01:09:16boyscoutHandle loop(&block) in c-ng. - d0c15ee - Brian Ford
01:09:16boyscoutEnsure block args find ::Proc. - 3f0395d - Brian Ford
01:16:42boyscoutCI: Build 3f0395d failed. http://ci.rubini.us/rubinius/builds/3f0395de6365587d56649e1736eeb8740cc739cc
01:18:21rueAww
01:31:29brixeneffen f
01:31:58brixenwell, full compiling CI runs with kernel and lib compiled with c-ng under rbx
01:32:28brixenbut something isn't marshaling right running c-ng under mri
02:48:34boyscoutMake Tuple a real class for MRI bridge. - 7524602 - Brian Ford
02:48:34boyscoutYou have to run ./configure before running rake when compiling Rubinius for the first time - 11923e8 - Markus Prinz
02:50:05boyscoutCI: Build 11923e8 failed. http://ci.rubini.us/rubinius/builds/11923e8ebe2db2793065bdabbc5bd3442316e807
02:58:50brixenboyscout: ur gonna end up in negative merit badge territory soon
03:05:42brianmariolol
03:31:15ddubthinks demerit badges would be cool
07:07:49brianmariowhat's the best way to represent a function with double-pointer in FFI?
07:08:11brianmarioer, a function with a double-pointer parameter
07:08:16slavavoid*
07:08:27slavawait, you mean double* or foo**?
07:08:27brianmarioI mean, in ruby
07:08:40slavaat the ABI level all pointers are the same
07:08:45brianmariohm
07:09:00brianmarioso here's what's in C
07:09:00brianmarioyajl_gen_get_buf(yajl_gen hand, const unsigned char ** buf, unsigned int * len)
07:09:14brianmariohere's what I've been trying in ruby
07:09:14brianmarioattach_function :yajl_gen_get_buf, [:pointer, :pointer, :pointer], :int
07:09:53brianmarioI also tried
07:09:58brianmarioattach_function :yajl_gen_get_buf, [:pointer, :string, :int], :int
07:10:02brianmariowith no luck either
07:13:19justin-georgeHmm. I just ran rake clean and rubinius deleted itself.
08:30:43rue:string should have worked
08:30:55ruejustin-george: Altogether?
08:31:20justin-georgerue: the entire top-level directory (the one with .git in it)
08:31:28justin-georgequite amusing :)
08:33:58rueThat is weird...reproducible?
08:35:01justin-georgelet me try again
08:35:14justin-georgewhen I renamed the directory to foobar it *didn't* delete itself
08:36:04justin-georgehmm...
08:36:18justin-georgeah, rake uninstall
08:36:21justin-georgenot rake clean :)
08:37:01justin-georgeperhaps that's desired behavior
08:38:32justin-georgeindeed, nevermind, my bad
20:36:45dgtizedand then just use 1 for the write's in quit_handler?
21:26:41tarcieridbussink: yeah Rammstein is better in german too
21:28:17Zoxcbets the german versions will be less cool if he learns more german...
21:30:02dbussinkZoxc: they will only increase in coolness ;)
21:30:26dbussinkit's also about the feeling the german pronounciation encompasses :P
21:31:05ZoxcI think german sounds a little childish =P
21:37:52dbussinkevan: do you think that getting rid of warnings that are shown with shorten-64-to-32 is useful?
21:38:58ZoxcScandinavian languages seems pretty suited to balads. See http://www.youtube.com/watch?v=yvWwyEwwhVI vs. http://www.youtube.com/watch?v=XzcORqLFsZw :)
22:23:34evandbussink: they should all be evaluated
22:23:36evanyes
22:35:09evanyippy!
22:35:16evangot rails 3 to show the application info panel
22:35:26evansoooo many little bugs.
22:35:29evanlike, rails 3 does
22:35:36evans = regexp.source
22:35:43evans.gsub!(/.../, "")
22:35:52evanand guess what we were doing in .source
22:36:43brixenum...
22:36:51brixenthe suspense is killing me
22:37:26evanreturning @source
22:37:29evanand not duping it.
22:37:32evanbefore we return it.
22:37:35brixeneeks
22:37:49brixenthat would confuse things
22:37:50evanwhich actually didn't matter much for the actual regexp
22:37:56evanbecause onig internalizes the pattern
22:38:12evanbut rails then created a new regexp using the 1st one
22:38:14evanand that uses source
23:33:57brixenok, bin/rbx -v runs after compiling kernel with c-ng running under MRI
23:34:01brixengonna run specs now
23:34:05brixen<sweaty palms>
23:34:16DefilerOoh; moment (minute) of truth!
23:34:35evan:D
23:35:06brixenrunning!
23:35:13brixenand running and running
23:35:24brixenevan: do you have local changes to lib/compiler-ng ?
23:35:48brixen(still running)
23:36:03brixenbtw, by running, I mean bin/mspec ci -T --melbourne --ast -B full
23:36:29brixenso, running while compiling under c-ng, including compiling lib
23:36:41brixenoh for focks sake get_handle!
23:36:48evannope
23:36:55evanget_handle giving you troubles?
23:36:57brixenevan: ok, i'm gonna rename that dir
23:37:03evank
23:37:04brixenoh, just the same old abort
23:37:11evani'm about to push a bunch of commits
23:37:14evan11 to be exact
23:37:16brixenthat I get about every 1 out of 2 commits
23:37:17brixenok
23:37:18evangets rbx running on rails 3
23:37:20brixensweet
23:37:23brixenpush! :)
23:37:33evani had to update and get your changes from last night
23:37:34brixener, 1 out of ever 2 spec runs I mean
23:37:37evanrebuilding and testing now.
23:37:40brixenk
23:37:51boyscoutVerify if RUBY_ENGINE is set to rbx in melbourne.rb, as MRI 1.9 now defines RUBY_ENGINE - aac7c52 - David Waite
23:37:53evanrails 3 uses 1.9's named captures on Regexp if they're available
23:38:07evanbrixen: ug.
23:38:13evanyou better pull that change from ddub in.
23:38:18brixenyeah
23:38:20evananyway,
23:38:29evanwe have the same API as 1.9 for named captures now
23:38:31brixenI asked him to push, I figured he'd check the channel
23:38:35evank
23:38:35brixensweet
23:38:46evanwe nearly did before
23:38:51evanit's just 2 extra methods on Regexp
23:38:56evanotherwise, you use MatchData#[]
23:40:42boyscoutCI: aac7c52 success. 3003 files, 11767 examples, 35776 expectations, 0 failures, 0 errors
23:44:35brixenoh man, I've got some funny commit msgs coming
23:45:02evanheh
23:45:20slavahi evan
23:45:29slavanew byte compiler for rbx, and new gc for factor, exciting times in language-land
23:46:19evan:)
23:46:21evanhows the gc coming?
23:47:57slavaits merged but I'm working the bugs out
23:48:33evanwhats new about it?
23:49:46slavamark compact for old generation and code heap
23:49:54slavareplacing copying for old generation and mark sweep (no compact) for code heap
23:50:47slavaevan: code heap compaction is not that important, its mostly there so that saved images are a bit smaller on disk
23:50:58slavaevan: how does llvm-jit handle reclaiming unused compiled code blocks btw?
23:51:18evanbadly.
23:51:46evanbut because we only compile things that are hot
23:51:51evanwe put little pressure on that.
23:51:54slavaI'm also using 4-bit tags now; so all built-in types fit in the tag, and one is for library-defined classes
23:52:06slavasimplifies method dispatch and inline cache machine code sequences a bit
23:52:09evanwhy not just go for a full 8?
23:52:19slavaand 256-byte align all objects? :)
23:52:24evanhell yeah
23:52:39slavanah, 16 byte alignment is optimal for SSE operations
23:52:54slavait means I can use aligned load/store instructions to access vector data in byte arrays, which speeds things up
23:53:23evancool.
23:54:03slavawhat's the llvm-jit issue?
23:54:13evanwith?
23:55:02slavacode gc
23:55:19evanoh
23:55:28evanit's just not been working on much
23:55:32evanso it doesn't try very hard
23:55:58evanit does simple attempts to reuse regions
23:56:01evanbut no moving