Index

Show enters and exits. Hide enters and exits.

00:19:57evanfinally!
00:20:24evanwe were reporting some really wrong backtraces when type errors happened in FFI
00:56:36boyscoutCast glob's arg to a String - c36eafb - Evan Phoenix
00:56:36boyscoutUntag Dir.glob failure - 11d8afb - Evan Phoenix
00:56:36boyscoutHandle TypeErrors during FFI properly - fbbe050 - Evan Phoenix
00:56:36boyscoutShow the exception that shouldn't have happened - c1c2b0d - Evan Phoenix
00:56:36boyscoutClamp mode in File[.#]chmod - 41fb3be - Evan Phoenix
00:56:37boyscoutRemove File[.*]chmod failing tags - 9f22617 - Evan Phoenix
00:57:05evana little spit polish :D
00:58:31rueAdded GCC register syntax?
00:58:51evanin one place, yes
00:59:07evanit should only be being run on i386
00:59:21evanthough, i just realized it might try to be used on x86-64
00:59:23evanwhich is probably wrong
01:03:02evanrue: ?
01:03:28boyscoutCI: 9f22617 success. 2709 files, 10761 examples, 33729 expectations, 0 failures, 0 errors
01:06:27rueNo, it should not conflict
01:08:23rueSuppose cannot rely on it
01:09:43evannope, can't
01:09:55evanbut no reason to not try and use it on certain platforms
01:10:00evangcc will just ignore it if it doesn't work out
01:11:22rueMm, alternative would be more cumbersome
01:15:44evanhuh?
01:17:04rueIf you wanted to use %esi in a more portable manner
01:20:29evanah.
01:20:37evanmore #ifdef's would be needed likely
01:20:48evanasm needs a literal string there
01:26:43rueInterpreter is running faster currently, with JIT compiled in but not enabled. Full CI (-fs) is ~78s JITted, ~73s interpreted
01:29:35rueIt is overall slower, though, unless more specs have been enabled. I seem to have reported ~70s last time.
01:31:28rueToodles
08:14:57boyscoutHandle lambda's with 1 or fewer args specially - db8f4b3 - Evan Phoenix
08:14:57boyscoutChange spec to accurantly reflect test - 38bcf6f - Evan Phoenix
08:14:57boyscoutRemove Proc tags - 643948e - Evan Phoenix
08:18:25boyscoutCI: 643948e success. 2709 files, 10764 examples, 33778 expectations, 0 failures, 0 errors
13:13:35nemerle_mhello sirs, i'm trying to build current git/master of rubinius ( on debian ), and i've encountered some errors:
13:13:35nemerle_mwhile building extensions: " Unable to resolve type 'size_t' " " Unable to send 'isatty' on 'Platform::POSIX' (Module) "
13:14:00nemerle_mare there some common solutions i might try ?
13:36:27badboynemerle_m: which gcc version?^^
13:58:21nemerle_m4.4.0 :)
13:59:13nemerle_mhacked around it by adding size_t to ffi.rb
14:01:59nemerle_mnow String::to_double test fails
14:28:41badboyhm...i get another error with gcc 4.4
14:29:06badboyhttp://github.com/evanphx/rubinius/issues#issue/13
14:29:53nemerle_mstrange, i had no problems like that
14:32:27nemerle_moh, and the problem with to_double is caused by strict aliasing issues within the external library (libgdtoa)
14:34:25badboyi will compile it again..
14:35:25badboyagain...
14:35:39badboyinvalid conversion...
14:42:04nemerle_madding -fno-strict-aliasing to compiler flags in libgdtoa 'solves' that vm/test/runner failure
14:43:38nemerle_mbadboy_: your problem might be related to the version of bison you're using ?
14:44:48badboy2.4.1, and you?
14:45:56nemerle_msame
14:46:02badboyhm...
14:51:19nemerle_mthe error there should be easy to fix, just try adding 'const' specifier before 'char *tmp;'
14:59:18badboynemerle_m: i tried that and it come up withanother error
15:22:14cremessegmentation fault when running specs against latest trunk (from about 30 minutes ago) using LLVM: http://pastie.org/513761
15:22:25cremescan i provide any additional details?
15:24:52cremeswhen i run with the "-fs" switch, it dies on Float#abs - returns the absolute value
18:35:11boyscoutExclude undef'd methods from Kernel#methods - 88661bd - Evan Phoenix
18:35:11boyscoutDisallow adding methods to Numeric subclasses - deef443 - Evan Phoenix
18:37:36boyscoutCI: deef443 success. 2709 files, 10764 examples, 33778 expectations, 0 failures, 0 errors
20:13:39seydarsalutations, children
20:44:44boyscoutFix Thread::value - d7da228 - Evan Phoenix
20:44:44boyscoutUpdate fails tags for Thread - ac6cdaf - Evan Phoenix
20:52:11boyscoutCI: ac6cdaf success. 2709 files, 10767 examples, 33781 expectations, 0 failures, 0 errors
20:59:51boyscoutAdded benchmarks for reading/writing locals vs ivars. - cff7c69 - Brian Ford
20:59:51boyscoutSome initial app benchmarks. - 79f14f5 - Brian Ford
20:59:51boyscoutAdded basic graphing of benchmark results. - c00fa28 - Brian Ford
20:59:51boyscoutTemporarily commiting benchmark graph. - c68283c - Brian Ford
20:59:59brixenevan: see benchmark/results/web/index.html
21:00:31brixenif you tell me where we will run the nightly benchmarks, I can set up the baseline and a template to work with rubini.us
21:00:47brixenI need to finish the baseline and delta graphs
21:00:52brixenbut first, lunch
21:02:16boyscoutCI: c68283c success. 2709 files, 10767 examples, 33781 expectations, 0 failures, 0 errors
21:46:00evanbrixen: wow! the JIT makes that big a difference?
21:46:27evanthe JIT kicks rdoc's ass!
21:46:33evantests locally
21:48:50brixenwell, actually not I think
21:48:58brixenthe jit was bus erroring on that
21:49:10brixenbut it wasn't getting reported to the runner oddly
21:49:30evan:/
21:49:31evandamn.
21:49:36brixenyeah
21:49:40evanso how should I be running these?
21:49:56brixenrake -q bench:dir DIR=benchmark/app
21:50:08brixenadd VM='rbx blah'
21:50:14brixenor VM=ruby
21:50:17brixenor whatever
21:51:00brixenbut you can just run the command in benchmark/app/bm_rdoc.rb manually with the jit on
21:51:13brixenjust haven't gotten to tracking that down yet
21:52:57brixenit's possible that we don't set $? correctly if a bus error/segfault occurs
21:53:11brixensince that's what benchmark/utils/monitor.rb checks
21:53:29dbussinkbrixen: hmm, most other benchmark scripts can also be run standalone
21:53:50brixendbussink: how do you mean?
21:54:31dbussinkbrixen: just by running ruby benchmark/rubinius/bm_string.rb
21:54:48brixenyeah, well it doesn't give you yaml output either
21:54:52brixenanyway, so can these
21:55:00brixengo read benchmark/utils/README
21:55:45evani can't get the rdoc one to run
21:56:02evandid you mean to be using the rdoc-2 directory there?
21:56:26brixenrdoc-2.4.3 yes
21:56:37evanit doesn't seem to support --threads or --output
21:56:47evanand you don't use -I to make sure it's used
21:56:48evanjust -S
21:56:54evanwhich will likely use rdoc 1, the installed one
21:56:59brixendo you have rdoc installed?
21:57:04evanrubinius does
21:57:17brixenle'me run it here
21:57:29evanbin/rbx -S rdoc -h
21:57:34evanruns in a fresh checkout
21:57:40evanbecause we ship with rdoc 1, like MRI
21:59:33evan../../bin/rbx -Irdoc-2.4.3/lib rdoc-2.4.3/bin/rdoc --all --force-update --threads=1 --output scratch rdoc-2.4.3/lib
21:59:36evanthat worked
21:59:40evanif i'm in benchmark/app
21:59:52evanthough, i had to change darkfish.rb
21:59:59evanto not try include rdoc 2.4 from rubygems
22:00:59brixenworked fine for me http://gist.github.com/130896
22:01:32brixenthe benches run in their dir, yes
22:01:38brixenI didn't need to change any files
22:01:47evanwell
22:01:50evani don't see how
22:01:53evanthats going to be using rdoc 1.0
22:02:08evandid you update rdoc in rbx/lib
22:02:08brixenwhat is this rdoc 1.0
22:02:09evan?
22:02:16brixenI just installed the gem
22:02:22evanoh
22:02:23evansee
22:02:28evanwe HAVE rdoc without the gem.
22:02:30evanthats why.
22:02:46brixenwe probably should have the current version if we are going to have it
22:02:54evanls lib/rdoc/
22:02:57brixenI didn't know we did
22:03:00evanin the repo root
22:03:14brixeneww
22:03:16brixenwhy do we?
22:03:21evanbecause MRI did
22:03:27evanand we pulled it over with the rest of stdlib
22:03:36brixenhm
22:04:05brixenwell, we can do it differently
22:04:09brixenhow do you want it to work?
22:04:14evanif you've got rdoc's source there
22:04:15evanwhy not use it?
22:04:22evani don't get why it's there if you require the gem to be installed
22:04:42brixenwell, for one, it doesn't create .rbc files
22:04:49brixenwhich I could not get to exclude properly
22:04:56brixenreally, do it however you'd lke
22:05:08brixenjust use the bench template
22:05:13evanno no this is fine
22:05:14brixenso we can get the data
22:05:20evani'm just trying ot make sense of it to run it here
22:05:30brixenrdoc is a pile of steaming ...
22:05:36brixenthis worked for me
22:05:48evanok, so to run them, you need a set of gems install
22:05:54evanrdoc, rails, what else?
22:06:12evanbtw, JIT just crashed here too
22:06:15evanrunning rdoc
22:06:30brixenrdoc and rails are the only gems atm
22:06:35evanok
22:06:52evanwe should probably ditch rdoc in lib/
22:06:56evanso we don't have to maintain it
22:07:10brixenanyway, like I said, if there are .rbc files, rdoc will process them and that skews the results
22:07:18brixenso, I only use the rdoc dir as data
22:07:25evanah!
22:07:29evanso just coincidence
22:07:36evancould have used any source there
22:07:38evani gotcha
22:08:06brixenI definitely think we should remove lib/rdoc
22:08:16brixenor at least update it to current rdoc version
22:08:35evanlets ditch it
22:08:39evanno reason for us to maintain it
22:09:02evanif we need to 'ship' with it, we'll ship with the rdoc gem pre-installed
22:09:46brixenthat would make more sense
22:10:04evanok! nuking now.
22:13:07boyscoutNuke old version of rdoc - 3d591d1 - Evan Phoenix
22:15:46slavahi evan
22:15:50evanhi slava
22:15:59evanbrixen: seems like the generation step takes the longest
22:16:03evannot the parsing
22:16:06evanwhich i find surprising
22:17:13boyscoutCI: 3d591d1 success. 2709 files, 10767 examples, 33781 expectations, 0 failures, 0 errors
22:18:41brixenyeah, parsing is pretty fast
23:25:20evanwow
23:25:25evanno really, WOW.
23:25:38evanline 355 of lib/rdoc/parser/ruby.rb
23:27:03brixenwow indeed
23:27:33evanthats a doozy