Index

Show enters and exits. Hide enters and exits.

08:34:04dbussinkmorning!
08:36:10danlucraftmorning
08:36:41slavahey
08:39:28dbussinkevan: you there?
08:41:03slavahe's not usually around on sundays
08:45:36dbussinkslava: i did get a hold of him for what is this morning over there
13:29:40tarcierisup slava
13:29:42tarcieriif you're still awake
13:29:56tarcierinifty post re: local mutable variables
13:30:00tarcieriwish I could have those :/
14:22:18slavatarcieri: hehe
14:22:27slavatarcieri: we hardly ever use them
14:22:49slavatarcieri: mutable data types, on the other hand...
14:26:08tarcieriheh
14:26:39tarcierislava: being completely immutable in all cases is annoying
14:26:41tarcieri:(
14:26:46tarcieriErlang dogma
14:27:20tarcierithey're against destructive assignment because it's "mutable"
14:27:38tarcieripsst, if you can bind anything your local bindings are "mutable" :O
17:30:28DefilerHrm.. is there an example I can look at somewhere of calling C++ code in a CAPI extension?
17:31:05Defileror do I just need to do extern "C" VALUE foo?
18:08:41brixenDefiler: not sure what you are asking but melbourne is a C/C++ hybrid-ish sort of thing
19:36:26dbussinkDefiler: you can take a look at rice
19:37:08dbussinkevan: ping?
19:37:13evansup
19:37:36dbussinkevan: do you remember i talked about an issue in the datamapper specs that only occured after a while running specs?
19:37:54dbussinkevan: found that it's also something jit related, the specs run fine with -Xint
19:38:09evanah ok.
19:38:13dbussinkevan: i even got an abort yesterday, so that might be helpful too
19:38:24evansure
19:38:25evanyeah
19:39:52dbussinkevan: i think it was actually also on some code referencing a module
19:40:00evanooh
19:40:01evanok
19:40:04dbussinkevan: dunno if that is a coincidence or not than :)
19:40:13evanprobably not actually :D
19:40:16dbussinkevan: with the issue i saw with the profiler
19:40:17dbussinkhehe :)
19:40:21evanthere are few coincindences in the JIT
19:40:29dbussinklet me see if i can manage to reproduce it
19:41:20evancool.
19:43:04dbussinkevan: good thing that the specs at least run good with -Xint
19:43:12dbussinkevan: about 4x times slower than mri
19:43:22evanyeah
19:43:36dbussinkevan: actually, a friend of mine wrote a ruby only png library
19:43:56dbussinkit's pretty heavy on fixnums and pack, dunno if it's interesting to profile / benchmark a bit with it
19:44:02evanabsolutely!
19:44:04dbussinkhttp://github.com/wvanbergen/chunky_png
19:44:05evanget it
19:44:26dbussinkhe got saving a png 12% slower compared to rmagick which is a native extension
19:44:34dbussinkpretty nice for pure ruby
19:46:05evani'll try it
19:47:03dbussinkrmagick doesn't work with rubinius though
20:04:42dbussinkevan: did you see this one: http://github.com/doudou/rubinius/commits/master ?
20:05:07evanhe sent me some messages
20:05:07evanyes.
20:06:03dbussinkdid you see whether there are rubyspecs for this?
20:20:29brixendbussink: that's a good question
20:20:46brixenwe should probably start requiring either rubyspecs or tags removed on these patches
20:20:58brixenie either rubyspecs exist and the patch removes tags
20:21:07brixenor the patch includes new rubyspecs
20:21:10dbussinkbrixen: seems like a good approach yeah
21:12:40dbussinkevan: ok, i have a crash in gdb now
21:12:52dbussinkit's an USZ actually
21:13:44dbussinkevan: it's happening in the compiler
22:01:58evandbussink: oh! thats good.
22:01:59evanticket!
22:02:13dbussinkevan: what kind of information would you want?
22:02:30dbussinkevan: it takes a while to run before the repro occurs
22:02:41evanany info
22:02:43evanas much as you have.
22:03:38dbussinkevan: well, if i should add some specific gdb voodoo, please let me know :)
22:08:04dbussinkevan: ah, i have a much simpler case that fails pretty fast, makes it easier for you to reproduce it i guess :)
22:20:15dbussinkargh, github 404's
22:25:35dbussinkevan: http://github.com/evanphx/rubinius/issues/issue/171
22:25:46dbussinkevan: that also contains a way i can reliably reproduce it here
22:28:18dbussinkbut i'm going to get some sleep
22:28:32dbussinkevan: if you need more info, i'd be happy to get it for you