Index

Show enters and exits. Hide enters and exits.

01:27:17toulmeanbrixen, evan, I'm trying to install rubinius 1.0.1 on apache infra
01:27:28toulmeanunfortunately it fails at that step:
01:27:29toulmeangcc -c -o readline.o -I. -fPIC -I/home/toulmean/rubinius-1.0.1/vm/capi/include -I/usr/include readline.c > /dev/null
01:27:55toulmeanbecause of those errors: readline.c:12:31: error: readline/readline.h: No such file or directory
01:27:56toulmeanreadline.c:13:30: error: readline/history.h: No such file or directory
01:28:04toulmeanI'll attach a full gist
01:30:03toulmeanhttp://gist.github.com/437162
01:36:25toulmeantrying to work around with using jruby -S rake, it sounds like it fails over melbourne as an optional component (mri_bridge something ?)
01:36:37toulmeananyway. Might just be for nothing.
01:37:57toulmeanhmf
01:38:03toulmeandifferent errors
01:38:05toulmeanhttp://gist.github.com/437170
01:40:49toulmeanI think I'm stuck for now. If you have insights, feel free to ping me!
01:49:43toulmeanit looks like when installing with mri you need this debian package: libreadline5-dev
01:49:46toulmeancorrect ?
01:54:04toulmeanok looks like I'll need to get those installed from what I read.
02:29:00brixentoulmean: doc/getting_started.txt has all the stuff you need to install on debian/ubuntu
03:57:49dkubbis there a way to transform a Proc into a sexp?
03:58:51dkubbeither directly or through a series of steps. the ideal being something like Proc#to_sexp, but I'll take anything at this point :)
04:02:33brixendkubb: no, there's not
04:02:52brixenthere are no sexp in rbx anymore
04:03:24dkubbbrixen: ahh ok, is there a way to access the AST at runtime?
04:03:38brixenyes, parse some source code :)
04:03:52brixenProc#to_sexp is ridiculous imo
04:04:10brixena consequence of not having the parser be a useful utility
04:04:32brixenlook in lib/compiler/compiler.rb for how to do it
04:04:38dkubbbrixen: I was thinking of porting an Ambition-like API to rbx and I was looking at how to approach it
04:05:05brixenyou can parse some code and get an AST
04:05:44dkubbcool. I just wanted to make sure it was possible before diving in :)
04:08:14brixendkubb: http://gist.github.com/437262
04:10:35brixendkubb: if you are just parsing, look at lib/melbourne.rb actually
04:14:20dkubbbrixen: thank you. I'll check this out and see how far I can push it
04:14:27brixencool
04:14:40brixenyou might already know, but melbourne wraps MRI parser
04:15:03brixenbut you get a Ruby AST back that the compiler can process, if you need that
04:15:32dkubbmy intention was to try to parse stuff like User.all.select { |user| user.id == 1 } into a DM::Query object, which would be transformed into "SELECT * FROM users WHERE id = 1" if executed with a database adapter
04:15:42brixenalso, you should check out lib/compiler/ast/transforms.rb
04:15:47dkubbI've done it before with MRI + ParseTree, just wanted to see if I could do it in rbx
04:16:18brixenyeah, that shouldn't be too hard
04:16:43dkubbit's mostly just for fun. no one uses it in production since it's a bit brittle. I still think it's beautiful when I look at it tho :)
04:17:11brixenyeah, AST can be brittle, mostly because MRI never made the AST part of Ruby
04:17:38brixenI'm not a big fan of stuff based on AST transforms b/c it doesn't fit well with the rest of Ruby
04:17:45brixenbut if it floats your boat... :)
05:29:31kstephensdkubb: might wanna look at the ambition gem.
05:30:45dkubbkstephens: yeah I have. I ported it to DataMapper a year ago
05:31:19dkubbkstephens: my approach uses the same as Ambition. I use ParseTree and get the sexp for the proc, and then build up a query as I process it
05:32:17dkubbkstephens: although I don't generate SQL directly like Ambition does. I build a DataMapper::Query object that can be used with any DM adapter
05:43:11dkubbkstephens: the challenge will be to figure out an approach that works without access to the Proc at runtime
06:10:55kstephensI was toying around with ParseTree Friday trying to make a generic Ruby macro processor: http://github.com/kstephens/ruby_unify
16:31:59evanmorning!
16:32:06brixenmorning!
16:32:22brixenhow was chi-ca-go?
16:32:29evanfun!
16:32:35brixennice!
16:33:07LeNsTR^__^
16:33:35evanhappy to be home though.
16:34:58brixengoing to try to get rbx to build on fedora 8
16:35:00brixenfor dblack
16:35:09evanok
16:35:18brixenbecause he's one of the few people I'll support on an os released in 2007!
16:35:38evanhehehe
16:57:46evanwelp, still working on the signal delivery issue
16:57:53evanhopefully have it fixed today.
17:04:47toulmeanevan, brixen: I now have rubinius installed on apache infra!
17:04:55evanwoo!
17:05:03brixentoulmean: yay
17:05:03evantoulmean: I saw your ticket about $LOADED_FEATURES
17:05:06evanwe'll have to look into that.
17:05:07toulmeanyes
17:05:17toulmeanthere is more after that, I'm a bit puzzled.
17:05:26toulmeananyways.
17:05:31toulmeanfirst things first.
17:05:59toulmeanevan: think $LOADED_FEATURES is a bad hack from buildr ? I could do it in a different way
17:06:11evana very bad hack
17:06:20evanany direct manipulation of $LOADED_FEATURES is a bad hack.
17:06:27toulmeanI can probably use eval with a regexp to catch what I want from the file
17:06:35evantoulmean: so, i'm looking at the ticket
17:06:37evanand confused
17:06:47evanit puts "java" in loaded features, then requires jruby
17:06:56evandoes jruby require java? is that why you do that?
17:07:00evanalso, what exception does it raise?
17:07:21evana LoadError for java, i expect
17:07:22toulmeanevan: it adds jruby to the LOADED_FEATURES, not java
17:07:23evanbut you don't say
17:07:27evanoh
17:07:29evanyour ticket is wrong.
17:07:33evanoh oh
17:07:34evanno
17:07:37evanit's me thats wrong.
17:07:40evanNEVERMIND
17:07:47evanwe should fix this though
17:07:47toulmeanevan: oh wait let me add formatting.
17:07:48evanbut it's a hack.
17:09:27toulmeanevan: the thing is it works with MRI
17:09:32toulmeanrubySpec for it ?
17:09:44evanyes, please.
17:09:48brixenthere should be a rubyspec for it
17:10:01toulmeanevan: I can file it. Helps ?
17:10:02brixenplease check the existing ones to see where the hole is
17:10:10toulmeansure
17:10:37brixenbasically, require 'jruby' should be a no-op because you put 'jruby' in $L_F, right?
17:11:12toulmeanbrixen: yes
17:11:17toulmeanfunny hack hey ?
17:12:10toulmeanthis is buildr requiring a file that would heavily depend on jruby, but making sure to not load it using LOADED_FEATURES, so it can grab a constant in the file and use it to compile java
17:12:13brixenwell, it's a lame hack inspired and supported by MRI's totally brain-dead code loading
17:12:27toulmean:)
17:12:36toulmeanit's ok if you don't want to do it
17:12:50brixenoh, we try to support the rest of it, this is just a bug
17:13:04toulmeana bug in MRI ?
17:13:05brixenbut I'm sure there is a better way for you to do it :)
17:13:16brixenyes, #require is a bug in MRI IMO
17:13:26brixenbut no, this is a failure of rbx to implement said bug
17:13:35toulmeanah ok
17:13:44toulmeanyour codebase is just too clean
17:13:47toulmeanthat's the problem
17:13:48toulmean:)
17:14:03brixenwe try to use OO abstractions, that's the problem
17:14:58kronos_vanoI'm trying to fix the ticket today, but: $LF << 'jruby'; require 'jruby' #=> false. $LF << 'pp', require 'pp' # => true. WTF? :(
17:15:44brixenkronos_vano: if you are working on a ticket, please indicate that
17:15:47toulmeanthat's nice
17:15:59toulmeanguys I'm looking up rubyspec
17:16:05brixensaves wasting time on something someone else is doing
17:16:11toulmeanlook at line 43 of load.rb
17:16:33brixenload.rb where?
17:16:42toulmeanerr
17:16:48toulmeancore/kernel/shared/load.rb
17:16:58toulmeansorry wrong file actually.
17:17:02toulmeanload != require
17:18:26toulmeanthe interesting stuff is more in require.rb, same folder line 330
17:18:35kronos_vanobrixen, I always indicate. here is my tag: "kronos". But i dunno how to fix it for now, so i didn't indicate
17:19:08brixenkronos_vano: well, if you are working on it, let us know
17:19:17kronos_vanosure
17:19:30brixentoulmean: yes, there are specs for not requiring the file if it's in $L_F
17:19:31toulmeanthanks all for working on it
17:19:42brixentoulmean: that's why I'm wondering why your case is different
17:19:52toulmeanbrixen: yes, none of them indicate a path without .rb
17:20:03brixenah, well there it is then :)
17:20:09brixenshould be an easy spec/fix
17:20:09toulmeanalso it's possible you try to bind to an actual file
17:20:32toulmeanmy case was a bit more complex, but I wanted to reproduce a minima:
17:20:48toulmeanfirst you had the $L_F << 'jruby'
17:21:00toulmeanthen require 'buildr/nailgun'
17:21:09toulmeanthen nailgun.rb does require 'jruby'
17:21:18toulmeansee, the path changed.
17:22:05brixentoulmean: show me how what you are doing is different than this http://gist.github.com/437974
17:22:21brixenbasically, write me a repro I can run, please :)
17:23:03evanshouldn't $L_F have jruby.rb?
17:23:07evannot just "jruby" ?
17:23:23brixendunno, i'm trying to follow toulmean's steps
17:23:33evanyeah, I think toulmean's steps are wrong.
17:23:38brixenme too :)
17:37:59toulmeanok will do
17:38:06toulmeanwill use mature language as well
17:40:11toulmeanbrixen: hmmm.
17:40:29kronos_vanohttp://gist.github.com/437991 This fix case when $L_F contain name with rb extension
17:41:54toulmeanbrixen: ok, I can't repro small scale, I'll give the full steps from my use case
17:42:08kronos_vanofeature_provided? method doesn't check name without '.rb' :/
17:48:07toulmeankronos_vano: updated the issue.
17:48:58kronos_vanotoulmean, can you test with http://gist.github.com/437991 ?
17:49:33toulmeanerr ?
17:49:43toulmeanok one sec.
17:52:12toulmeankronos_vano: still not.
17:52:20toulmeanbut then.
17:52:26toulmeando I need to run rake again ?
17:52:31toulmeanafter applying the patch ?
17:52:37brixenyes
17:52:45toulmeanrak'ing then
17:52:48toulmean:
17:55:09kronos_vanotoulmean, rake build
17:55:14toulmeankronos_vano: yes!
17:55:18toulmeanit works. kthx.
17:55:38brixentoulmean: so were you adding jruby.rb or jruby to the $L_F?
17:55:45toulmeanjruby
17:56:06toulmeanbrixen: can I ask roadmap questions ?
17:56:12brixensure
17:56:27toulmeanis there a 1.0.2 coming up ?
17:56:34evantoulmean: not atm, no.
17:56:42evanwe can do one if there is the need.
17:57:49toulmeanok no worries.
17:57:54evanwhy?
17:58:35toulmeanevan: I have more issues, and I needed to see if there was a timeline
17:58:41toulmeanto rush fixes in or not.
17:58:56evanstuff going into master right now is for 1.1
17:59:06evanwhich will come out in august.
18:01:32toulmeanok good deal.
18:02:35toulmeanjust wondering - you guys will never support rcov right ?
18:02:47toulmeanor the other way around - rcov will never support rubinius right ?
18:02:53evanwe could probably support the front end
18:02:57evanbut likely not the backend
18:03:18evanwe've got a better profiler than rcov anyway
18:03:19evan:)
18:03:45toulmeanyes. I commented it out using a check on RUBY_ENGINE
18:04:17brixentoulmean: I'll look at rcov
18:04:22toulmeanno need
18:04:48brixentoulmean: I have some profiler work to do for wycats to support ruby-prof style collapsable graphs
18:05:04toulmeanon Buildr, I can install all dependencies now (rbx -S rake setup)
18:05:12toulmeanI removed the need to sudo
18:05:35toulmeanthen rbx -S spec fails when it uses sh to call rbx to compile java files
18:05:44toulmeanit gives a status 1 without more explanations
18:06:14evantoulmean: sounds like the call to javac isn't working
18:06:29toulmeanyes but it works if I run the command directly
18:06:30evanmaybe some directories were created when you were using sudo
18:06:37evanand they're not root owned.
18:06:41evans/not/now/
18:06:44kronos_vanobrixen, Why Go? (i'm about gor)
18:06:54toulmeanevan: well same gems work well with mri
18:06:56toulmeanhmm
18:07:00toulmeanlet me check.
18:07:25brixenkronos_vano: why not? ;)
18:08:38toulmeanevan: no, it doesn't change much.
18:08:51evantoulmean: i'm not sure what you mean.
18:08:57toulmeanok.
18:09:11kronos_vanobrixen, :D. I prefer javascript
18:09:13brixenwell, blast, doesn't seem like I can get a fedora 8 system running
18:09:50brixenkronos_vano: heh, well get busy running javascript on rbx then :)
18:10:20toulmeanevan: http://gist.github.com/438044
18:10:53evantoulmean: no clue
18:11:29evantoulmean: you'll have to dig in to see whats different
18:11:36evanjavac should really be outputing something
19:09:04kronos_vanoPatch for #365 http://gist.github.com/438128
19:10:28brixenkronos_vano: that spec should not use 'should_non raise_error'
19:10:36brixener should_not
19:11:05kronos_vanoWhat they should use then?
19:11:14kronos_vano.should be_false ?
19:11:22brixenit "returns false when requiring a file that already exists in $LOADED_FEATURES"
19:11:25brixenyes
19:11:44brixenit "returns false when requiring a file with no extension that already exists in $LOADED_FEATURES"
19:12:22brixenkronos_vano: is this true for a path or only a filename?
19:13:12kronos_vanoI don't test with path. I'll try
19:13:22brixenalso, is this behavior changed if the file actually exists?
19:14:13kronos_vanobrixen, yes
19:14:26kronos_vano$LOADED_FEATURES << 'pp'; require 'pp' => true
19:14:38kronos_vanobut
19:14:46kronos_vano$LOADED_FEATURES << 'pp.rb'; require 'pp' => false
19:15:21brixensounds like me need more than one more spec then ;)
19:15:29brixener me == we :)
19:15:58kronos_vanobrixen, I really try to write another, but all of them passed before patch :(
19:17:11brixenthey all passed because we implement all that behavior
19:17:25brixenall passing =\=> all behavior is spec'd
19:18:36kronos_vanoright.
19:19:05kronos_vanoabout paths, did you mean that: http://gist.github.com/438144 ?
19:20:50brixenkronos_vano: no, I mean similar to existing specs at core/kerner/shared/require.rb:311-332
19:21:04brixenand those specs give you an example for how to write these btw
20:37:15evanman signals are a pain.
20:37:19evani should see how python handles them.
20:39:18slavawhat problem are you having?
20:39:37evanwell, how/where I should dispatch them mainly.
20:39:51slavaideally you'd be able to run ruby code from the signal handler itself
20:39:58evanfuck
20:39:59evanno.
20:40:09evanthats never happening.
20:40:10slavaclozure common lisp can do it
20:40:15evangood for them.
20:40:19slavanative threads without a GIL + lisp signal handlers
20:40:21evanit's so problematic.
20:40:31slavait means every instruction has to be a safepoint
20:40:34evanthe system is in ANY state.
20:40:37slavaright
20:43:32slavaevan: so the basic approach you use is to queue up signals and then run them in the main thread?
20:43:43evanyeah
20:43:50evanshould be easy.
20:44:01evani'm fighting with some aspects of it though.
20:44:11slavacan you interrupt infinite loops with ^C?
20:44:38evanyes.
20:45:54evani'm now getting this weird behavior where even though i've registered a handler
20:46:05evanthe process exits as though the default handler were hit.
20:46:22slavasounds like you forgot to reset the signal mas
20:46:24slavamask
20:46:32slavadoes this occur when you get the same signal twice in a row?
20:46:47evanperhaps yes.
20:46:56evanreset the mask to what?
20:47:18slavahow are you exiting from the handler? longjmp?
20:47:24evanreturn.
20:47:36slavaso you change the instruction pointer in the ucontext to point to some code, and then return?
20:47:41slavathat should work fine
20:47:41evanno no
20:47:53evanthe handler just records that a signal occured
20:47:55evannothing more.
20:48:14slavaok
21:05:51evanslava: the complicated part is that if the main thread is in a blocking operation
21:06:00slavaoh
21:06:02evanit needs to be woken up to handle the signals
21:06:07slavacan't you spin up a new thread to run handlers in?
21:06:21evani've considered it
21:06:27evanbut ^C still needs to be handled on the main thread.
21:07:14evanfor IO operations, I think i'm going to introduce a way to advertise a pipe fd that can be used to wake up select
21:07:32evanatm, a signal is sent with pthread_kill to get the syscall to return EINTR
21:14:55bakkdoorhi
21:15:04evanpython, it appears, doesn't wake up the main thread at all.
21:15:10evanit just waits until the main thread checks.
21:15:12evanbakkdoor: hi.
21:15:26ekiddHello! Rubinus is looking really great!
21:15:32evanthanks!
21:15:35ruePerhaps the channel concept could be worked in again
21:15:44ekiddIs there any documentation on setting up rubygems with Rubinius?
21:15:53evanekidd: it's builtin
21:15:55evanthere is nothing to do
21:16:05evanbin/rbx gem <commands>
21:16:10ekiddAh, nice!
21:17:49evanrue: there still has to be a mechanism to trigger a channel
21:17:52evanwhich is the level i'm at.
21:18:03slavaevan: you should make all io nonblocking :)
21:18:25evanthats an option, it just has it's own set of headaches.
21:18:36evanthats a bit of "now you have 2 problems" solution.
21:19:08slavaheh
21:19:18slavasignals just suck in general though
21:19:36evanthat they do.
21:23:28evanman, IO in a signal handler is such bad news.
21:33:55ekiddSweet! I'm running MongoDB's GridFS in Rubinius. It's definitely on the slow side, since the bson_ext gem doesn't play nicely with Rubinius and I had to drop back to the native Ruby gem.
21:34:04ekiddBut it's working.
21:34:27brixenekidd: the bson_ext gem probably uses #pack #unpack a lot, right?
21:35:01ekiddbrixen: The C version seems to also poke around inside objects. But yeah, I think the native version uses pack and unpack.
21:35:08ekiddI'm just amazed that everything runs. :-)
21:35:15brixenheh, cool
21:35:36ekiddI've worked on compiler projects before, and this is definitely the part of the development process where things get easier...
21:35:57brixenekidd: if you want me to look at some perf issues, give me a gist with all the steps I need to install and run
21:35:59ekidd...Oh, if I wasn't devoting every waking hour to a startup, I'd have such fun with this!
21:36:05brixenheh
21:36:33brixenekidd: you'd be very successful if your startup's product were an hour multiplier
21:36:40ekiddYou people just completely rock.
21:36:45brixenjust enter the day and how many hours you want to multiply
21:37:07brixenfor example, I'd like 6 hrs extra on 6/15/2010 :)
21:37:32ekiddYeah, I think there's a market for that. :-)
21:37:33parndtI'd like to go back to 2010-02-11
21:37:42slavaI'd like to go back to 1997
21:38:17brixenevan: so, I have built rbx on fedora 8
21:38:21evanok
21:38:38brixenevan: but to do so, I had to build llvm from source
21:38:49ekiddbrixen: Here's the code I'm running: https://gist.github.com/a240685f8f650615cbd1
21:38:51brixentrying to use the prebuilt, I got this while linking http://gist.github.com/438188
21:39:27ekiddYou need mongodb installed, but that's pretty easy--it's pretty straightforward to set up.
21:39:34brixenevan: I'm testing the llvm pkg on ubuntu 10.04
21:39:43brixenekidd: sweet, I'll take a look
21:39:44evanbakkdoor: yeah, we'll need to build a prebuilt specificly for it probably
21:39:52evansince it's a pretty old version of libc
21:40:14brixenevan: the problem is, how do we distinguish it?
21:40:27brixenit gets the same host string as a much more recent linux
21:40:34bakkdoorevan: hm?
21:40:44brixenbakkdoor: he meant me :)
21:40:45evanbakkdoor: eh?
21:40:46ekiddbrixen: Note that MongoDB GridFS is still a bit piggish in MRI, too--it doesn't run nearly as fast as the pure C GridFS client. Call it 5 files/second.
21:40:51evanoh
21:40:52evansorry.
21:40:53bakkdoorbrixen: ah ok :)
21:40:54evanyes, brixen.
21:40:56bakkdoorno problem ;)
21:40:57brixenheh
21:41:00ekiddRubinius is rather, um, _slower_. :-)
21:41:04evanbrixen: we don't.
21:41:18evanbrixen: prebuilt just isn't support on older linux for now.
21:41:30ekiddOTOH, unlike MRI 1.8.7, it doesn't actually _fail_ when trying to run this code, AFAICT.
21:41:35evanwe can add a --build option to configure to take a triple
21:41:38evanso they can use
21:41:42ekiddMore compatible than the real thing!
21:41:50evan./configure --build x86_64-linux-fedora8
21:41:52evanor something
21:41:56evanand we can a prebuilt for that.
21:41:57brixenevan: well, if the prebuilt I built on f8 works with newer ubuntu, could we just use that one?
21:42:04bakkdoorbrixen: btw, how is poison coming along? I'm anxiously waiting to see some examples on additional languages running on rubinius :)
21:42:05evanit won't.
21:42:06brixenfor the i686-linux
21:42:12evanit will not work.
21:42:18brixenevan: ok
21:42:40evanekidd: hah
21:42:40evan:)
21:42:50brixenbakkdoor: I pushed what I have up to now, I need to restart with a better parser
21:43:12bakkdoorbrixen: alright. i can wait ;)
21:43:20brixenbakkdoor: I'll make "'hello' print" execute
21:43:42bakkdoorbrixen: ok
21:43:55brixenbakkdoor: evan added custom callsite logic, so rbx is basically rockin' :)
21:44:40brixenevan: so, f8, how do we want to instruct ppl to build?
21:45:08brixenshould we re-enable telling configure to use source directly? add a --no-prebuilt switch?
21:46:39evanno
21:46:44bakkdoorbrixen: yeah i'm pretty sure I'll port fancy to rubinius some time soon. lots of things in common and I think supporting the platform is generally a good idea ;)
21:46:49evanwe've got that already anyway
21:47:05evanthey'll just have to build LLVM themselves
21:47:21evanand use --llvm-path
21:47:23evanto point to it
21:49:48brixenhmm, just linked ok on ubuntu with f8 prebuilt
21:49:54brixenand the specs are running
21:50:28evanyou mean for 10.1?
21:50:41evani'm just uncomfortable building LLVM with a much older version of libc
21:50:47evanand using it in ubuntu 10
21:52:03brixenyeah, me too
21:52:14brixenI just don't want to force ppl to build llvm :/
21:52:16brixenit sucks
21:52:45evanbest way then is to upload that prebuilt as a special name
21:52:50evanand add --prebuilt-name
21:52:53brixenwhy not --prebuilt fedora8 => llvm-1686-linux-fedora8.tar.bz2
21:52:56brixenyeah
21:53:00brixensomething like that
21:53:03evansure
21:53:06brixenk
22:00:29cremesekidd: i saw that you said bson_ext has problems with rubinius so you fell back to the native bson driver
22:00:57cremesi did the early work to get mongo_ext working with rubinius; is the C extension compiling or failing?
22:01:06ekiddcremes: Yes, it claims that it can't serialize Rubinius' time object.
22:01:16cremescan you gist me the error?
22:04:06ekiddcremes: http://gist.github.com/438390
22:04:23evan alias_method :write_nonblock, :write
22:04:24evan:/
22:04:25evanFAIL.,
22:04:54cremesekidd: thanks; i'll look at it this weekend
22:05:21ekiddcremes: I just installed the gems a few minutes ago, so they're all the latest versions as of now.
22:05:22evanit probably uses RTIME or something
22:06:01cremesperhaps... it didn't use to so this would be an incompatibility that was recently introduced
22:14:30evancremes: http://github.com/mongodb/mongo-ruby-driver/blob/master/ext/cbson/cbson.c#L426-429
22:14:32evanthats the problem.
22:14:40evanit's detecting Time by digging into the object raw
22:57:00brixenevan: could you move the f8 prebuilts in my home dir to assets so I can test pls
22:57:06evansure
22:57:10brixenthanks
22:57:40evandone
22:57:46brixenexcellente
23:38:39cremesevan: that code is new; i'll try to figure out an alternative this weekend when i can focus on it
23:38:48cremesthanks for doing the legwork to pinpoint it
23:40:26evancremes: no prob.
23:40:36evancremes: using rb_kind_of() would be better.
23:41:34cremesyeah
23:42:01cremesit doesn't look like it is accessing anything private; it's calling to_f on it...
23:42:50evanyep
23:43:01evanjust a bad shape test
23:43:26cremes"bad shape"?
23:43:42evan"This thing looks like a Time object"
23:43:46cremesah
23:44:04evani used "shape" as a generic term for indicating the type a thing is
23:44:14cremesgot ya
23:44:28cremesi'll add this to my weekend queue
23:45:05evanawesome
23:45:06evanthanks!
23:45:37cremesnp; hopefully brixen will get to his pack/unpack optimization soon so that mongo is usable under rbx... (hint, hint)
23:49:15evanwe're going to have to rewrite them entirely likely
23:51:19brixencremes: I can take a hint! :)
23:52:33brixencremes: if you could just convince everyone to use the latest mac os x release, we'd spend far less time on these build issues
23:52:44brixenand more time doing "real work" (tm) :)
23:53:01evanbrixen: which are you working on? fedora?
23:53:08brixenyeah
23:53:11evanok
23:53:18evannext time, push back.
23:53:23brixenadding --prebuilt-name
23:53:31brixenwell, we need a tad more flexibility here
23:53:34evanwe're not going to spend out time building vms for all linux distros
23:53:41brixenyeah
23:53:45evani'd rather have the user of that distro be helping out
23:53:52brixenthis way, they can build and we can use it
23:53:53brixensure
23:53:56evannot us sinking our own time.
23:54:06brixenyep
23:54:44brixenlsb_release on linux is a sweet thing, btw
23:54:55evanah cool.
23:54:56brixenI didn't know about it till dwaite told me about it
23:55:23brixenlsb_release -irs appended to the config.guess output is what I'm shootingfor
23:55:52brixenie llvm-i686-linux-gnu-fedora-8 or ubuntu-10.04 etc
23:56:04evanok
23:56:18cremesbrixen: i will give a sermon to my mac faithful about the wonders of snow leopard so as to save the project some time :)
23:56:21evanare you going to be trying both?
23:56:28brixencremes: sweet
23:56:39brixenevan: I want to do specific with a fall-back to generic
23:56:51evank
23:56:58brixenso it will check #{config.guess}-#{lsb_release} then config.guess
23:57:56brixencrazy http://dailydeadbirds.com/