Index

Show enters and exits. Hide enters and exits.

06:08:50brixenstepheneb: if you configure with --prefix, you need to run rake install
06:09:08brixenyou can't expect rbx to work if it doesn't have the files it needs
06:09:17brixenthat's why you get the error message
06:09:25brixenthe other part I'm looking into now
06:09:27stephenebbrixen: rake install also generates errors when run
06:09:36brixenyes, that's the part I'm looking at
06:12:18stephenebAs part of debugging and trying to understand the build and install process I ..
06:12:37stephenebcopied over the lib/ and runtime/ directories to the prefix dir
06:13:06stephenebThat allowed rake install to get further but it then stopped with this error:
06:13:22stephenebwhile running the first invoke of gem_bootstrap
06:13:32stepheneband "Compiling library files for install..."
06:13:42stepheneb./lib/etc.rb:4: uninitialized constant Etc::FFI (NameError)
06:13:59brixenwhat ruby are you using?
06:14:06brixenand what env stuff do you have set?
06:14:14brixeneg RUBYLIB, RUBYOPT, etc
06:14:30brixenyou should just run rake install after running configure --prefix
06:14:36brixenyou should not be copying stuff around
06:14:43brixenthat's what rake install is for
06:14:48brixenso let's just get that working...
06:15:25stepheneb$RUBYOPT => rubygems
06:15:37stephenebRUBYLIB => nil
06:16:35brixenwhat version of ruby?
06:16:53stephenebI copied some of the directories just to give me a better handle on what was used when
06:17:10stephenebruby 1.8.6 (2008-08-11 patchlevel 287) [universal-darwin9.0]
06:17:10brixenthat just confuses the issue tho
06:17:19brixenwe need rake install to just work
06:18:32stephenebSure, that sounds fine -- I was following the instructions in getting_started.txt
06:18:45brixenok, let's get it clarified then...
06:18:52brixenwhere does it say copy stuff?
06:19:05brixenwe need our directions to be correct too :)
06:19:27stephenebI'm not making myself clear yet ;-)
06:20:42stephenebThe only reason I copied stuff was as a way to get more understanding of the build and run process -- it was just a way to get more data
06:20:59brixensure, I got it
06:21:10brixenjust really want this to work and I've run it with no issues
06:21:19stephenebI'm tying too slow also ... I meant to say after: "instructions in getting_started.tx"
06:21:22brixenso I'm having trouble understanding the Etc::FFI issue
06:21:24stephenebthat I just ran rake
06:22:13stephenebI just ran rake clean and then rake install and am getting the same error
06:22:25stephenebrunning this command:
06:22:41stephenebgenerates this error:
06:22:42stepheneb./lib/etc.rb:4: uninitialized constant Etc::FFI (NameError)
06:23:02brixenhm, le'me fix rake
06:23:11stephenebopps, irc ate the first line => /System/Library/Frameworks/Ruby.framework/Versions/1.8/usr/bin/ruby -Ilib lib/bin/compile.rb -V -T default lib
06:23:29stephenebwas the command being executed in the rake Install process that generated the error
06:23:35brixenok
06:23:49brixenso the system ruby is trying to load stuff from rbx's lib dir
06:23:52brixenhmm hmmm
06:24:01brixeneff, why does this not happen for me...
06:24:31brixenok, let's try something...
06:24:33stephenebThis is the end of the console log: http://gist.github.com/244815
06:25:06stephenebI'm on Mac OS X 10.5.8
06:25:45brixenme too
06:25:55brixensame ruby version
06:26:03brixenbasically the vanilla os x ruby here
06:26:09stephenebyup
06:26:19stephenebI also have macruby installed in /usr/local
06:26:20brixenoh dur
06:26:26brixenunset rubyopt
06:26:28stephenebbut that shouldn't cause a problem
06:26:34stephenebok will try that
06:26:52brixenyou must be loading rbx's rubygems
06:27:19brixenbut I'm going to make this work without using -Ilib I think...
06:27:26stephenebyes, rake install is now getting much further
06:27:31brixenyeah, frick
06:27:34brixenhaha
06:27:41brixenRUBYOPT has bit us soooo many times
06:27:53stephenebme too, I should have thought of it
06:28:06brixenwell, we can make this more robust
06:28:19stephenebnow the prefix dir is filled up with rbx goodness
06:28:28brixenhehe
06:28:36brixenI'll set rubyopt to repro this...
06:29:43stephenebit seems that setting a custom prefix dir should only matter when running install
06:30:08brixenhow do you mean?
06:30:14stephenebI was thrown off when I ran rake and bin/mspec threw an error
06:30:23brixenwell, yes, I'm fixing rake
06:30:38brixenbut you HAVE to install if you configure rbx to have a different dir than the build dir
06:30:47brixenbecause the VM has to know where to load the kernel
06:30:49stephenebit nice that rubinius will just *run* form it's source dir where I've got it checke dout
06:30:55brixenit's not on exe like MMRI
06:30:59brixener MRI
06:31:09brixenit's a small exe (the VM) and the rest it loads
06:31:18brixenjust like the jvm with its classpath var
06:32:05stephenebOr perhaps running the standard rake when using a custom prefix should run install before running mspec
06:32:29stephenebbut that also doesn't make sense
06:32:37stephenebinstall should be a separate step
06:32:37brixenwell, I can't make the default rake run rake install
06:32:50brixenbut I can make the standard (default) rake work...
06:33:03stephenebahh ... that would be good
06:33:04brixenby just setting RBX_RUNTIME and RBX_LIB directly
06:34:04stephenebdoes rbx use the -S param the same way jruby does?
06:34:19stephenebif I run: jruby -S gem list
06:34:19brixenwell, it uses -S the way it's documented in MRI
06:34:27brixenit searches some internal dirs and then your PATH
06:34:45brixenrbx -S gem should run rbx's bundled rubygems
06:35:06brixenrbx -S gem install foo; rbx -S foo ... should work too
06:35:11brixenwe bundle rake and rdoc
06:36:40stephenebyup, that all works for me so far:
06:37:27stephenebrunning: /Users/stephen/dev/ruby/builds/rubinius/bin/rbx -S gem list
06:37:41stephenebresults in just the installed gems
06:37:56brixencool
06:38:52stephenebI've been playing around running a few simple benchmarks on all the ruby VMs I can run
06:39:07brixenfun
06:39:31stepheneband while rbx is pretty fast on somethings it's quite slow running rexml
06:39:52brixenhmm yeah, not surprising
06:40:04brixenwe haven't worked on optimizing string much
06:40:07brixenin the jit
06:40:10brixenbut soon! :)
06:40:52stephenebrbx is about 4x slower than MRI 1.8.6 and macruby about 1x slower
06:41:26stephenebin a test using rexml to parse a large xml file
06:42:00brixeninteresting
06:42:18brixenif you have a simple bench, gist it or something
06:42:30stephenebok
06:42:31brixenevan is working on a nice new benchmarking framework
06:42:40brixenthat puts the benches in tiers
06:43:02brixenoh, actually make it a ticket :)
06:43:21brixenwe'll consider being slower than 1.8.6|7 a bug of sorts :)
06:44:55stephenebbrixen: see: http://gist.github.com/244822
06:47:31stephenebThe xml test comes simulates some work that we do -- but I'll bet the slowdown could be shown much more simply
06:47:48stephenebwith probably a simple String#gsub type of test
06:48:01stephenebI know rexml uses lots of regexes
06:48:43stephenebI'm pretty sure the extreme slowdown for macruby is dues to issues with String
06:50:13brixenhmm ok
06:50:36brixenoh so rbx is faster than macruby...
06:51:00stephenebyes
06:51:22stephenebbenchmark displays strangely on macruby
06:51:22brixenI think we'll be able to get rbx faster on the tak bench...
06:51:33stephenebbut rbx is 4x slower, macruby 10x slower
06:51:50brixenyeah, above you said 1x slower heh
06:51:54brixenI was confused
06:51:59stepheneboops
06:52:12stephenebmacruby is about 92s
06:52:18stephenebrbx about 40s
06:52:23stephenebmri about 9s
06:52:38brixenyeah, that sys time output is weird on macruby huh
06:52:51brixenwhy is the real less than total?
06:52:53stepheneband jruby with --server about 6s
06:53:04stephenebI don't know??
06:58:19stephenebI just tried running: $ time macruby xml_benchmark_rexml.rb ...
06:58:47stepheneband the shell 'time' command also showed this strange behavior
06:59:06brixenhmm
06:59:07stephenebreal 1m53.495s, user 3m8.339s
06:59:30stephenebthat doesn't make much sense and doesn't have anything to do with running the ruby vm
06:59:57stephenebbut macruby is definitely takeing about 90s
07:00:09stephenebwhich is pretty dismal
07:00:14stephenebfor that test
07:00:27stephenebpretty fast on the tak test though
07:01:56brixenyeah, they do some special stuff with floats iirc
07:02:30brixenbut I think we'll be able to get the jit to understand code like that well enough to optimize
07:02:37stephenebevidently! so does maglev it seems
07:02:41brixenwithout messing with IEEE float values
07:02:52brixenI don't think maglev is doing the same thing, but dunno
07:03:54stephenebI don't know what maglev is doing either but aprox 30x MRI speed is pretty good
07:04:20stephenebthey just seemed to come out of cloaking mode so I added them to the mix
07:04:50stephenebthey are pretty raw however, the rexml test didn't even run
07:05:24stephenebthanks for the help! It will be easy now to keep up with rbx
07:06:23stephenebwould you like a performance ticket for the rexml test or wait for something simpler (maybe String.gsub might show the effect)?
07:08:57brixenyeah, please do file a ticket
07:14:26stephenebbrixen: here you go: http://github.com/evanphx/rubinius/issues/#issue/102
07:16:31brixenawesome, thanks
07:48:18boyscoutMake specs run from rake regardless of install config. - 0c1fb8e - Brian Ford
07:51:19boyscoutCI: 0c1fb8e success. 3005 files, 11472 examples, 35615 expectations, 0 failures, 0 errors
07:59:44stephenebcool, rake (and mspec) is working now with a custom --prefix
08:04:57brixencool
08:14:02stephenebbut the bin/mspec run hung after 425 tests (well I stopped it after it stopped there for about 5m)
08:14:20stephenebhere's end of the console log: http://gist.github.com/244848
08:14:47brixenthere is a random spin that is happening
08:15:03brixenwe haven't tracked it down yet, but there's already a ticket for it
08:15:12brixenrun it again and it probably won't happen
08:15:38brixenbut if you can repro it reliably, add a note to #93
08:18:05stephenebtried again and it completed fine => 3005 files, 11469 examples, 35608 expectations, 0 failures, 0 errors
08:23:16boyscoutFix Kernel#__send__ - d26d8a4 - Evan Phoenix
08:23:16boyscoutChange spec for Kernel#send - 78116c3 - Evan Phoenix
08:23:16boyscoutFix Process.wait and Process.wait2 - f2f1fb5 - Evan Phoenix
08:23:16boyscoutCleanup Syslog - 5fc6d8e - Evan Phoenix
08:23:16boyscoutInclude sys/types.h because some extentions expect that - 863a9ec - Evan Phoenix
08:23:17boyscoutAdd Gem.user_home to the path to look for gems - 87539ea - Evan Phoenix
08:23:19boyscoutAdd rbx specific default_exec_format - 88ecae7 - Evan Phoenix
08:23:21boyscoutCleanup doc - a2867ab - Evan Phoenix
08:23:23boyscoutDisambiguate lines output - 1efeda5 - Evan Phoenix
08:23:25boyscoutSet std::terminate to provide better error message - 4ff7d5e - Evan Phoenix
08:23:27boyscoutImprove flag and directory usage in building - 87981fe - Evan Phoenix
08:23:29boyscoutAdd -c support - 9287afc - Evan Phoenix
08:27:36boyscoutCI: 9287afc success. 3005 files, 11472 examples, 35615 expectations, 0 failures, 0 errors
08:34:10brixenbooyah
09:02:30rueGood stuff, I think the Gem.user_home should fix a ticket
22:16:52evanwebpokes around the freenode web UI
22:17:02evanla lal al
22:24:55rueHow did that work out?