Index

Show enters and exits. Hide enters and exits.

00:02:48warren_s leaves the room.
00:11:23twbray leaves the room.
00:17:36nari leaves the room.
00:26:15malumalu leaves the room.
00:29:47twbray enters the room.
00:32:13twbray leaves the room.
00:40:23blowmage leaves the room.
00:47:42rudebwoy leaves the room.
00:47:47rudebwoy enters the room.
00:50:01joearnol_ enters the room.
00:51:11joearnold leaves the room.
00:51:41benny enters the room.
00:54:11lopex leaves the room.
00:56:37bitsweat enters the room.
01:09:55joearnol_ leaves the room.
01:10:05joearnold enters the room.
01:11:11nari enters the room.
01:11:13joearnol_ enters the room.
01:12:47joearno__ enters the room.
01:14:07joearno__ leaves the room.
01:14:21joearno__ enters the room.
01:15:30joearnol_ leaves the room.
01:15:32joearn___ enters the room.
01:15:32joearno__ leaves the room.
01:17:41joearnold leaves the room.
01:25:43boyscoutAdd a rudimentary Backtrace#to_mri for compatibility. - 099a089 - Eero Saynatkari
01:25:43boyscoutFix Method#call to pass block again. - 08cb015 - Eero Saynatkari
01:26:51rueWow. http://axiombox.com/rubx/
01:28:17boyscoutCI: 08cb015 success. 1438 files, 7105 examples, 23402 expectations, 0 failures, 0 errors
01:29:01botanicus leaves the room.
01:40:15bitsweat leaves the room.
01:56:26ezmob leaves the room.
02:02:55enebo leaves the room.
02:07:37enebo enters the room.
02:12:02jashmenn leaves the room.
02:12:44asap18 leaves the room.
02:24:05rudebwoy leaves the room.
02:25:20cored enters the room.
02:28:04joearn___ leaves the room.
02:28:34joearnold enters the room.
02:30:37enebo leaves the room.
02:32:42rudebwoy enters the room.
02:36:49ryanlowe enters the room.
02:37:51binary42 leaves the room.
02:46:50VVSiz_ enters the room.
02:53:37VVSiz leaves the room.
02:54:30joearnold leaves the room.
03:22:38tarcieriw00t, you guys documented rbc
03:22:43tarcieristudies
03:23:26evanwe did?
03:23:30tarcieriyeah
03:23:33tarcierivm/rbc_files.txt
03:23:36tarcierierr, doc/vm
03:23:59evani wonder if it's up to date...
03:24:01evanlet me check.
03:24:04evanhm, no, it's not.
03:24:09evanbut i've got something better for ya
03:24:18tarcierinice
03:24:35tarcieriwants to copy your approach as much as possible :)
03:24:41evan:D
03:25:38evankernel/compiler/compiled_file.rb
03:25:45tarcieriok
03:25:51tarcierievan: there's a huge problem if I store nested classes/modules in the attributes tuple of Erlang's equivalent of a .rbc
03:25:56evanit's a ruby implementation of the code to create and load .rbc files
03:26:03tarcieristuff in that tuple gets loaded into the Erlang "code server"
03:26:12evanthey're even simplier than doc/rbc_files.txt makes it seem
03:26:15tarcieriso I end up with massive code duplication
03:26:21evani simplied them a while back
03:26:29evanthey used to use binary encoding of numbers
03:26:33evani switched to ascii
03:26:40outerim leaves the room.
03:26:44evanso .rbc files can easily be edited and debugged.
03:27:01tarcieriyou just compare mtime of a .rb vs a .rbc and decide whether or not to recompile?
03:27:23evanatm
03:27:32evanyes
03:27:41tarcieriok
03:28:11evanwe also have a compiler version number, so if the compiler changes, .rbc files are discarded
03:28:20evanbut we don't have the compiler version number hooked up atm
03:28:26evanand we might not hook it up again
03:28:32evanit's really complicated to get right.
03:28:39tarcieriheh, fun
03:28:48tarcierididn't even think of that
03:30:24tarcieriI'm presently using Erlang's eval and it sucks balls performance-wise
03:30:26evandon't bother
03:30:36tarcieriI suspect you're dealing with a similar problem but have addressed it a lot better
03:30:40evanthe compiler doesn't change in production
03:30:43evanso it was only for dev
03:30:55tarcieriok
03:31:03evanand we just have the rake tasks remove .rbc files if the compiler changes
03:31:28tarcieriyeah I always rake distclean after I pull, but I pull pretty infrequently
03:31:50tarcieriI pulled after you merged stackless into master but ran into some problem... guess I'll try again
03:32:02tarcierierr
03:32:04tarcieristackfull
03:32:04tarcieriheh
03:32:07evanyeah
03:32:17jashmenn enters the room.
03:32:19evanif distclean doesn't fix it, let me know.
03:32:29tarcierithis was like... awhile ago
03:32:39tarcieriprobably something that was broken at the time which has since been fixed
03:32:41tarcieritries again
03:32:45evanwe did run into a problem where some .rbc files weren't being removed
03:33:03tarcieritrying to avoid recloning because he's lazy
03:33:10evanyou shouldn't need to reclone
03:33:15tarcieriI just pulled and now I'm trying to build again
03:33:21evank
03:33:23tarcieriI'll let you know
03:33:25tarcierigoes for more wine
03:33:26evanhows reia these days?
03:33:32evansome people asked me about it recently
03:34:46cored leaves the room.
03:34:57tarcieriI have it autoloading the 3 file standard library now... it's got directory, file, and socket support now, woo!
03:35:15tarcieriexcept it's horribly slow and loading that stuff, thanks to the horribly gimpy eval-based approach I'm using
03:35:22tarcieriErlang's eval is a pretty crazy hack
03:35:34slavahow does it work? just compile the code right there and run it?
03:35:55tarcierislava: it uses eval to execute the toplevel code, and any class or module declarations are compiled when they're declared
03:36:24tarcieriI want to move to where everything is compiled ahead of time and "wired up" when it's declared
03:36:28slavaok
03:36:52tarcierievan: sh: line 0: cd: vm/assembler/libudis86.a: Not a directory
03:36:54evantarcieri: rubinius compiles all the methods and such
03:36:59tarcieriI believe that's the error I got before
03:36:59evantarcieri: umm.
03:37:00evancrap.
03:37:10evanlibc didn't add a rake task I guess
03:37:12tarcieriit was definitely a "Not a directory" error
03:37:12evanone sec.
03:37:36tarcierithat was executing distclean
03:37:39evank.
03:37:41evanoh
03:37:44evanum
03:37:49evanjust run rake now.
03:37:55tarcieriok
03:38:19tarcierievan: you compile when you eval too, right?
03:38:26tarcieriI don't know why Erlang didn't just do that
03:38:49evaneval("a = 1")
03:38:51evanstarts the compiler, yes.
03:39:00evanbut nothing uses eval in kernel
03:39:07evanto add methods, they're compiled up front
03:39:14tarcieriyeah, I'm talking about for irb/string eval
03:39:16evanand put into the literals tuple of the method for the class body
03:39:41tarcierihere's one of the gems from Erlang's eval:
03:39:42tarcierihttp://gist.github.com/79022
03:40:25evanyikes.
03:41:01tarcieriarity sensitivity rearing its ugly head
03:42:48slavagross
03:42:52tarcierievan: so when you eval, you create a CompiledMethod, call it, then it's GC'd after that?
03:42:52slavawe use macros to deal with this in factor
03:43:21drbrainthat's basically how call_cfunc works in MRI
03:43:21twbray enters the room.
03:43:23evantarcieri: yep
03:43:32tarcierievan: I should totally just do that
03:43:42evantarcieri: though, thought of another idea on the plane down here
03:43:48evanif the eval is for something simple
03:43:54evanlike eval("a = 1")
03:44:05evanrather than going through the trouble and creating a CompiledMethod, etc, etc.
03:44:14evaninterpreter it
03:44:16evanall in ruby
03:44:19evanusing a simple AST walker
03:44:30evani've got the beginnings of one
03:44:31slavathat's what sbcl does
03:44:32tarcierimetacircular ruby evaluation or something?
03:44:35evanthat we use for Rubinius.asm
03:44:38evanwhen ou do
03:44:41evanRubinius.asm do
03:44:43slavaif the form passed in to eval is from a smal lrestricted set it interpetes it
03:44:45evan blah
03:44:45slavainstead of compiling it
03:44:45evanend
03:44:54evanit runs the code in the block inside an AST walker
03:45:01evanslava: right
03:45:05evanfor trivial evals
03:45:07tarcierievan: having a Ruby DSL for outputting assembly is fucking awesome, BTW
03:45:09evanit should be fine
03:45:14evan:D
03:45:15tarcieriheadius seems to have done something similar with BiteScript
03:45:19evanyep
03:45:22evanhe uses instance_eval
03:45:29evanwhich I really dislike
03:45:39tarcieriheh
03:45:42evanthats why our DSL is built with methods against a generator object
03:45:44headiusyou can use it either way
03:45:44tarcieriwants to add instance_eval to Reia :)
03:45:53evanwhich makes it strictly not a DSL, but close enough.
03:46:00headiusduby and compiler2 use it as a normal api
03:46:04evanah nice
03:46:14tarcierievan: build seems to have worked, btw
03:46:17headiusthe instance_eval way is only for making certain sections look like an assembly source file
03:46:24tarcierievan: it just ran 752 tests and they all passed!
03:46:30evanyippy!
03:46:35tarcierirunning some more now I guess
03:46:37evanheadius: yeah
03:47:03evanthe fact that instance variables suddenly work differently in an instance_eval irks me.
03:47:23tarcieriheadius: fwiw I like instance_eval-based DSLs, even if they're slow :)
03:47:34headiusyeah, if you're maintaining any state you can just have the block pass a builder in and it won't be instance_eval'ed anymore
03:48:02tarcieriafaict in Reia instance_eval is the only way a block/lambda can modify hidden state/instance variables
03:48:20tarcieribecause you're asking the block/lambda to be evaluated in the context of the receiver
03:48:28tarcieriwithout it the receiver is trying to modify the caller's state
03:48:30tarcieriwhich is no no no!
03:48:35tarcieriat least in Erlang
03:48:51evanhehe
03:50:03slavaman, I had a really cool ARM assembler DSL back in the day
03:50:06slavait was so slick
03:50:19tarcierinice
03:50:43tarcierievan: have you ever considered assing more DSLs to Rubinius for different types of bytecode?
03:50:59evanwhat kinds of bytecode?
03:51:01tarcierilike say... one for x86, or CUDA, or whatever
03:51:10evanand whats assing?
03:51:12tarcieriLLVM, for that matter
03:51:13tarcieriheh
03:51:19tarcieriI... don't know
03:51:21tarcierisome Ruby code?
03:51:21tarcieriheh
03:51:52evani guess where would the generater code be used?
03:52:06evanin the case of x86 and llvm, could be run "normally"
03:52:38tarcieriwell, you have a Ruby compiler written in Ruby... is that what you're already doing?
03:52:54slavathe jit won't be in ruby though will it?
03:52:56evanfor the rubinius bytecode, yes.
03:53:02evanatm, the jit isn't written in ruby
03:53:08tarcieriok
03:54:04slavahave I pimped my PowerPC asm dsl beofre?
03:54:18tarcierievan: I suppose it would be better as part of an FFI-based gem
03:54:25tarcierithen it'd be portable
03:55:09tarcieriI think it'd be pretty cool to have a Ruby DSL describing CUDA bytecode
03:55:22tarcieribut I suppose you could just talk to the CUDA libraries over FFI
03:55:41binary42 enters the room.
03:55:41moconnor enters the room.
03:56:12tarcieriFinished in 128.923688 seconds
03:56:13tarcieri1438 files, 7104 examples, 23400 expectations, 0 failures, 0 errors
03:56:14tarcieriyay!
03:56:16evanwhats CUDA?
03:56:22slavaGPU shiz
03:56:32tarcierievan: nVidia's language for executing code on a GPU
03:56:38evanOH WORD
03:56:47evanvertex shading as method dispatch!
03:56:55tarcierirofl what?
03:57:20tarcierihaha
03:57:27tarcieriusing a GPU to accelerate the virtual machine, eh?
03:57:29slavaevan: http://gitweb.factorcode.org/gitweb.cgi?p=factor/.git;a=blob_plain;f=extra/spheres/spheres.factor; hb=HEAD
03:58:00slavathat's glsl but cuda is similar
03:59:37tarcierione of my coworkers is using CUDA to accelerate an atmospheric model
04:00:25evanslava: cute!
04:00:33tarcierilike, I used to work in climate modelling, and one of my coworkers still does (and is using CUDA)
04:01:01tarcieriwhat if scientists (or whoever) could write most of their code in Ruby but describe their hotspots in a DSL that compiles to x86, CUDA, whatever
04:01:01evanbtw
04:01:06evanFX is way more badass not in the US
04:01:11evanno dubbing
04:01:20evanR rated nudity
04:01:23evanthe works.
04:01:31tarcierihaha
04:01:36slavatarcieri: or if Ruby was fast enough so that you don't need the x86 DSL :)
04:02:12tarcierislava: heh, keep reaching for that rainbow
04:02:15evanslava: but how do you get it fast enough without reducing to machine code?
04:02:23evandamn chickens and eggs!
04:03:12tarcierislava: the model I worked with was written in Fortran and C but had huge performance bottlenecks due to how it worked
04:03:19evanif we have the ability to save machine code
04:03:23evancould bootstrap like factor :)
04:03:43tarcierislava: it alternated between a CPU-bound period and an I/O bound period... its CPU usage looked like a sine wave, and the I/O bound period was proportional to the number of nodes you had
04:03:48slavawell you'd still have a x86 DSL in the compiler
04:04:23ruetarcieri: Did you read the SecondLife MQ evaluation thingy?
04:04:50wmoxam enters the room.
04:05:17evanslava: how many highlevel and lowlevel operations does your SSA sets have?
04:06:46slavalow level has 116
04:06:58slavahigh level has 22
04:07:59evanwhats are the higlevel ones?
04:08:04evani can imagine what the lowlevel are
04:08:07evanlots of memory movement
04:08:10evancasting, etc.
04:08:37tarcierirue: nope
04:08:53slavahigh level ones are calls, stack manipulation, branches, loops, ffi stuff
04:09:28evanok
04:09:39slavaif you have the source handy, they're defined in basis/compiler/tree/tree.factor for HLSSA and basis/compiler/cfg/instructions/instructions.factor for LLSSA
04:10:07tarcierirue: a coworker and I went through message queue hell... we've been on RabbitMQ about a year now
04:10:08tarcierirue: but we don't need a disk log like Twitter does
04:10:24tarcierirue: looks like Twitter seriously evaluated RabbitMQ but it wasn't suitable for their needs due to their whole system being stateful
04:10:40tarcierirue: one of the RabbitMQ developers replied on my blog defending them
04:11:01evanslava: i do, ok.
04:11:19tarcierirue: still doesn't explain Starling *facepalm*
04:11:33jashmenn leaves the room.
04:11:55tarcierirue: the author of Starling (who I guess has since been ousted from Twitter) replied with... lies, accusing me of spreading misinformation
04:12:00imajes leaves the room.
04:14:43jashmenn enters the room.
04:15:44tarcierifwiw, Ezra told me pretty much the exact same thing the RabbitMQ developer did, days before he responded. Ezra is a sharp guy :)
04:16:44tarcierinamely that the RabbitMQ disk log is still a WIP
04:28:14qbproger leaves the room.
04:31:33outerim enters the room.
04:44:24twbray leaves the room.
04:51:54yipstar leaves the room.
05:13:09imajes enters the room.
05:15:56slavatarcieri: http://gist.github.com/79022
05:15:59slavathis is just a perfect demonstration of why erlang sucks :)
05:31:00shoe leaves the room.
05:33:00kronos_vano enters the room.
05:33:21asap18 enters the room.
05:36:30twbray enters the room.
05:37:15headius leaves the room.
05:51:17shoe enters the room.
05:52:45mdalessio leaves the room.
06:00:42tarcierislava: rofl
06:00:56wmoxam leaves the room.
06:00:58tarcierislava: yeah I'm not a fan of how Erlang handles arity
06:01:08tarcierislava: Reia has since moved away from that
06:01:32tarcieriit's causing some problems but affords me a lot more flexibility
06:06:13twbray leaves the room.
06:06:21binary42 leaves the room.
06:10:05twbray enters the room.
06:12:37asap18 leaves the room.
06:17:16outerim leaves the room.
06:25:08mediogre enters the room.
06:34:16headius enters the room.
06:42:10ennui enters the room.
06:43:06ennui leaves the room.
06:43:08jashmenn leaves the room.
06:46:25mediogre leaves the room.
07:10:52kronos_vano leaves the room.
07:11:16headius leaves the room.
07:24:12twbray leaves the room.
07:28:43joachimm enters the room.
07:41:07botanicus enters the room.
07:44:17moconnor leaves the room.
08:16:29krawek leaves the room.
08:22:09joachimm leaves the room.
08:30:27kronos_vano enters the room.
08:57:28Maledictus enters the room.
09:01:04naeu enters the room.
09:22:18mvr enters the room.
10:11:13botanicus leaves the room.
10:39:17malumalu enters the room.
10:58:59malumalu leaves the room.
11:05:58malumalu enters the room.
11:11:15boyscoutCalling a private setter method permits self as a receiver - fd956a9 - Eugene Pimenov
11:15:18boyscoutCI: fd956a9 success. 1438 files, 7106 examples, 23403 expectations, 0 failures, 0 errors
11:20:27imajes leaves the room.
11:21:57chris2 enters the room.
11:22:41lstoll leaves the room.
11:41:37Maledictus leaves the room.
12:34:22mvr leaves the room.
12:39:11botanicus enters the room.
12:39:42malumaluwah, this drives me crazy >_>
13:00:38asap18 enters the room.
13:12:33qbproger enters the room.
13:12:46cout leaves the room.
13:16:59qbproger leaves the room.
13:21:17chris2_ enters the room.
13:31:20chris2 leaves the room.
13:33:46yipstar enters the room.
13:44:32stepheneb enters the room.
13:45:10gavinstark enters the room.
13:50:38gnufied enters the room.
13:57:31nari leaves the room.
14:02:20malumaluhm, can someone point me to the file which generates runtime/platform.conf?
14:06:33lstoll enters the room.
14:07:35outerim enters the room.
14:12:07yipstar leaves the room.
14:12:16dbussinkmalumalu: you can find it in rakelib/platform.rake
14:12:27dbussinkbut it consists of different parts
14:12:59dbussinka constants part, a structs part and a types part
14:13:11outerim leaves the room.
14:13:30outerim enters the room.
14:14:11malumaludbussink: thanks
14:17:56dbussinkmalumalu: what part is it you're looking at?
14:26:14nari enters the room.
14:29:20malumaluI just tested something by changing rbx.platform.rlimit
14:29:48malumaludbussink: but that didn't work, so I still have a spec failing
14:31:33jashmenn enters the room.
14:32:33binary42 enters the room.
14:36:13twbray enters the room.
14:46:52macournoyer enters the room.
14:48:23gnufied leaves the room.
14:48:53gnufied enters the room.
14:54:50PhilK enters the room.
14:58:03twbray leaves the room.
15:01:33twbray enters the room.
15:03:34evanmorning.
15:07:25jashmenn leaves the room.
15:12:12kronos_vanoevan, why you don't use your lighthouse tracker? There are some bugs/patches
15:14:21jashmenn enters the room.
15:19:09evankronos_vano: we do
15:19:15evanwe're just not always on top of it
15:22:48evankronos_vano: since you've reminded me, i'll go through it now
15:24:31botanicus leaves the room.
15:25:33stepheneb leaves the room.
15:27:45lstoll leaves the room.
15:28:56wmoxam enters the room.
15:29:23twbray leaves the room.
15:32:14twbray enters the room.
15:33:27twbray leaves the room.
15:34:45kronos_vanoevan, tnx :)
15:41:42jashmenn leaves the room.
15:56:57MarkMenard enters the room.
16:11:31twbray enters the room.
16:12:40moconnor enters the room.
16:13:24therealadam enters the room.
16:19:26boyscoutRemove silly way to set ivars - c58a454 - Evan Phoenix
16:19:26boyscoutFix header alignment - 50c157b - Evan Phoenix
16:19:26boyscoutFix Rubinius.asm block, they're full expressions - 30404a9 - Evan Phoenix
16:21:26jashmenn enters the room.
16:23:27moconnor leaves the room.
16:25:12boyscoutCI: 30404a9 success. .................................................................................................... .................................................................................................... .................................................................................................... ...............................................................................
16:25:18evan:(
16:25:58botanicus enters the room.
16:26:08rueSuccess
16:27:24rue"Mission Accomplished"
16:28:07rueShould have Lighthouse mail boyscout
16:30:53twbray leaves the room.
16:36:48evanyeah
16:38:39rueAlso not suck
16:39:56naeu leaves the room.
16:47:14imajes enters the room.
16:54:31brixenevan: I can regularly get an abort with no output compiling the specs on linux
16:54:38brixenthat's what all the dots are
16:54:53brixenand something is not setting $? either
16:55:06brixenwhen I ran with -fs to track it, it would not happen
16:55:19brixenso I back-burnered it
17:00:14malumaluhm, I found a wrong spec(frozen/core/process/setrlimit_spec.rb:72-78), what should I do?
17:00:39malumaluor: I think I found a wrong spec
17:01:45evanbrixen: hm. ok.
17:01:54malumaluit line 77, max = lim should be lim = max, I guess
17:02:03malumalu*line 76
17:08:15kronos_vano leaves the room.
17:10:25twbray enters the room.
17:12:58blowmage enters the room.
17:17:22enebo enters the room.
17:17:52ruemalumalu: Yes, I think it should be the other way
17:18:48nari leaves the room.
17:18:50malumaluI just send a pull request on github
17:22:08ruemalumalu: Unless evan is still around, can you pastie a format-patch? Be faster
17:23:11malumaluk, moment
17:25:32jashmenn_ enters the room.
17:31:32shoe leaves the room.
17:33:02pastierue: http://pastie.org/439673 by malumalu.
17:34:56shoe enters the room.
17:37:19jashmenn_ leaves the room.
17:39:22jashmenn leaves the room.
17:42:01ruebrixen: Should these go through RubySpec? ^^
17:43:40seydar enters the room.
17:44:10seydar leaves the room.
17:47:17joachimm enters the room.
17:47:30brixenI've been commiting to both spec/frozen and rubyspec on simple stuff
17:47:40brixenjust edit the patch to change the paths
17:47:52brixenI'm concerned about that fix though
17:48:01brixen1. why is it happening
17:48:09brixen2. why was the test reversed
17:48:21brixenmaybe it needs to test both ways
17:48:43brixenif max < lim blah elsif lim < max blah
17:49:25brixener max < lim .. max > lim rather
17:49:37brixenoh nvm
17:50:07brixenanyway, I do not know why it was added, so I have to assume the test was that way for a reason
17:50:11rueI cannot imagine why it should ever be higher than the maximum?
17:50:13brixenlacking any better info
17:50:26brixenI cannot either
17:50:51brixenat the same time, seems ridiculous to have to check that
17:50:58brixensince it's returned from a syscall
17:51:54mediogre enters the room.
17:53:21rueAh, you can probably raise the maximum in some cases
17:53:36rueStill, it would need better info
17:53:47ruemalumalu: What was the error you were getting?
17:53:51enebo leaves the room.
17:54:52malumaluEINVAL: Invalid argument - Invalid argument
17:55:20malumaluso lim > max when setrlimit was called
17:56:41rueOS X man says "The per-process resource limit is a limit on the amount of virtual memory locked, while the system-wide limit is for the number of locked physical pages."
17:57:41malumaluI always looked at http://linux.die.net/man/2/getrlimit
17:59:49brixenwell, seems odd that we cannot trust sanity coming back from getrlimit
17:59:59brixenbut it is logical that lim < max
18:00:08brixengo ahead with the patch
18:01:11brixenalthough, concerns me if max =
18:01:16brixenmax == 0
18:01:26mediogre leaves the room.
18:01:37brixenwhat will happen to the process when you setrlimit lim = 0
18:01:56brixenprobably nothing, cus I don't think we mlock anything
18:06:22brixenrue: are you going to commit it or should I?
18:11:02joearnold enters the room.
18:17:07ruebrixen: M, go ahead, I am poking around Sun's getrlimit
18:17:33brixenwell, if you're going to get to it, n/p
18:17:47brixenI'm trying to put some docs and specs together for the profiler
18:17:56brixenbut I'm leaving to look at an apt in about 15 min
18:18:22blowmage_ enters the room.
18:25:56ezmob enters the room.
18:29:59enebo enters the room.
18:30:17blowmage_ leaves the room.
18:30:48blowmage_ enters the room.
18:30:56blowmage leaves the room.
18:32:23libcCould somebody close https://rubinius.lighthouseapp.com/projects/5089/tickets/757-51a2393-does-not-compile-on-amd64 please ?
18:33:00boyscoutFix 64bit compilation issue introduced in “Rewire JIT” - 20bb4d1 - Eugene Pimenov
18:35:11ruelibc: ^^ fixes?
18:35:23libcrue, yes
18:36:10boyscoutCI: 20bb4d1 success. .................................................................................................... .................................................................................................... .................................................................................................... ...............................................................................
18:37:31ruelibc: OK, add ticket numbers to commit messages when appropriate too
18:38:02ruelibc: I will send a Lighthouse invite, is the e-mail you use for Git OK or do you already have/want a different one?
18:38:47libcrue, really? I don't see commits in overview… I though lighthouse is not connected to git
18:39:00libcrue, it's fine
18:39:22tilmanlibc: i think he's asking that you put a reference to the ticket in your git commit messages
18:39:23rueIt is not :) That is why it helps if you add something like "Closes #60"
18:40:04libcoh =) I thought he's talking about "[#42 state:resolved]" as rails does
18:41:00headius enters the room.
18:43:43rueGreat, mark as ham is broken again
18:46:24kronos_vano enters the room.
19:09:16blowmage enters the room.
19:14:51headius leaves the room.
19:18:34nicksieger leaves the room.
19:21:24twbray enters the room.
19:25:18blowmage_ leaves the room.
19:30:54evani'm back.
19:31:12evanlibc: good job on the JIT
19:31:24evanlibc: i'm going to be doing a bunch more JIT work this week
19:36:19kronos_vano leaves the room.
19:39:37libc_ enters the room.
19:39:52kronos_vano enters the room.
19:44:07tilmanevan: do you want to stick with ascii for rbc files?
19:44:19evanunless we have a reason not to
19:44:38tilmanlike performance and size?
19:45:17evanwe've yet to have either problem
19:45:19evanthat i'm aware of.
19:45:53tilmanaha
19:46:04tilmanimo vm startup time is quite bad
19:46:06evani'm open to ideas
19:46:12evanwell
19:46:14evanin profiling
19:46:20evan.rbc parsing has never shown up
19:46:23tilmani've got a couple of patches that reduce the size of the compiled kernel and the startup time noticably
19:46:25evanin VM startup
19:46:37evanwhy are you sitting on these patches then?
19:46:46evan:D
19:47:09tilmanbrixen told me he prefered ascii rbc's for readability/debugability
19:47:15tilmanso i wanted to check with you firt
19:47:17tilmanfirst*
19:47:18evanas do I
19:47:21libc leaves the room.
19:47:24evanwhat do your patches do?
19:47:31evanas I said
19:47:35evani'm not opposed to new ideas
19:47:44evanjust need to be sure there is a problem that actually being solved
19:48:18tilmantime bin/rbx hw.rb
19:48:28tilmanprinting hello world takes half a second
19:49:05tilmanno, i don't run that all day :)
19:49:17evanyeah, i know it could be faster
19:49:27ruelibc_: Looks like Unicode in that last commit message again.. are you able to produce ASCII quotes? I am not quite up to date with modern cyrillic keyboards
19:49:30evanmy question is, have you profiled startup to see whats slow?
19:49:45tilmanmaking the rbc format smarter (ascii -> binary) helps with that
19:50:01tilmanevan: no, i've had an old-fashioned look at the rbcs and noticed they were fat
19:50:11tilmanwhen there's no obvious (to me) need for them to be
19:50:27libcrue, what's wrong with unicode in commit messages?
19:50:29evanwhen I changed the format
19:50:33evanit was a couple of reasons
19:50:34boyscoutFixes a wrong spec. - 5b48063 - Maximilian Lupke
19:50:36evanit was much easier to debug
19:50:52evanand it was much easier to write ruby classes to encode/decode
19:51:00evanand it doesn't have byte encoding problems
19:51:01ruelibc_: Trying to keep to lowest common denominator
19:51:20evanplus, for encoding of integers, there is no size limitation
19:51:29evanso Bignum's can be easily encoded
19:51:40tilmani've used variable length integers for that
19:51:54boyscoutCI: 5b48063 success. 1438 files, 7106 examples, 23403 expectations, 0 failures, 0 errors
19:52:03evanhow much does changing them to a binary encoding change things?
19:52:08evantime wise.
19:52:09libcrue, It would be hard to write a program without "able to produce ASCII quotes" ;-)
19:54:04tilmanevan: i took notes on the size of runtime/ only, but i think startup time improved 15-20% with my limited patching
19:54:20tilmani didn't go the whole way yet, just experimented a bit
19:54:38evansure
19:54:49evanthis is probably a great example of where doing stuff in a fork would be awesome
19:54:53evanyou can do whatever
19:54:55evanget it working
19:55:00evanand then, when you feel like it's good
19:55:04evanwe can fold it in
19:55:20tilmanokay
19:55:38evanthat way, it's easy for me to check out what ya did
19:55:56evanand criticize :)
19:56:24tilmanworks for me
19:56:30evandope
19:56:46tilmani'll ping you after the holidays
19:56:48ruelibc_: Heh, well, an editor might have a convenience for it
19:56:54evantilman: oh also
19:57:00evani did the .rbc change we switch to C++
19:57:06evantilman: and i was being super lazy
19:57:18evanso i just used iostreams to decode the ascii int stream :)
19:57:21evanas you've seen
19:57:38evantilman: if we make it a bit more opaque
19:57:51evanwe should also write some tools to make decoding to human form easy
19:58:01tilmanyes, i was going to mention that
19:58:05libcrue, ok, I won't use “” in commit messages… (btw it have nothing to do with cyrillic keyboard, it's mac alt + [ )
19:58:09tilman:)
20:01:29rueOh, yeh. I only use alt + u + <ao>
20:03:07evanI used shift-q and shift
20:03:14evanjust shift-q.
20:04:29brixenevan: my 2 cents again are use .rba's if size is an issue, read into a buffer and iterate for loading .rbc, and leave .rbc a simple \n delimited ascii
20:04:42brixencompression will be really good on the ascii
20:04:46brixenif space is an issue
20:04:55brixenit's not for 99% of installs I think
20:05:01evani think tilman isn't looking at the space, but rather the decode time
20:05:12brixenwell, he mentioned both
20:05:15brixeninitially
20:05:18evanyeah, he did.
20:05:56shoeI hear space and time are related anyway... somehow.
20:06:09brixenreading .rbc could be a lot faster, but it probably won't be faster if we start complicating the format
20:06:18brixenI'd really like to not do that
20:06:35scooprmake it mmap'able and it'll be superfast! ;)
20:06:38tilmanof course reading a binary format is faster than reading ascii
20:06:41brixenshoe: it's all relative man :)
20:06:48rueshoe: LLVM has an FTL optimizer
20:06:50shoelol
20:07:00libcshoe, general relativity? einstein?
20:07:18evantilman: not in ruby, it's not.
20:07:22brixenevan: are you back stateside?
20:07:26evannegative.
20:07:30brixenahh
20:07:41evanhackin' with argentines
20:07:46brixennice!
20:07:56evani'll be back thursday morning
20:08:03evanmy flight leaves tomorrow nite
20:08:07brixenI'm at my fave coffee shop/art gallery/small band venue
20:08:11brixenah ok
20:08:15evannice
20:08:21brixenI need to get out more
20:08:31brixengoing a bit crazy in the burbs
20:08:43brixenoh, and it's like 70 deg here!
20:08:45tilmanevan: this code surely will stay in c++ though?
20:08:47brixeninsane
20:08:58evantilman: i'd prefer to move it to ruby
20:09:01evaneventually.
20:09:05brixentilman: it's written out from ruby now
20:09:11brixenwhen the compiler runs in MRI
20:09:16evanyes, the writer is always ruby
20:09:27evanbut we use both a C++ and a ruby reader
20:09:31tilmanbrixen: yes, i'm mostly interested in the reading side atm
20:09:35tilmanok
20:09:52evantilman: so, feel free to play around and experiment in a branch
20:09:53brixenwell, writing precedes reading :)
20:10:01evantilman: and we'll check it out
20:10:06tilmanok
20:10:08evantilman: no promises though, k?
20:10:27tilmangot it
20:11:29nicksieger enters the room.
21:16:45jashmenn enters the room.
21:17:12chris2 leaves the room.
21:21:08gavinstark leaves the room.
21:21:18benny leaves the room.
21:40:56malumaluI think I found two specs which contradict each other
21:42:00malumalufrozen/core/struct/new_spec: line 51 & 62
21:42:51malumaluone spec expects the objects to respond to to_sym, the other expects them to be Symbols
21:43:15malumaluso one is pretty useless I think
21:55:42rueI think it is related to the one below it
21:57:49PhilK leaves the room.
22:06:14lstoll enters the room.
22:12:41dbussink leaves the room.
22:18:14macournoyer leaves the room.
22:20:34boyscoutfix reverse_each - 09950d3 - Ivan Samsonov
22:20:34boyscoutCheck that roots aren't null. - f6f89be - Evan Phoenix
22:22:13boyscoutCI: 5b48063 success. 1438 files, 7106 examples, 23403 expectations, 0 failures, 0 errors
22:27:41mutle_ enters the room.
22:30:00lstoll_ enters the room.
22:39:17evanso
22:39:23evanBlue Ruby claims they run rails
22:39:24evanhttp://www.slideshare.net/schmerdy/blue-ruby-sdn-webinar-1260181
22:39:28evanslide 23
22:39:34evani call shenanigans
22:41:56therealadamdoesn't slide 22 say no?
22:42:02therealadamand that they run 0% of the stdlib?
22:42:24evanyeah
22:42:25therealadamand they run fib(20) in 1.8s
22:42:32evanexactly.
22:42:35evantheir slide deck is confusing.
22:42:44evani don't get what the line graph is.
22:42:48evanwhats an "RFC runtime" ?
22:43:00therealadamenterprisey
22:43:10evanso it's a pudding pop?
22:43:25evan<cosby> And then ya put the puddin pop in ya mouth </cosby>
22:43:30therealadam+/- bill cosby, yes
22:45:40mutle leaves the room.
22:48:56rueThe really weird thing about this, though
22:49:10rueI do not recognise a *single one* of those names
22:50:00lstoll leaves the room.
22:53:25soveran enters the room.
22:57:55rueYou would think that at some point one of them would have popped up on -core or something
23:00:58drbrainrue: unless their corporate masters forbade it due to possible licensing issues
23:01:30blowmage leaves the room.
23:03:13blowmage enters the room.
23:03:57evanfor those with a webinar fetish: https://sap.na.pgiconnect.com/p16473929/
23:04:04evanimpl details start at minute 55
23:04:37evanrue: yeah, as far as I can tell, none of them are in the ruby community anywheres
23:05:41rueI think the demo app just means the app is built (not that it runs)
23:06:08rueOr, well, based on the numbers it would take them about a month to find out if it did
23:06:41rueevan: The really sad thing is that it is still more open than MagLev is :/
23:07:07evan:(
23:07:13ruedrbrain: It is possible, I suppose, but I would think highly unlikely?
23:07:31evanone of their slides has sexps on it
23:07:36evani wonder what parser they're using
23:07:39evanruby_parser maybe
23:09:22rueThat or the bison, almost necessarily
23:09:52evanperhaps
23:10:38botanicus leaves the room.
23:12:37evaninteresting thing about blue ruby is that it's pretty much not competing with anyone else
23:12:43evansince it's ruby for SAP programmers
23:12:54evanand I don't think i've met a SAP programmer
23:14:30wmoxam leaves the room.
23:20:47slavai evan
23:20:48jashmenn leaves the room.
23:20:49slavahi*
23:20:53evanallo
23:25:27jashmenn enters the room.
23:27:07jashmenn leaves the room.
23:27:23qbproger enters the room.
23:27:39jashmenn enters the room.
23:31:56soveran leaves the room.
23:33:31nicksieger leaves the room.
23:36:04enebo leaves the room.
23:39:57imajes leaves the room.
23:41:45therealadam leaves the room.
23:57:50binary42 leaves the room.