www.parrot.org | Parrot 1.6.0 "half-pie" released: feel free to tear out the JIT! | Testing priorities: Exception and MultiSub
Set by moderator on 18 September 2009.
darbelo I *think* it's mergeable with some cleanups. But the sheer amount of EVIL that's left over in frame_builder.[ch] makes me want to have it *really* well tested before merging. 00:01
JIT is dead, but I haven't finished cleaning up the severed libs, and there's still some blood on the walls. So to speak. 00:03
cotto_work sounds like it put up a fight 00:04
darbelo Actually the hard part was keeping parts of it alive :) 00:05
I mean, there's a lot of JIT that got left in, mostly, to support the call frame builder. 00:07
cleaning up *that* sucker is a job for another branch. 00:08
chromatic If you want to make a list of tasks on the wiki, we can all pile on. 00:13
00:15 fperrad joined
darbelo I'm not sure how to partition it, the job is basically, "Pick a struct in frame_builder.h, see if we can do without it and kill it." 00:15
Same of functions in fram_builder.c 00:16
Whiteknight is getting close to resolving the clone_p_p segfault 00:19
cotto_work that sounds like it'll split nicely into small mostly-orthogonal subtasks
Whiteknight although it's not pretty
Austin whiteknight ++ 00:20
What's your take on it?
cotto_work darbelo, should people who want to help work in the branch or wait util the merge?
Whiteknight Austin: I've created a little registry API. When we start deep cloning we create a registry, and search for mapped items there before creating clones 00:21
darbelo cotto_work: I guess I'll clean up as much dead JIT code as I can tonight, and then merge. 00:22
Austin What's the cost?
purl somebody said the cost was not VERY relevant.
cotto_work Whiteknight, you weren't joking.
I guess it's an ugly problem though. 00:23
Whiteknight Austin: no idea, obviously depends on the size of the cloned aggregate and the number of duplicates
cotto_work: this is just the first attempt at it, there may be better solutions that I haven't thought about
darbelo Eeewww.
00:23 Wolong_ joined 00:24 mokurai joined
cotto_work yeah. first make it work, then make it sane 00:24
00:24 jrtayloriv joined 00:25 ruoso joined
darbelo *hates* the realclean dance. 00:25
00:25 fperrad_ joined
cotto_work goes home 00:26
friday++
Whiteknight friday++ 00:28
Tene ā†‘friday
darbelo Well, crap. There's an intermittent failure in t/tools/parrot_debugger.t 00:29
I've spent about an hour chasing it in the kill_jit branch.
I've also just reproduced it in trunk. 00:30
I am not happy.
darbelo whishes for *consistent* failures. 00:31
Whiteknight oh, in parrot you just have to define FAILURES_ARE_ALWAYS_CONSISTENT to 1 00:32
in fact, I don't know why that's not the default
dalek rrot: r41347 | darbelo++ | branches/kill_jit (3 files):
Kill some more leftover jit functions.
 00:33
00:36 PacoLinux_ joined 00:43 jrtayloriv joined 00:45 ZeroForce joined
Coke . 00:48
ZeroForce ..
darbelo ... 00:49
00:51 Zak joined 00:53 payload joined
Tene ! 00:55
Whiteknight yay! my ugly API functions finally build. Now I need to actually *use* them and wait for segfaults to happen 00:56
Tene yay! 00:57
ZeroForce throws some kudos to Whiteknight.
jrtayloriv my $x = '...'; loop { say $x; $x = $x ~ '.'; } 00:58
ZeroForce jrtayloriv: hehe
dalek rrot: r41348 | darbelo++ | branches/kill_jit/src (2 files):
Remove more dead JIT code.
 01:01
01:08 kid51 joined
Whiteknight !!! Fixed the clone segfault! 01:08
(of course, I created a segfault in PGE...) 01:09
01:11 TiMBuS joined
dalek rrot: r41349 | darbelo++ | branches/kill_jit/docs (10 files):
Remove JIT references from the docs.
 01:18
Austin lol 01:19
Whiteknight++
Whiteknight Austin: patch incoming (though it doesn't work 100%, it does fix the hash issue) 01:20
Austin Does it blow up PGE? 01:22
'Cause I've already worked around the hash issue -- stop using symtables. 01:23
Whiteknight yeah, it breaks PGE 01:24
Austin (Whereas if you blow up PGE, the next sounds you hear will be my keyboard falling to the floor, then a taxi pulling to a stop at your front door...)
Whiteknight yay! a sleep over! 01:26
cotto with the fishes, even 01:27
Austin cotto++ 01:28
Wow. It's surprising how much stuff broke when I start trying to resolve overloaded names. 01:29
Whiteknight is never surprised by how much things break when doing something small and innocuous 01:30
Austin The thing is, I was expecting encapsulation to provide much better protection than it did. 01:33
Whiteknight just because your code is encapsulated doesn't mean parrot respects that 01:34
Austin I was "reaching in" too much, and the implementation approach I took (return an array instead of a scalar) was confusing in the context of all the other arrays I'm passing around.
darbelo goes on the hunt for food. 01:38
Whiteknight has bed now. Later
dalek rrot: r41350 | darbelo++ | branches/kill_jit (4 files):
Remove the jit_info member of the PackFile_ByteCode struct. Pull a whole lot of dead code with it.
darbelo See y'all later.
Austin Laugh. It's not parrot breaking encapsulation. It's a PEBKAC problem.
01:38 darbelo left
Whiteknight PEBKAC? 01:38
purl Problem Exists Between Keyboard and Chair or at www.userfriendly.org/cartoons/archi...80506.html or a pretty safe answer or see PEBCAK or www.thinkgeek.com/tshirts/sysadmin/6692/
Whiteknight ah 01:39
Austin xkcd: Blockbuster Mining. Harriet the Spy 01:41
Whiteknight haha, yeah
Whiteknight really has to go to bed now. Goodnight
cotto clock? 01:46
purl cotto: LAX: Fri 6:46pm PDT / CHI: Fri 8:46pm CDT / NYC: Fri 9:46pm EDT / LON: Sat 2:46am BST / BER: Sat 3:46am CEST / IND: Sat 7:16am IST / TOK: Sat 10:46am JST / SYD: Sat 11:46am EST /
cotto he must have a really early day tomorrow
Austin Well, he /is/ married. 02:00
chromatic He wants to get a head start on Talk Like a Pirate Day. 02:02
12:01 EST, it's all YARRRRRRRR in Casa Whiteknight.
Austin That .. is .. so .. wrong .. 02:03
Does anyone know if {{{ :multi() }}} is valid in PIR? 02:04
chromatic I believe so. 02:06
Austin How do I check 'isa' across hll's? 02:10
I'm doing {{{ $I1 = isa $P0, [ 'parrot' ; 'Undef' ] }}} but it doesn't seem to be working. 02:11
Well, actually, it /is/ working. But it's working wrong.
02:11 joeri left
chromatic Prescriptivist. 02:13
Austin ?
chromatic A little linguist humor. Very little.
Austin Ok.
Aha. Using the convenient 'defined' opcode. 02:26
02:35 janus joined 02:40 zak_ joined 02:53 rg1 joined
kid51 must sleep 03:08
purl $kid51->sleep(8 * 3600);
dalek rrot: r41351 | darbelo++ | branches/kill_jit/MANIFEST:
Update MANIFEST.
 03:29
03:30 kyle_l5l joined 03:43 Austin joined 04:01 zak_ joined 04:05 dukeleto joined
dukeleto 'ello 04:06
Austin Hello, dukeleto 04:12
dukeleto Austin: how goes?
Austin I keep learning new things about Parrot. :(
04:13 dukeleto_ joined
dukeleto_ yes, sometimes learning is painful 04:13
Austin Also, frequently pointless. Where's my DWIM? 04:14
And now, I believe I'm going to cry. 04:16
I've been chasing this @##$%#^$^#$%@#$% bug for hours now.
$I0 = defined $P0
It returns false on some truly bizarre things.
An empty string gets passed in, it returns false.
A Capture object is passed in - false. 04:17
Turns out, I wasn't loading $P0 04:18
04:22 Zak joined 04:34 Andy joined 04:46 theory joined 05:02 mokurai joined
dalek rrot: r41352 | darbelo++ | branches/kill_jit/src (2 files):
Another batch of JIT leftovers removed.
 05:21
05:27 mikehh joined 05:58 Zak joined
mikehh kill_jit branch - All tests PASS (pre/post-config, make test / fulltest at r41352 - Ubuntu 9.04 i386 05:58
kil_jit branch - one codetest TODO pass - t/codingstd/c_function_docs.t - ok 19 - compilers/pirc/src/bcgen.c # TODO Missing function docs 06:02
kill_jit branch - different subtest number from my previous test - but same program 06:04
06:14 zak_ joined 06:25 Zak joined
mikehh All tests PASS (pre/post-config, smoke, nqp_test, fulltest) at r41352 - Ubuntu 9.04 i386 06:29
06:40 HG` joined, payload joined 06:43 zak_ joined
dalek rrot: r41353 | darbelo++ | branches/kill_jit/src (2 files):
Another cleanup pass. Now with extra struct pruning.
 06:46
mikehh partcl r742 builds on parrot r41352 - make test PASS - Ubuntu 9.04 i386 06:49
rakudo (4b141a8) builds on parrot r41352 - make test / make spectest (up to 28295) PASS - Ubuntu 9.04 i386 06:55
06:59 flh joined 07:09 cotto_w0rk joined
jrtayloriv glances furtively over his shoulder and scribbles profanity on the wall while nobody is looking. 07:14
07:15 zak_ joined 07:18 bacek joined
bacek o hai 07:18
jrtayloriv Howdy bacek!
bacek jrtayloriv: not so bad. How is your gc refactoring going? 07:19
07:19 Austin joined
jrtayloriv done -- I merged it a day or two ago. 07:19
I'm reading the GC book by Jones right now, and hoping to do much more afterwards.
cotto gc book? 07:20
jrtayloriv cotto, "Garbage Collection: Algorithms for Automatic Dynamic Memory Management" by Richard Jones
Totally delicious.
cotto I don't think I've made a commit most of this week. 07:21
That's a bad feeling.
07:21 darbelo joined
jrtayloriv cotto, I didn't have internet at all for a week -- *that* was a bad feeling ;) 07:21
Although, I got a lot of reading done offline, which was nice. 07:22
cotto If we'd switched to git, that wouldn't have been a problem. ;)
bacek cotto :)
dalek rrot: r41354 | bacek++ | trunk/examples/pir/make_hello_pbc.pir:
[examples] Add example how to generate working PBC using Packfile PMCs.
cotto is totally on the "can it be git tiem nao plz" bandwagon 07:24
07:24 fperrad joined
darbelo thinks git is a better fit to the parrot workflow than svn can ever be. 07:25
cotto bacek++ for a very shiny example, even if it's a bit heavy on magic numbers 07:26
dalek TT #1011 closed by bacek++: Packfile PMCs cannot create a bytecode file from scratch 07:27
cotto it reminds me of looking at the jit code in that it's code to write code. 07:28
bacek cotto It's example for PBC creating, not how to write PIR compiler :)
cotto cosa? 07:29
bacek "cosa"?
dalek rrot: r41355 | darbelo++ | branches/kill_jit (20 files):
Sync with trunk.
cotto cosa is an unimplemented and untested idea for graphically creating bug-free programs 07:30
bacek darbelo: merge it back!
cotto holy shit... People still believe it's possible?
cotto I thought it seemed like a good idea until I realized that the author didn't have any actual code.
afaict he's a crackpot
darbelo cosa is also 'thing' in spanish 07:31
purl okay, darbelo.
cotto and in other languages
darbelo i don't speak other langiages :) 07:32
cotto basically, don't waste your valuable hacking time reading up on it
darbelo or languages.
purl languages are tools. just like php programmers.
cotto purl wins
purl MENTALITY!
darbelo purl++
cotto I actually like PHP. I'm happiest when I have something to gripe about, and PHP gives me an unbelievable amount of that. 07:33
It's kinda like a pug, but 100x
Austin lol 07:34
cotto and absolutely not like pugs
Austin If you like PHP, you'll love NQP.
cotto NQP at least has a design goal 07:35
Austin So does PHP. "Sure, we support that!"
mikehh darbelo: how close are you merge - I ran some tests on r41352 - all ok
cotto nqp isn't supposed to be of interest to anyone other than language devs, and it's odd because much of it comes from perl 6. 07:36
mikehh darbelo - on i386
Austin bacek: In your example, do the pfconst[1..4] entries have a particular order requirement?
cotto but yes, it yields good griping material
eor
bacek Austin: no. But they have to be properly referenced from bytecode. 07:37
Austin okay.
bacek Austin: I actually just created hello.pbc from hello.pir and then reconstruct it using Packfile PMCs :) 07:38
Austin How else?
cotto 'night all 07:39
Austin What is the difference among 'name', 'subid' and 'ns_entry_name' ?
Good night, cotto.
bacek Austin: it should be in some PDD... But I don't know which
Austin $P1 = new 'Sub', $P0 07:40
This works?
Could you invoke it? $P1()
Okay, that is cool, but incredibly scary. 07:42
bacek++ for making that happen, and getting Packfile PMCs to work
darbelo mikehh: I'm merging now.
But it's slooooooooooooow. 07:43
bacek Austin: no, you can't invoke Sub made from "new 'Sub', $P1". This Sub isn't "attached" to running interpreter.
07:46 desertm4x_ joined
mikehh darbelo: well make test PASSed at r41355 on Ubuntu 9.04 i386 07:48
darbelo mikehh: good to know.
mikehh darbelo: waiting to to test trunk 07:49
darbelo mikehh: don't hold your breath. This is slooooooooooooow. 07:50
bacek darbelo: use git-svn. At least you'll not loose time waiting for svn :) 07:52
darbelo And it just gave a server error.
bacek sigh... 07:54
wanna help from proper vcs? :) 07:55
darbelo I'm trying it again on another checkout. And it so horribly slow i want to cry. 07:56
Another server error. 07:57
bacek passing napkins to darbelo evilly laughing looking at local git checkout
darbelo cries.
bacek: Want to merge it? svn has decided I can't. 07:58
chromatic At least it gives you merge conflicts with all of its slowness.
darbelo chromatic: No, it gives me "svn: Server sent unexpected return value (304 Not Modified) in response to GET request" 07:59
chromatic That's it, I'm changing my vote!
bacek darbelo: let me run make test before committing it :) 08:00
darbelo bacek: test it as much as you want to.
08:01 iblechbot joined
darbelo Well to be fair, the server has it right here, I didn't modify the file it's barfing on. 08:02
But I would love to know why my working copy thinks I did.
There was a CVS option in the poll right? 08:03
jrtayloriv I should have voted for Mercurial, just to be special.
mikehh thought of voting for Bazaar 08:04
mikehh it's really easy to use - but it has a python interface :-{ 08:05
jrtayloriv mmm 4:08 here ... sleepy time.
Oh crap -- it's actually only 4:06. Nevermind. 08:06
7 08:07
OK -- 4:08 here ... sleepy time.
bacek darbelo: incoming 08:08
purl incoming is pause.perl.org/incoming/
bacek make test passed in Linux/i386
darbelo wonders if he should have gotten some tests on windows before merging. 08:09
Austin pause.perl.org - invalid security certificate 08:10
bacek "windows"? Ah. Some piece of software mis-called "operating system".
mikehh ok got it :-} 08:11
bacek mikehh: I'm running fulltest on Linux/i386 now. 08:12
darbelo: can you check that I merged it properly 08:13
darbelo Checking now.
dalek rrot: r41356 | bacek++ | trunk (83 files):
Merge branch 'kill_jit' back to trunk.
 08:14
bacek Next time I'll probably ask allison for merging branches using plain svn...
darbelo The deletions/additons look right. Configure game sane output, and it's building ok so far. 08:15
bacek: Did you get any conflicts? 08:16
bacek darbelo: quite few of them. MANIFEST, src/jit.c, src/pic.c, packfile.[ch]
jit.c and pic.c I just removed. packfile was related to #ifdef 08:17
mikehh bacek: I am on i386 at the moment - will try amd64 in a bit 08:18
bacek mikehh: ok. 08:19
darbelo: my local diff between trunk and branch showing only absense of "src/pic.c".
darbelo bacek: looks good to me. bacek++
bacek wonders about it. I think it was removed in kill_pic branch. 08:20
darbelo: it's just about using right tools for the job :)
git++
darbelo svn-- 08:21
bacek karma svn
purl svn has karma of 16
bacek karma git
purl git has karma of 241
bacek "...and the winner iiiiis..."
karma hg 08:22
purl hg has karma of 7
darbelo karma c
purl c has karma of 8014
bacek karma bzr
purl bzr has karma of 7
darbelo karma cvs
purl cvs has karma of -48
darbelo karma rcs
purl rcs has karma of 5
darbelo karma sccs
purl sccs has karma of 1
darbelo karam cssc
karma cssc
purl cssc has karma of -1
darbelo karma darcs
purl darcs has karma of 7
bacek TapTinder looks good so far 08:23
darbelo All tests successful. here.
And it will be your fault if it breaks :)
bacek "manifest" tests are failing here. But they always fail on git-svn checkout 08:24
darbelo: no way!
mikehh I'll fix that
bacek t/codingstd/c_function_docs (Wstat: 0 Tests: 273 Failed: 0)
TODO passed: 46
darbelo mikehh: did it for you. 08:25
bacek mikehh: it's not related to MANIFEST. Test just trying to invoke "svk"
mikehh bacek: yeah got that before :-}
dalek rrot: r41357 | darbelo++ | trunk/MANIFEST:
Update MANIFEST.
darbelo examples/pir/make_hello_pbc.pir was missing from MANIFEST.
mikehh some codetest failures
bacek so. we don't have undocumented functions anymore? All of them were in jit?
bacek think that last roadblock for Lorito was just removed. 08:27
chromatic I'm sure a lot were in JIT.
darbelo mikehh: git doesn't handle svn props, they got lost in the merge. fixing them now.
bacek No JIT, pluggable runcores, Packfile PMCs are capable for creating PBC.
darbelo bacek: There's still some JIT leftovers in frame_builder.[ch] 08:28
bacek darbelo: I believe YOU can fix it :) 08:29
darbelo Not without breaking the frame builder :)
But maybe it can be made less ugly ;)
mikehh darbelo: make html - Failed to process docs/jit.pod., docs/native_exec.pod. and tools/build/jit2c.pl. 08:32
darbelo svn ci
mikehh: What is the makefile for that. 08:34
dalek rrot: r41358 | darbelo++ | trunk (4 files):
Fix svn props after the merge.
 08:35
chromatic A 9.272% smaller binary too. 08:42
dalek rrot: r41359 | darbelo++ | trunk/examples/pir/make_hello_pbc.pir:
Make make_hello_pbc.pir pass codetests.
mikehh darbelo: looking into it now - just running the rest of the tests 08:43
darbelo mikehh: I have no perldoc here, and have to head out in a bit. I'll look into it later. 08:45
chromatic They probably have a mention in one of the .pm files under lib/
Yep, fixing.
darbelo Ok. See y'all later. 08:46
08:46 darbelo left
dalek rrot: r41360 | chromatic++ | trunk/lib/Parrot (3 files):
[docs] Removed references to JIT documentation and utilities since removed.
 08:49
08:59 JimmyZ joined
dalek kudo: 0d77e42 | mberends++ | tools/test_summary.pl:
[tools/test_summary.pl] use Time::HiRes to work on Windows, code cleanup
 09:14
rrot: r41361 | mikehh++ | trunk/t/codingstd/c_function_docs.t:
remove jit files from c_function_docs data and TODO pass
 09:29
mikehh All tests PASS (pre/post-config, smoke, nqp_test, fulltest) at r41361 - Ubuntu 9.04 i386 09:58
10:01 fperrad_ joined 10:07 MoC joined
dalek rrot: r41362 | chromatic++ | trunk (3 files):
[interp] Removed binop_mmd_funcs pointer and n_binop_mmd_funcs count from

very nice. I recommend a make realclean after this commit.		 10:10
mikehh rakudo (0d77e42) builds on parrot r41361 - make test / make spectest (up to 28298) PASS - Ubuntu 9.04 i386 10:15
10:15 cognominal joined
mikehh partcl r742 builds on parrot r41361 - make test PASS - Ubuntu 9.04 i386 10:19
10:24 Whiteknight joined 10:27 darbelo joined
moderator www.parrot.org | Parrot 1.6.0 "half-pie" released: feel free to put in a new JIT! | Testing priorities: Exception and MultiSub 10:29
dalek rrot: r41363 | darbelo++ | branches/kill_jit:
Kill the branch that killed the JIT.
 10:31
10:43 tokuhirom___ joined 10:47 payload joined 10:54 masak joined 11:00 tokuhirom___ joined
mikehh All tests PASS (pre/post-config, smoke, nqp_test, fulltest) at r41362 - Ubuntu 9.04 i386 11:04
Whiteknight that frame builder is U G L Y ugly 11:11
mikehh partcl r742 builds on parrot r41362 - make test PASS - Ubuntu 9.04 i386 11:21
darbelo Whiteknight: It is made out of distilled JIT ugly. 11:25
It's there to save those little pieces of code that no matter how hard we killed, hell wouldn't take :) 11:26
Whiteknight I think the frame builder would be an excellent opportunity to test out LLVM or one of the other JIT solutions 11:27
darbelo Kill it with fire!
Whiteknight That would be a great start on the road to JIT
mikehh rakudo (0d77e42) builds on parrot r41362 - make test / make spectest (up to 28298) PASS - Ubuntu 9.04 i386 11:31
darbelo IIRC the only externally callled functions there were Parrot_jit_clone_buffer, Parrot_jit_free_buffer, get_nci_*, set_nci_* and Parrot_jit_build_call_func. 11:32
And now it's all there in one place, for a more convenient killing :) 11:34
... and 60% of the file is just one function. Eeew. 11:40
flh most HLLs depend on Perl5 for building, since they have a "Configure.pl" file: has anyone ever tried to ship instead a "Configure.pir" or "Configure.nqp"? 11:51
darbelo flh: I'm building a Configure.pir for decnum-dynpmcs, but I got sidetracked with the JIT-killing. 11:52
It's pretty trivial to do, in my case. All I have to do is call /tools/dev/gen_makefile.pl on my makefile templates. 11:54
flh so you're still using perl5 :) 11:57
darbelo I'm using whatever and install-dev'ed parrot provides me. I don't look at it :) 11:58
mikehh cardinal - builds on parrot r41362 - rake test:all - 3 files fail to compile - Ubuntu 9.04 i386 12:00
darbelo But yeah. I have two pieces of perl code as part of my project, the Confiugre script (on it's way out) and the test harness, which i'm not going to replace yet. 12:01
I guess somebody needs to write a TAP parser for parrot before that can happen. 12:03
flh: Oh, and cardinal got rid of perl. They replaced it with ruby. 12:07
12:11 JimmyZ joined 12:31 hugues23 joined 12:40 preflex joined 12:51 hugues23 left 12:57 kid51 joined
dalek rrot: r41364 | darbelo++ | trunk/src (2 files):
Remove useless functions and #defines from the frame builder.
 12:58
13:00 joeri joined 13:20 HG` joined
dalek a: ec38101 | fperrad++ | (6 files):
allows require of *.pbc compiled from .lua
 13:39
a: b63f6c9 | fperrad++ | src/lua51.pir:
paths with / & \\\\
a: bb6b94e | fperrad++ | config/makefiles/root.in:
compile & install Test.More
a: 59c7153 | fperrad++ | src/lib/luapackage.pir:
try .pbc before .lua (faster)
13:48 Andy joined
dalek a: 7a11a6a | fperrad++ | config/makefiles/root.in:
fix cleanup
 14:05
14:21 payload joined 14:39 Psyche^ joined 14:44 sri joined 14:57 sri joined 15:02 JimmyZ joined 15:04 sri joined
dalek tracwiki: v17 | cotto++ | GitObjections 15:26
tracwiki: note that bacek++ has been cleaning up after the svn server's messes
tracwiki: trac.parrot.org/parrot/wiki/GitObj...ction=diff
15:27 jrtayloriv joined
cotto anyone used gcspy? 15:29
15:32 sri joined
jrtayloriv cotto, Looks interesting. Apparently, we'd have to write some extra code in the GC to use it, though. 15:40
cotto I think you just preload a library and it takes care of the rest. I'm playing with it now. 15:43
jrtayloriv cotto, OK -- maybe that was just for the older version then. This page --> www.experimentalstuff.com/sunr/proj...index.html <-- which was linked to from the gcspy site, says that a "driver" must be written for it. But that was the old GCSpy page, so maybe that's not true anymore. 15:46
15:46 desertm4x__ joined
jrtayloriv Looks like a very neat application. Let me know if it works -- I'd like to play around with it, if so. 15:47
jdv79 moritz: how about that patch?
cotto will do. I just got it built.
jdv79 my pulls got blocked cause git had a hissy fit about it:(
moritz jdv79: sorry, I forgot that - will merge now 15:48
jdv79 thanks 15:49
moritz ah, merging was trivial 15:50
dalek rrot: r41365 | fperrad++ | trunk/MANIFEST.generated:
[library] install uuid.pbc
jdv79 weird, git said something like build/Makefile.in needs update, cannot merge 15:51
whatever 15:52
dalek kudo: 2c40a5b | moritz++ | (2 files):
support smolder reports in harness and Makefile. Patch courtesy by jdv79++
 15:53
rrot: r41366 | fperrad++ | trunk/tools/dev/mk_inno_language.pl:
[languages] recurse subdirs for library
 15:54
cotto odd. the build completes but I don't see any class files.
nm
there aren't supposed to be any, apparently
academic code is weird. 15:55
15:56 davidfetter joined
cotto at least there's code this time 16:02
meh. I'll have to either wrestle with it later or find someone who knows java. 16:11
anyone know java and want to play with gcspy?
darbelo What if I don't know java but still want to play with gcspy? 16:12
cotto there it is 16:15
go for it!
darbelo where?
cotto gcspy?
gcspy is www.cs.kent.ac.uk/projects/gc/gcspy/ 16:16
it's a little odd to get the app started. I'll put up directions on the wiki if it looks like it'll prove useful.
it's definitely got that classy late '90s/early '00s java app feel 16:17
darbelo Hmm, depending on the difficulty of hooking it up to the parrot gc it could be veeeeery useful. 16:18
Still, worst case is writing a new gc. And that's pluggable now, so not that bad. 16:19
cotto afaict it doesn't require any modification of the program being tested. 16:20
darbelo ... only the least possible amount of code is added to the system being visualised ... 16:21
"only the server side of GCspy needs to be adopted for a particular system" 16:22
But from what I can see there the driver's only job is just pushing data to a socket. 16:23
cotto I'm close, but I need to go celebrate International Talk Like A Pirate Day at a friend's house. 16:26
16:26 sri joined
jrtayloriv cotto, You're getting there by boat right? 16:27
darbelo Arrr.
I'll give this another look later. Have to do something else now. 16:32
17:06 JimmyZ joined 17:07 JimmyZ joined 17:12 ruoso joined 17:15 mberends joined 17:23 rhr_ joined
mikehh All tests PASS (pre/post-config, smoke, nqp_test, fulltest) at r41366 - Ubuntu 9.04 i386 17:35
17:39 Andy joined 17:40 mokurai joined, theory joined
mikehh partcl r742 builds on parrot r41366 - make test PASS - Ubuntu 9.04 i386 17:42
17:50 payload joined
mikehh rakudo (2c40a5b) builds on parrot r41366 - make test / make spectest (up to 28320) PASS - Ubuntu 9.04 i386 17:59
18:10 nathanmccauley joined 18:22 Austin joined 18:48 chromatic joined
dalek ose: r109 | Austin++ | trunk (33 files):
Got :multi working. Built a better test framework.
 18:54
18:59 payload joined 19:02 kyle_l5l joined
dalek rrot: r41367 | darbelo++ | trunk/src (2 files):
Replace a bunch of nested, interdependant structures with a single char*.
 19:11
19:23 chromatic joined
kyle_l5l I added an INTVAL to the visit_info struct in pmc_freeze.h, and now I'm getting segfaults everywhere. What step did I forget? 19:31
chromatic Probably realclean.
19:35 kid51 joined
dalek rrot: r41368 | fperrad++ | trunk/tools/install/smoke_languages.pl:
[languages] chdir to the directory of each language
 19:39
kyle_l5l chromatic, still segfaults... 19:42
19:43 darbelo left
kyle_l5l wait 19:43
ok, rakudo needed make realclean too 19:49
dalek rrot: r41369 | fperrad++ | trunk/tools/dev/mk_inno_language.pl:
[languages] add a subdir lang
 20:14
rrot: r41370 | mikehh++ | trunk/src/frame_builder.h:
codetest failure - trailing whitespace
 21:19
chromatic Undef's set_pmc doesn't work. 21:20
mikehh All tests PASS (pre/post-config, smoke, nqp_test, fulltest) at r41370 - Ubuntu 9.04 i386 21:28
chromatic ... probably because calling pmc_reuse_by_class() is very wrong. 21:29
21:29 bacek joined
mikehh partcl r742 builds on parrot r41370 - make test PASS - Ubuntu 9.04 i386 21:32
chromatic: have you run make cover recently? 21:40
chromatic I don't ever run it. 21:44
mikehh ok 21:45
do you do something similar?
chromatic No, not really. I couldn't get gcov working, so I ignored it. 21:46
bacek good morning
purl And good moroning to you, bacek.
bacek mikehh: tapir2.ro.vutbr.cz/cover/cover-results/ 21:47
66.5%
purl 0.665
mikehh i got 69.0% at r41366 on i386 21:49
however I got a lot of intermediate output like the following for example: 21:50
cover: running gcov extend_5.c
File 't/src/extend_5.c'
Lines executed:90.91% of 11
t/src/extend_5.c:creating 'extend_5.c.gcov'
t/src/extend_5.c:cannot open source file
the lasd is what bothers me - I got it a lot 21:51
last
dalek ose: r110 | Austin++ | trunk/ (4 files):
Got :named() parameters generating
 21:53
mikehh needs to investigate further 21:58
however I think I need to go and harrass amd64 - I have been on i386 long enough 21:59
oh forgot to report 22:08
rakudo (2c40a5b) builds on parrot r41370 - make test / make spectest (up to 28320) PASS - Ubuntu 9.04 i386
bbl 22:14
22:15 tetragon joined
chromatic Hey, a 6.327% performance improvement in Rakudo's Hello, world! 22:24
22:26 Andy joined
dalek rrot: r41371 | chromatic++ | trunk/include/parrot/call.h:
[include] Rearranged parrot_runloop_t struct to try to minimize cache misses;
 22:41
rrot: r41372 | chromatic++ | trunk/src/pmc/class.pmc:
[PMC] Moved more logic from default PMC's isa_pmc() into Class's isa_pmc(),

improvements here, but this change improves Rakudo's "Hello, world!" benchmark by 6.327%.
22:42 Andy joined
dalek ose: r111 | Austin++ | trunk/ (3 files):
Got :named:flat arg adverbs working.
 22:47
23:37 quek joined 23:38 mikehh joined
jrtayloriv Man, this channel's pretty quiet today. Is everyone out raping and pillaging? 23:56
dalek ose: r112 | Austin++ | trunk/t/language (4 files):
Moved new-style test files to t/language
ose: r113 | Austin++ | trunk/t/0 (4 files):
Got arithmetic-expr tests running new style.
chromatic I made a performance improvement, and now I'm going to fix an Undef morph to Object bug. 23:57

Powered by Cro and the RakuĀ® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!