Parrot 3.3.0 released | parrot.org | Log: irclog.perlgeek.de/parrot/today” | Accepted GSoC Students announced! | GSoC student information emails coming out soon
Set by moderator on 26 April 2011.
00:01 jrt4 joined 00:04 jrtayloriv left 00:09 sigue left, sigue joined 00:14 nopaste left 00:15 TonyC left 00:17 nopaste joined 00:18 TonyC joined 00:20 jrt4__ joined 00:23 jrt4 left 00:28 sigue left 00:33 varta joined
whiteknight finally got his GCI t-shirt 01:00
apparently I had the wrong zipcode entered
01:03 varta left 01:05 varta joined 01:10 varta left 01:11 varta joined 01:30 whiteknight left, woosley joined 01:31 woosley left 01:32 woosley joined 01:51 plobsing_ joined 01:56 plobsing left 01:57 ShaneC joined
cotto ~~~~~ 02:01
dukeleto .-~v~-. 02:21
cotto That somehow makes me think of Pan's Labyrinth. 02:23
dukeleto it is my ascii art rendition of a facehugger pod 02:25
02:27 theory left
cotto I feel hugged. 02:30
dukeleto cotto: have your LFNW plans in order yet? I am looking up train stuff now. 02:35
dalek rrot: 6a28a16 | petdance++ | src/string/api.c:
consting a STRING opinter
 02:40
cotto dukeleto, I'm thinking I may just drive there. 02:42
I want to be back here by 5 on Sunday and don't want to rely on the train being on-time. 02:43
dukeleto cotto: indeed. and I just found out Bellingham has an airport 02:44
02:51 bubaflub left
soh_cah_toa oh lfnw, why must you be this weekend? 03:00
hmm....lfnw or study? 03:01
cotto soh_cah_toa, where are you?
soh_cah_toa new jersey
cotto that's a bit out of the way
03:01 bacek left
soh_cah_toa washington? i can make the drive 03:01
benabik soh_cah_toa: Washington state, not DC.
soh_cah_toa ugh 03:02
cotto important distinction, that
soh_cah_toa nw.. duh
why did i just think i lived in the northwest?!
benabik If it was in DC, I'd try to show up. I have family down there.
soh_cah_toa oh man, there looks like there's gonna be some killer presentations too 03:03
your both doing presentations!? man, am i mad :( 03:07
cotto soh_cah_toa, yapc::na is closer to your neck of the woods. I'm sure my talk will be more polished by then. 03:08
soh_cah_toa yeah, i was thinking of yapc::na. fosscon too. i'd have to see how that'd work out w/ gsoc
depends how much progress i'm making 03:09
cotto I didn't have fosscon on my radar
atrodo soh_cah_toa> we're having a hackathon on thursday the week of yapc, so it'd be perfect for gsoc
soh_cah_toa hmm...
cotto yeah
atrodo Anyone know off hand the right mailing list to invite the perl6 people to the yapc bof? 03:11
soh_cah_toa what's a bof? 03:13
atrodo birds of a feather. A meeting for people of similar interests 03:14
soh_cah_toa are they organized or just "hey, i know you on irc!"? 03:16
atrodo Both
cotto dukeleto, what do you know about fosscon? 03:17
soh_cah_toa fosscon.org: Fosscon 2011 is a community driven conference for fans of free and open source software(FOSS), being held in Philadelphia, PA on July 23rd, 2011. 03:18
i don't see many speakers yet, though
dalek rrot: 666f58d | petdance++ | src/oo.c:
Now that PMCNULL cannot be NULL, update some function annotations
cotto oh noes! the commit of the beast! 03:19
soh_cah_toa ha! didn't even notice that 03:20
cotto better revert
atrodo too late, git never forgets
soh_cah_toa git literally never forgets. i love it 03:21
benabik I can make git forget. Bwa-hahahah.
Although they keep making it harder. Stupid reflogs. 03:22
(By stupid, I mean awesome.)
dukeleto ~~ 03:23
soh_cah_toa actually, git is stupid
dukeleto cotto: i know little, but I have heard good things about it
cotto dukeleto, ok. still plenty of time to submit a talk
dukeleto atrodo: perl6-compiler ?
cotto atrodo, I'd ask in #perl6 on freenode 03:24
atrodo dukeleto: that was my best guess as well
dukeleto just made a train ticket and booked a hotel for LFNW 03:25
it's on
cotto ~~ donkey kong
atrodo cotto> Yea, that's my next step if I don't get a definitive from someone around here
03:25 jrt4__ left
dukeleto atrodo++ # thanks for organizing a hackathon 03:25
cotto atrodo++ 03:26
I love hackathons.
atrodo I do it for the karma
soh_cah_toa you two, dukeleto and cotto, better save your presentations. i want to see them afterwards
dukeleto atrodo: you can probably send the email to perl6-language as well, the overlap is undefined 03:27
cotto soh_cah_toa, I'm keeping the source on github
atrodo the perl6-compiler list looks to be nearly exclusively bug reports
soh_cah_toa yay
cotto I'll also put the slides online when they're ready. It's a very handy backup trick I'm stealing from pmichaud. 03:28
dukeleto atrodo: some people are only on -language and some are only on -compiler, but most people that would go to YAPC are probably on both. But maybe not :)
03:28 baa joined
atrodo excited about YAPC 03:29
dukeleto atrodo: indeed. It will be my first YAPC
03:33 soh_cah_toa left
benabik I want to go to YAPC, but my initial estimates say that it'll cost a good chunk of my GSoC money to go. 03:34
cotto I've always found it to be a worthwhile experience. 03:37
If I had to pick one conf, it'd be a yapc. 03:38
03:40 bacek joined
dalek rrot: 1051308 | petdance++ | / (5 files):
now that PMCNULL is not NULL, we can set function annotations
 03:44
03:51 hudnix left 03:58 baa left
dalek rrot: c1f77f7 | petdance++ | / (2 files):
now that PMCNULL is not NULL, we can update function annotations
 04:03
rrot: 31daebd | petdance++ | src/library.c:
consting local args
 04:10
rrot: 3456694 | petdance++ | src/pmc/sub.pmc:
consting local var
 04:16
04:50 jrtayloriv joined 05:00 ppant_ joined
cotto anyone remember what state whiteknight lives in? 05:10
05:16 birdwindupbird joined
dukeleto cotto: Pennsylvannia 05:19
cotto: why?
cotto dukeleto, I'm mentioning where Parrot's hackers come from in my talk 05:20
dukeleto cotto: ah.
05:22 birdwindupbird left 05:23 birdwindupbird joined
cotto It's shaping up fairly nicely, though I have trouble seeing it as using 45 minutes efficiently 05:23
I still need to raid the past year+'s #ps reports though 05:26
Hmmm. irclog seems to be down 05:28
05:29 jrt4 joined
cotto nm 05:29
05:32 jrtayloriv left 05:38 jrt4__ joined 05:42 jrt4 left 05:46 jrt4 joined 05:50 jrt4__ left 05:51 fperrad joined 05:56 bacek left 06:11 bacek joined, Coke left 06:32 alin left 06:33 jrt4 left 06:35 Coke joined 06:36 mj41 joined 06:47 mtk left 06:53 mtk joined
cotto atrodo++ for the hackathon stuff 07:14
moritz 's head jerks up at "irclog seems to be down" 07:16
cotto moritz, no worries 07:19
was just slo
w
moritz happens :(
especially when some caches expire
cotto time for sleeps 07:26
'night
07:39 cosimo left 07:47 alin joined 07:48 UltraDM joined 07:49 dodathome joined 07:52 alin left, alin joined 08:19 alin left 08:32 bacek left 08:46 bacek joined 08:47 contingencyplan left 08:53 particle left 08:56 jjore left 08:59 cotto left 09:12 particle joined, cotto joined 09:13 jjore joined
dalek p/ctmo: 7aa35a5 | jonathan++ | src/ops/nqp.ops:
Add a way to take a type object and find out if its representation considers itself as representing a boxed primitive.
 09:19
09:50 birdwindupbird left 09:52 lucian joined, kuku joined 09:54 kuku left, woosley left 09:56 kuku joined 10:04 lucian left 10:07 lucian joined
mikehh sti;; getting occasional intermittent failures with t/pmc/socket_ipv6.t 10:12
still
tadzik nopaste.snit.ch/42225 10:14
oh wait, that's not neceserilly master 10:16
mikehh re-running the failing test nearly always results in a PASS (that is the test in which t/pmc/socket_ipv6.t failed - in this case testb from fulltest) 10:21
All tests PASS (pre/post-config, make corevm/make coretest, smoke (#15662) fulltest) at 3_3_0-57-g3456694 10:26
Ubuntu 11.04 beta i386 (g++)
bah - that was smoke #15622 10:28
tadzik okay, I still get that segfault. Ideas? 10:31
I think it happens during some hardcore file copying
mikehh tadzik: segfaults in parrot is nearly always the result of a gc problem - i.e. some object is collected too early 10:33
all I can suggest is that you have a null pointer in Parrot_Hash_get_pmc_keyed_str, but of course it could be something else entirely 10:39
tadzik that worked fine, must have become broken recently 10:40
mikehh all I can suggest at the moment is trying a bisect - to find where it last worked 10:42
mikehh needs to re-boot and test in amd64 for a bit - bbiab 10:50
10:55 cosimo joined 10:57 mikehh left 11:04 lucian_ joined 11:08 lucian left 11:23 darbelo joined 11:26 rhebus joined 11:31 woosley joined 11:33 cosimo left 11:42 mikehh joined 11:47 ppant_ left 11:55 jsut_ joined 11:57 Patterner left
darbelo clock? 12:00
12:00 jsut left
darbelo aloha: clock? 12:00
aloha darbelo: LAX: Wed, 05:00 PDT / CHI: Wed, 07:00 CDT / NYC: Wed, 08:00 EDT / UTC: Wed, 12:00 UTC / LON: Wed, 13:00 BST / BER: Wed, 14:00 CEST / TOK: Wed, 21:00 JST / SYD: Wed, 22:00 EST
12:05 Psyche^ joined, Psyche^ is now known as Patterner 12:10 mikehh left 12:12 whiteknight joined
whiteknight good morning, #parrot 12:13
seen cgaertner
aloha cgaertner was last seen in #parrot 24 days 21 hours ago saying "btw, who would be willing to mentor the project? the proposal template ask for that information, I I don't remeber anyone actually mentioning that...".
whiteknight seen NotFound 12:21
aloha NotFound was last seen in #parrot 12 days 17 hours ago saying "Unless you do some tricks, like in my expriment.".
12:35 benabik left 12:38 lucian joined 12:42 lucian_ left 12:48 hudnix joined 12:51 lucian_ joined 12:55 lucian left 13:17 bubaflub joined 13:22 mikehh joined 13:48 autark_ left
tadzik hello there 14:12
darbelo o/
whiteknight good morning tadzik
and good morning, darbelo 14:13
darbelo good morning, whiteknight.
14:15 Andy_ joined
mikehh All tests PASS (pre/post-config, make corevm/make coretest, smoke (#15649) fulltest) at 3_3_0-57-g3456694 14:18
Ubuntu 11.04 beta amd64 (g++)
hmmnnn that was with --optimize
tadzik: did you get anywhere with your segfault problem? 14:23
tadzik mikehh: not at all, I have just come back home
(have come, have came?)
mikehh come 14:24
tadzik good, seems that primitive instincts do not lie
mikehh: irclog.perlgeek.de/perl6/2011-04-27#i_3616405 14:28
seems to be gone, it _may_ have been some 'old pirs laying around" 14:29
mikehh Had a friend long ago who's name was Andy Majoriebanks, who sais that there were well over 200 English spellings of his name
Andy_ Andie?
Andee?
Andye?
mikehh not the Andy bit, which is pronounced MarshBanks BTW
Andy_ I'm running out of it.
dalek website: rohit_nsit08++ | Hello Parrot 14:30
website: www.parrot.org/content/hello-parrot
14:40 contingencyplan joined 14:41 hercynium joined 14:58 benabik joined
dalek sella/versions: 2feb00e | Whiteknight++ | src/core/Rosella.winxed:
add a few small comments
 15:08
sella/versions: 7832c28 | Whiteknight++ | s (2 files):
Add in utilities for managing and retrieving version numbers to core. Set the core version to 1 for now
sella/versions: 9417d57 | Whiteknight++ | VERSION:
add a VERSION file, which will contain the definitive versions of all libraries
sella/versions: e1ba9b0 | Whiteknight++ | s (3 files):
setup.winxed builds a code file for setting up version numbers.
sella/versions: b107e17 | Whiteknight++ | .gitignore:
don't show build directory (or, as a bonus, generated html docs from the other branch) in the .gitignore file
sella/versions: b330f75 | Whiteknight++ | s (2 files):
fix build. Add comments to version
benabik ~~
whiteknight good morning, benabik
benabik Mostly good, yes. 15:09
15:14 lucian joined 15:17 UltraDM left 15:19 lucian_ left
dukeleto mornin' 15:20
bubaflub morning dukeleto 15:26
dukeleto bubaflub: how goes it? 15:31
bubaflub dukeleto: not too bad; workin' and finishing up finals stuff - looking forward to being done with this school work. anything new with you? 15:33
benabik grumbles about being on a Quarter system and having close to a month of school left. 15:34
dalek p/ctmo: 0ff3ff4 | jonathan++ | src/ (2 files):
First cut of natively typed attributes (int/num/str). Code gen could be better, but it works.
 15:36
dukeleto bubaflub: trying to get awake 15:37
bubaflub 100cc of coffee, stat!
benabik Most people prefer more than 200cc, unless you want espresso. 15:38
dukeleto downs the coffee 15:39
tadzik grumbles about a semester finishing 5 weeks after the beginning of GSoC 15:40
15:41 lucian_ joined
whiteknight benabik: where do you go to school? 15:41
Andy_ I was hoping my re-annotation of all those functions would clean up some splint warnings
but no.
benabik whiteknight: Rochester Institute of Technology. It'll be switching to semesters in 2013, but I'll be gone by then. 15:42
whiteknight ah right
cotto ~~ 15:43
15:45 lucian left
whiteknight good morning cotto 15:57
cotto hi whiteknight 16:00
16:00 rohit_nsit08 joined
rohit_nsit08 hello #parrot 16:00
Coke: ping
16:05 mj41 left
dukeleto rohit_nsit08: mornin' 16:24
rohit_nsit08 dukeleto: good morning, deciding the name of compiler, will set up the repo on github today 16:25
dukeleto rohit_nsit08: names are hard :)
rohit_nsit08 also wrote my first Post on parrot today
how does "CorellaScript" or just "Corella" sound en.wikipedia.org/wiki/Little_Corella 16:28
tadzik Corella is like Rosella, but sounds nice :)
rohit_nsit08 I like "senegal" too :-)
dukeleto rohit_nsit08: Corella sounds nice 16:30
16:30 woosley left
rohit_nsit08 okay, that's good. I name it "Corella" 16:35
benabik Kneel, javascript to PIR compiler, and rise "Corella".
rohit_nsit08 benabik++ 16:36
whiteknight tadzik: Rosella sounds just fine, thankyouverymuch 16:43
16:43 lucian joined
rohit_nsit08 whiteknight: hello 16:44
16:45 lucian__ joined 16:47 kuku left, lucian_ left
whiteknight hello rohit_nsit08. Excellent blog post today 16:48
16:49 lucian left
rohit_nsit08 whiteknight: thanks, just got it's name also "Corella" en.wikipedia.org/wiki/File:Cacatua_...lia-8c.jpg 16:49
whiteknight I saw that. I like the name 16:50
cotto_work ~~
dalek p/ctmo: cc013b6 | jonathan++ | src/ (5 files):
Setup the very basic KnowHOWAttribute in 6model core rather than leaving every HLL the problem of setting up an initial one. It's especially a pain if you're going to have a compile time meta-model and need the attribute meta-object available to the compiler. (FWIW, nqpclr ended up with this in core too, though with a different initial motivation.)
 16:53
p/ctmo: 2af2457 | jonathan++ | src/NQP/ (2 files):
Start using KnowHOWAttribute from the 6model core instead. Means we now create all meta-attributes during the compile. Toss the knowhow-specific workarounds.
rohit_nsit08 whiteknight: I have forked parrot and it is working on my system.I was thinking to create a new repo for new modified cafe with code generator in that or use the existing fork for that, any suggestions?
sorry it's "cafe" not parrot
whiteknight ruslanspivak.com/2011/02/08/registe...r-tinypie/ 16:54
ok
benabik Github really needs an easier way to only fork a couple branches. After forking parrot, I did a lot of copy/paste to "git push github :branch" all the things I won't be working on. 16:57
whiteknight benabik: if you have branches you don't care about, you can just ignore them 16:58
benabik whiteknight: I like keeping my repositories tidy. :-) 16:59
whiteknight that's fine too. I agree that github would do well to add a "fork only branch X" 17:00
17:01 lucian joined 17:02 bacek left, lucian__ left, bacek joined
dukeleto benabik: sounds like you are created that problem yourself ;) 17:06
17:06 rohit_nsit08 left
dukeleto benabik: can you "fork only some branches" in pure git? clone doesn't take any branch arguments, last time I checked 17:06
s/are created/are creating/
dukeleto obviously needs more coffee
benabik dukeleto: You can, but not with clone. Would have to init, remote add -t branch -t branch2 origin url, fetch, checkout 17:08
17:09 bacek left
dukeleto benabik: sounds like a pain 17:10
benabik dukeleto: Uhm, yes.
dukeleto benabik: but git could easily have a new porcelain command that automates that
benabik: git clone_branches URL branch1 branch2
benabik: Simple Matter of Programming ;)
benabik dukeleto: I'd probably just add the -t option from `git remote add` to clone. 17:11
dukeleto clone_branches is a crappy name, though
benabik: that sounds like less work and more DWIM. I like it.
benabik: we do have a lot of freakin' branches. It is like a graveyard. 17:12
benabik dukeleto: Need to use more forks. :-D
dukeleto benabik: many were migrated from svn, and haven't been touched in years.
benabik dukeleto: Ew. svn--
dukeleto benabik: but I think having history of failed attempts at things are useful
benabik: i migrated parrot from svn-- to git++ 17:13
benabik dukeleto: And many thanks to you for that!
dukeleto++
dukeleto benabik: and converted the test suite and build system to not care about svn--. It tooks months. You're welcome.
benabik dukeleto: We could make a parrot-graveyard repository for inactive branches...
dukeleto cotto_work++ did a lot of the git migration as well, the trac integration part was all his bag
tadzik whiteknight: I don't say it doesn't :)
benabik should probably head to class or something.
dukeleto benabik: that is an interesting idea 17:14
17:15 rhebus_ joined
dalek sella/test_asserter: 379d750 | Whiteknight++ | t/core/Error.t:
the empty test in Error.t is unimplemented
 17:15
sella/test_asserter: fd119ce | Whiteknight++ | t/ (41 files):
use asserter instead of Assert:: in all core tests
17:15 rhebus left, rhebus_ is now known as rhebus
dalek sella/test_asserter: f5aa0b1 | Whiteknight++ | src/test/Assert (2 files):
move functions from Assertions to Asserter.
 17:15
sella/test_asserter: dc00a38 | Whiteknight++ | setup.winxed:
remove Assertions.winxed from the build
sella/test_asserter: 6689693 | Whiteknight++ | src/ (5 files):
factor fail logic out into a separate function, since several places use it which do not have access to an asserter.
sella/test_asserter: b8c8419 | Whiteknight++ | src/test/Assertions.winxed:
remove old assertions file
sella/test_asserter: 872b50b | Whiteknight++ | src/ (5 files):
refactor some of the factory objects to take a hash of options instead of an ever-expanding list of named parameters
sella/test_asserter: bdf6980 | Whiteknight++ | src/test/Test.winxed:
add a quick description of available named parameters
sella/test_asserter: 2291c40 | Whiteknight++ | t/test/ (3 files):
fix tests
sella/versions: 8d12d86 | Whiteknight++ | t/core/Version.t:
+test for Versions
sella/versions: b5cc041 | Whiteknight++ | README.md:
mention VERSION in README
Heuristic branch merge: pushed 23 commits to Rosella by Whiteknight
whiteknight blah. heuristic branch merge fail
17:15 lucian left
benabik really should head to class. 17:16
TTFN, #parrot
17:16 benabik left
cotto_work seen soh_cah_toa 17:16
aloha soh_cah_toa was last seen in #parrot 13 hours 49 mins ago saying "yay".
17:17 rohit_nsit08 joined 17:18 lucian joined
dalek p/ctmo: 28dbe5e | jonathan++ | src/stage0/ (6 files):
Update bootstrap with various attribute changes.
 17:20
p/ctmo: 74ac547 | jonathan++ | / (2 files):
Toss the now-unused KnowHOWAttribute.pm.
p/ctmo: 835b51e | jonathan++ | src/Regex/ (3 files):
Fix lookups of meta-objects in regex library.
17:29 benabik joined 17:33 rhebus left 17:35 rhebus joined 17:54 theory joined 17:56 wagle left
dalek sella: d880d90 | Whiteknight++ | VERSION:
after the last branch merge, test is at version 2
 18:02
18:03 rhebus left 18:07 rhebus joined 18:09 wagle joined, Drossel joined
dalek p/ctmo: 177dadd | jonathan++ | src/HLL/SerializationContextBuilder.pm:
Toss a bunch of legacy package installations that we no longer need (this means that all type lookups are certainly working through the new packages).
 18:10
p/ctmo: 5d65440 | jonathan++ | src/NQP/Actions.pm:
Update some package lookups.
p/ctmo: c408fe7 | jonathan++ | src/Regex/ (3 files):
More legacy package lookup/installation cleanups.
18:10 Kulag left
dalek sella/gh-pages: e2c2821 | Whiteknight++ | .gitignore:
+.gitignore
 18:12
sella/gh-pages: 31d51e3 | Whiteknight++ | libraries/test.md:
update test documentation following the test_asserter merger
18:15 ShaneC left 18:19 ambs joined
cotto_work We should find a theme for parrot.org that makes the blogs easier to ready. 18:21
benabik cotto_work: I use the Google Reader theme. 18:22
whiteknight you know what we could do instead (in the future), set up a custom blog at parrot.github.com. Anybody can post there, all you have to do is open a pull request and a blog admin can pull it in 18:25
18:25 lucian_ joined
whiteknight then we could have any theme we wanted, not ever ever have to worry about permissions, and be able to use the same tools we use now for coding 18:25
dukeleto whiteknight: using gh-pages ? 18:26
whiteknight: sounds spiffy
whiteknight exactly
the reality is that we don't use a hell of a lot of functionality from the drupal repo. We could move the whole website over to github pages 18:27
18:27 rhebus left
whiteknight with the added benefit that the whole thing would be a git repo, so we could easily manipulate it with scripts, etc 18:27
releases become a hell of a lot easier
18:27 rhebus joined
whiteknight at least, that part of the release 18:27
cotto_work That's worth considering. 18:28
18:29 lucian left
whiteknight it would be one thing if we were really using all the features drupal has to offer. But somehow I feel like we aren't 18:30
and the website has several hundred, maybe thousands of user accounts, except none of them have permissions to do anything 18:31
I consider that a bit of a waste, though we could probably lock that down somehow
18:32 ShaneC joined
dalek Heuristic branch merge: pushed 21 commits to nqp by jnthn 18:33
18:34 ShaneC left
dalek p/ctmo: 0a24347 | jonathan++ | src/stage0/ (7 files):
Update bootstrap.
 18:34
18:43 lucian joined 18:47 lucian_ left
whiteknight on the bright side, it's absolutely painless for us to set up a prototype. I can start playing with that now, and if we don't like it, we don't switch to it 18:52
19:07 rhebus left 19:08 rhebus joined
Caelum Andy_: Hi! Did you get my pull request for vim-perl? is it satisfactory? 19:17
Andy_ Caelum: I probably did, and I haven't looked at it. 19:18
But I'm glad you sent it! Thank you.
19:18 rhebus left, rhebus joined
whiteknight Andy_ has been too busy being awesome in the parrot codebase 19:18
Andy_ whiteknight: :-) Thanks. 19:19
Caelum Andy_: ok, I'll ask you later, thanks :)
Andy_ But mostly I've been too busy trying to raise the flag of civility.
benabik Civility? Good luck.
19:19 mj41 joined
Andy_ benabik: Thanks for your support. 19:19
whiteknight It always amazes me when I hear those stories about the perl community, because the perl hackers I interact with tend to be very nice people 19:20
and I don't think the parrot community would tolerate any of that nonsense
at least, I hope we wouldn't
benabik Andy_: I do wish you well, I just don't like people sometimes.
Andy_ benabik: I don't, either, but civility is still possible.
19:28 lucian_ joined 19:32 lucian left 19:39 benabik left 19:47 lucian_ left
Caelum I haven't played with parrot in years, does it work on cygwin? 19:56
whiteknight that's a very good question. We frequently have many problems on cygwin
I don't know the current status 19:57
Caelum sounds interesting, I'll take a look :) 19:58
whiteknight Caelum: thanks! We never have enough people looking at cygwin so it falls into disrepair. Another set of eyes is always welcomed 20:02
20:17 rhebus_ joined, rhebus left, rhebus_ is now known as rhebus
dukeleto Caelum: it should build, but we might have some failing tests on that platform 20:19
20:21 lucian joined
dalek rrot/tt1931-nci-parameters-deprecation: 8c6fcbf | mikehh++ | MANIFEST:
re-generate MANIFEST
 20:29
20:30 whiteknight left
dukeleto crafts a gsoc mentor email 20:43
20:44 lucian_ joined 20:46 lucian left 20:49 lucian joined 20:54 lucian_ left
mikehh tt1931-nci-parameters-deprecation branch: 21:00
All tests PASS (pre/post-config, make corevm/make coretest, smoke (#15666) fulltest) at 3_3_0-97-g8c6fcbf
Ubuntu 11.04 beta amd64 (g++)
coke_ rohit_nsit08: are you on mumbai time? 21:02
... (if so, I expect I won't see an answer to this until tomorrow. ;)
msg rohit_nsit08 I'll carve out some time tomorrow (4/28) so we can chat. 21:03
aloha OK. I'll deliver the message.
21:05 lucian left
dukeleto mikehh: that branch is sensitive to config flags 21:11
mikehh: such as --without-libffi and some of the NCI-related config flags
21:12 bubaflub left
dukeleto mikehh: for instance, it fails with this: github.com/leto/Util/blob/master/b...rrot_plain 21:13
21:17 fperrad left
pmichaud good afternoon, #parrot 21:18
dukeleto pmichaud: howdy
cotto_work hi pmichaud
21:19 dodathome left
pmichaud I'm running spectests for the new Rakudo Star release and I'm noticing it's not significantly faster than the release we did in January.. is this expected? (more) 21:19
I thought that the new GC was supposed to significantly speed things up.
cotto_work pmichaud: Rakudo's been using the gms GC for a while. 21:20
plobsing_ not since January though, gms didn't land until what? march?
cotto_work grabs a git shovel 21:21
pmichaud Rakudo started using the gms GC in March 21:22
(March release)
the gms GC didn't exist in january
21:23 hercynium left
plobsing_ I expect gms to be faster than ms2. pmichaud: can you profile to see if the GCs are taking roughly equivalent time? 21:24
21:24 mj41 left
plobsing_ I suspect we may have accidental regression elsewhere that is eating up the perf win. 21:24
pmichaud gms is faster than ms2, yes.
but ms2 today is about 40% slower than it was in january
so rakudo+gms in april 2011 is only as fast as rakudo+ms was in january 2011 21:25
i.e., no speed improvement
plobsing_ so where is the loss?
pmichaud I don't know yet.
plobsing_ have you added significant new features?
cotto_work ipfy would be handy here
pmichaud no, we haven't made significant changes afaik 21:26
that's what surprises me a bit
plobsing_ ipfy would be usefull if its timeline weren't so short as to hide all the signal in noise
need a low-pass filter on that thing
cotto_work atrodo: ^
jnthn__ Agree there hasn't been any big changes of late. 21:27
Individual features maybe (like the INET stuff got re-done)
pmichaud the 2011.02 release of rakudo (using Parrot's default GC) is about 33% slower than the 2011.01 release 21:28
cotto_work pmichaud: so we should test a jan Rakudo + jan Parrot, current Rakudo + current ms2 Parrot, current Rakudo + current gms Parrot?
and profile to see what changed
also, what's your benchmark?
pmichaud time needed to run "make spectest"
cotto_work I was hoping you wouldn't say that 21:29
pmichaud well, that's what was mentioned in some of the posts about the new GC :)
cotto_work Sure. It's just long to test and impractical to profile. I agree that it's a good yardstick where it's feasible. 21:30
pmichaud right
anyway, on my machine:
jnthn__ pmichaud: Try running tools/benchmark.pl 21:31
pmichaud 2011.01 rakudo with 3.0.0 parrot: 29m45s
jnthn__ pmichaud: It cares a lot about invocation speed, for example.
cotto_work I wish my machine were that fast.
21:31 ambs left
pmichaud 2011.02 rakudo with 3,1,0 parrot: 42m12s 21:32
2011.03 rakudo with 3.2.0 parrot/gms: 30m15s
2011.04 rakudo with 3.3.0 parrot/gms: 28m59s
2011.04 rakudo with 3.4.0 parrot/ms2: 51+m
s/3.4.0/3.3.0
plobsing_ if we could get 1 (or a small handful) of the spectests whose performance deteriorated, that would be usable. would it be possible to have the test framework maintain timing info so we can compare and find the worst offenders? 21:33
cotto_work That's instructive.
plobsing_: I was thinking something similar.
plobsing_ also, how is rakudo startup? I know spectest is very dependant on that. 21:34
and that is dead easy to profile
pmichaud about 9 minutes of the spectest time is rakudo startup. 21:35
plobsing_ is that across the board? 21:36
21:36 rohit_nsit08 left, particle1 joined
pmichaud just a sec, I'll give a more accurate figure 21:36
but yes, across the board, about 9 minutes 21:37
2011.01 startup time: 1.413s 21:38
2011.04 startup time: 1.181s
21:39 lucian joined
pmichaud so, Rakudo startup in 2011.04 likely accounts for 648s, in 2011.01 accounts for 764s 21:39
21:39 particle left
cotto_work That much is encouraging. 21:39
pmichaud so, Rakudo startup improved by a lot between 2011.01 and 2011.04, but we lost performance on actual runtime 21:40
cotto_work but it means we're losing a lot of time elsewhere
21:40 theory left 21:41 rhebus left
pmichaud anyway, I was just curious, because I had somehow expected the new GC to improve things more. Oh well. 21:43
cotto_work pmichaud: we'll be looking into it.
21:44 rhebus joined
cotto_work thanks for mentioning it 21:44
dukeleto pmichaud: is there a graph of spectest times anywhere on the web? I think that visualization would help. 21:45
pmichaud dukeleto: no, not I know of. I'll probably start keeping a file or spreadsheet or something with timings though 21:46
in the general case it often becomes apples-to-oranges because the size and content of spectest changes over time also
dukeleto pmichaud: is isparrotfastyet.com/ relevant ?
pmichaud: are you seeing things similar to that?
21:47 rohit_nsit08 joined
pmichaud dukeleto: I'm just comparing the time needed to run spectest from 2011.01 to present 21:47
dukeleto pmichaud: that says that the fastest rakudo was 14 commits past 3.3.0
pmichaud: if you give atrodo++ a script for doing a spectest he can probably add it to isparrotfastyet.com 21:48
pmichaud: just in-case you want certain debug flags set and such 21:49
pmichaud dukeleto: it's just "make spectest"
probably want to avoid the git fetch time, though, so do "make t/spec" followed by "make spectest"
dukeleto pmichaud: yes, except when it's not. I have many scripts to run parrot with various config/debug flags
pmichaud so what does the script need in it? Can I see an example script? 21:50
some how "give a script for ..." isn't very descriptive. :-)
dukeleto pmichaud: what is your equivalent to github.com/leto/Util/blob/master/bin/new_parrot which configures, builds and does a spec test? 21:53
pmichaud dukeleto: how do we denote when the timing should begin or end? 21:54
dukeleto pmichaud: i can write it, but I asked you because I figured you knew something I didn't ;)
pmichaud: hmm. for now, the entire script will be timed
pmichaud that would seem to time the cost of building rakudo (and parrot) then 21:55
not the time of the spectest itself
dukeleto pmichaud: sure. write whatever you want to get timed in the script
pmichaud I only want the "make spectest" to be timed.
dukeleto pmichaud: ok.
pmichaud ...do I assume an up-to-date git checkout of Rakudo? 21:56
or does the script do that?
dukeleto pmichaud: i am going to put it in a jitterbug instance, which can run a script on every commit it gets
pmichaud: more exactly, from each github post-receive notification it gets
pmichaud you want this on the parrot commits, then? 21:57
dukeleto pmichaud: for example: jitterbug.sgn.cornell.edu:8090/ (my personal instance with parrot is currently down)
pmichaud I don't quite understand the process here. 21:58
dukeleto pmichaud: well, I can hold the parrot version constant, or I could run each commit under multiple versions of parrot
pmichaud you're asking me to write a script to run rakudo spectests
dukeleto pmichaud: no, you don't have to do anything. I was trying to ask you for information that didn't exist, and failed. No worries.
pmichaud the only thing rakudo does currently is send the --gc=gms and --optimize flags to Parrot's configure 21:59
*if* rakudo is being asked to build a Parrot
dukeleto pmichaud: just notifying you. Jitterbug is a things that eats github post-receive hook JSON and spits out a pretty interface with PASS/FAIL stats for commits and spews email as well
s/is a things/is a thing/
pmichaud: it also times the build of each commit
pmichaud: and it would be very simple to generate a graph of the build times 22:00
pmichaud: so the thing that you want for rakudo is a feature that I want to add to Jitterbug :)
dalek p/ctmo: 6435cc3 | jonathan++ | src/ (3 files):
Simply multi method construction and dispatcher testing.
p/ctmo: 245026d | jonathan++ | src/ (2 files):
Shuffle method-adding code into the SC builder.
p/ctmo: fef0b59 | jonathan++ | src/HLL/SerializationContextBuilder.pm:
First steps at getting method definitions to go through the compile time meta-model. Some regressions.
dukeleto pmichaud: sorry to make that so complicated.
pmichaud: just got my personal instance working again : new.leto.net:3000/project/parrot 22:02
pmichaud: so you would be interested in "make spectest" for fixed Rakudo version and across multiple parrot versions ? 22:05
pmichaud: the time for "make spectest" rather
pmichaud not exactly "fixed Rakudo version"
I'm interested in the time for "make spectest" for Rakudo releases
each Rakudo release is tied to a Parrot release 22:06
dukeleto pmichaud: ok, so both versions will be increasing
pmichaud: yes, gotcha
pmichaud if you want to do "fixed Rakudo", that's fine with me, but it's nearly impossible to have a fixed version of Rakudo that works across Parrots
dukeleto pmichaud: sure, that isn't what is wanted, anyway
pmichaud for example, iirc Rakudo 2011.01 won't necessarily work with later Parrots, because of changes to some of the internals.
and I'm quite certain that Rakudo 2011.04 won't work with Parrot 3.0.0 22:07
dukeleto pmichaud: i will always use the --gen-parrot flag so Rakudo uses whatever parrot it wants. That is what you want timed.
pmichaud yes, that's a good start
it doesn't always tell us where something significant changed, though, because a lot of Parrot commits can occur between what --gen-parrot uses 22:08
dukeleto pmichaud: i've just added various features to jitterbug recently that makes this possible, just as custom build commands and reusing git repos so they don't get recloned a million times
pmichaud: sure. but I think it will at least be a canary if I can get this to work
22:09 GodFather joined, GodFather left 22:10 soh_cah_toa joined
dukeleto pmichaud: so what i need is to add a post-receive hook to the rakudo github repo. i could fork and add it to my repo, and then mirror to my fork, as an alternate solution 22:10
pmichaud the parrot revisions Rakudo is using are: 3_3_0, 3_2_0-406, 3_2_0, 3_1_0-700, 3_1_0-327, 3_1_0-184, 3_1_0
dukeleto pmichaud: the URL new.leto.net:3000/hook/ needs to be added as a Post-Receive URL
pmichaud: exactly as parrot.git does at github.com/parrot/parrot/admin/ 22:11
pmichaud Rakudo releases only uses 3_3_0, 3_2_0, 3_1_0
dukeleto pmichaud: jitterbug will generate stats for each push to the repo, but I have scrips which allow you to run a script on a range of git sha1's as another way to skin that cat 22:12
pmichaud want me to add the Post-Receive URL now? 22:13
dukeleto pmichaud: do you want to add the post-receive hook to rakudo/rakudo/admin now? yes, please :)
pmichaud: sure, I am in the mood to get this done right now. 22:14
pmichaud done.
22:14 bacek joined
dukeleto pmichaud: by default, it will time "make test", and then I just need to tweak the conf to make it test "make spectest" 22:14
pmichaud++
22:14 rhebus left
pmichaud if you can, make sure it does a "make t/spec" first, or otherwise makes sure that the t/spec git repo is up-to-date 22:14
otherwise you'll end up timing the "git pull" that happens there also. 22:15
(it's probably insignificant to the overall time anyway :)
dukeleto pmichaud: do i need to do make t/spec folowed by make spectest ? t/spec just updates the spec tests, correct? 22:18
pmichaud: i am concocting a script now
pmichaud dukeleto: yes, that's what I do
typically: make t/spec; time make spectest
actually: make; make t/spec; time make spectest
completely: make realclean; perl Configure.pl --gen-parrot; make; make t/spec; time make spectest 22:19
soh_cah_toa hey, i don't mean to interrupt but what's this "automatic alias" on the blog entry page?
dukeleto soh_cah_toa: no clue
pmichaud blog entry page?
soh_cah_toa on parrot.org
dukeleto pmichaud: that was the script I was asking for earlier ;)
pmichaud: do you want any parrot flags by default, such as GC or debugging ?
pmichaud Rakudo's configure.pl provides them 22:20
22:20 lucian left
dukeleto pmichaud: groovy 22:20
pmichaud more specifically: --gen-parrot provides the Parrot configure flags that Rakudo thinks are best
(since March, these flags are --gc=gms and --optimize) 22:21
(prior to that, just --optimize)
afk, dinner
dukeleto Andy_: ping 22:30
msg Andy new.leto.net:3000/project/parrot shows your recent commit "shimming up some interps" 0aaee88 fails to build 22:31
aloha OK. I'll deliver the message.
22:32 whiteknight joined
Andy_ i'm here 22:32
whiteknight good evening, #parrot 22:33
dukeleto Andy_: hola
Andy_ ok 22:34
dukeleto Andy_: new.leto.net:3000/api/build/parrot/...rl-v5.10.1 shows an error at the very end of the file
Andy_ let's see your errors, dukeleto
looking
dukeleto structview is also spewing an insane number of warnings 22:35
dukeleto doesn't like that
whiteknight maybe they aren't warnings. maybe the compiler is listing all the things it likes
Andy_ what does your declaration of latin1_upcase_first look like? 22:36
22:36 rohit_nsit08 left
Andy_ Mine looks like latin1_upcase_first(PARROT_INTERP, ARGIN(const STRING *src)) 22:36
dukeleto Andy_: where do I find that ?
Andy_: i need to go digging for it
Andy_ I don't even know where this commit 0aaee88 is.
src/encodsrc/string/encoding/latin1.c
dukeleto Andy_: looks to be the same 22:39
Andy_ it's PARROT_INTERP, not SHIM_INTERP?
what branch are you on?
dukeleto Andy_: this commit on master github.com/parrot/parrot/commit/0a...3b7ccddd12 22:40
Andy_: yes, it is PARROT_INTERP
Andy_ That's about 6 weeks old.
dukeleto Andy_: lulz. Sorry. I just revived my jitterbug instance, and I think there was a lingering task in the database that just got ran 22:41
Andy_ That commit does have a SHIM_INTERP (incorrectly) in the latin1_upcase_first
ok.
dukeleto Andy_: sorry to bother you. I thought that was a recent commit
Andy_: should have looked at the date 22:42
Andy_ I will never forgive this grievous offense of being human.
dukeleto OH THE HUGE MANATEE!
dukeleto goes back to a dark cave to self-flagellate
Andy_ Hey, hey, that's your own business,pal.
22:46 Andy_ left, davidfetter joined
davidfetter hai dukeleto 22:46
22:51 benabik joined
dalek p/ctmo: 6ca730b | jonathan++ | src/ (3 files):
Partial fix for multi-method dispatch regression.
 22:53
p/ctmo: 5b9b016 | jonathan++ | src/NQP/Actions.pm:
Get adding of regex declarations done through the compile time meta-object. Well, mostly...there's a horrible issue with the way prefixes are stored to deal with first...
dukeleto davidfetter: hola 22:56
davidfetter dukeleto, so about the stuff you ran across while perusing plparrot.c... 23:00
dukeleto davidfetter: ... 23:02
23:03 kid51 joined 23:12 theory joined 23:14 theory_ joined 23:18 theory left, theory_ is now known as theory
cotto_work github.com/cotto/cotto-talks/blob/...tNW.slides 23:30
still a wip; thoughts welcome 23:31
23:38 cosimo joined 23:49 benabik left 23:53 bubaflub joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!