Parrot 0.8.2 Feliz Loro Released | parrot.org | 45 TT | 538 RT
Set by moderator on 17 December 2008.
00:05 ruoso joined 00:09 AndyA joined 00:10 tetragon joined
Whiteknight jonathan++ 00:15
00:22 AndyA joined
jonathan Whiteknight: Na? Ale d'akujem... 00:26
oh
For? But thanks..
jonathan is writing email in different language and gets confused at this time of night... 00:27
Whiteknight I love reading your Rakudo day blog posts 00:36
(Whiteknight's happiness)+
00:40 AndyA joined
peters sleep & 00:48
diakopter Infinoid: do you know what happened to feather's SVN::Web? 00:53
(that made it broken since a while before now)
jonathan loves online dictionaries. Especially ones that claim that the Slovak for "hybrid transformer" is exactly the same word they use for "crotch"... 00:55
diakopter hey, those terms are similar in english, too, I thought...? 00:56
jonathan I've never called it that! 00:57
diakopter that's what she said.
apache2: /tmp/buildd/subversion-1.5.1dfsg1/subversion/libsvn_subr/path.c:373: svn_path_dirname: Assertion `is_canonical(path, len)' failed. 00:58
oh; that's what feather gets for being sid. 00:59
jonathan reverts back to English mode 01:04
diakopter feather's (and sid's) subversion is r34233; this bug is fixed in r34233
oops
feather's (and sid's) subversion is r32289 01:05
jonathan Ah. :-|
It's an svn bug?
diakopter seems likely 01:06
jonathan :-|
Whiteknight i'll show you my hybrid transformer
diakopter based on [the very few] google search results of that error msg
jonathan Whiteknight: Erm. Please no.
jonathan looks away just in case
diakopter ponders Whiteknight's namesake/etymology. 01:07
Whiteknight somebody once asked me if it was a KKK reference 01:09
(it's not)
Whiteknight wonders if you can fly a diakopter?
diakopter if I can or if one can? I can't fly myself, afaict. 01:10
01:23 ruoso joined 01:25 geof joined
Infinoid diakopter: the upgrade to subversion 1.5 changed some APIs, specifically, the format of the path arguments got a bit stricter 01:31
you will see assertion failures related to that all over apache's error.log. there are a couple patches floating around cpan RT to work around it in various ways 01:32
diakopter: (done backlogging now) yeah, that.
Coke trac.parrot.org is generating warnings. 01:38
jonathan There's a Parrot Virtual Machine Workshop 2009? 01:40
jonathan looks
Coke wonders what svn status of " S" is. 01:49
jonathan didn't see that one before 01:50
chromatic File affected by svn switch, if its parent isn't switched. 01:51
cd compilers/imcc
svn switch some_tag_elsewhere
cd -
svn status -q
Coke looks like an svn switch I tried completely fubared my parrot checkout. 01:52
Coke wipes and does a fresh checkout.
Coke opens a bug with OSU regarding our trac install. 01:53
s/U/L/
OSL?
purl?
purl Coke?
Coke I am amused that OSL's issue tracking system is RT. 01:54
and yet we're having them run trac for us.
OSL is osuosl.org/
jonathan The sweet, sweet smell of irony.
Coke it smells a lot like bacon.
jonathan wishes people would talk less about bacon on this channel today :-| 01:55
I so want some now.
diakopter that's what she said.
chromatic She wants bacon?
diakopter I don't know what she meant. 01:56
needless to say, she was disappointed.
jonathan Hmmm. Be sensible and sleep OR have another beer and finish this letter...
pmichaud "Today's episode is brought to you by beer and the letter 'O'". 01:57
jonathan Fine, beer...
pmichaud: How's the initload stuff going?
pmichaud just about to get started... paula got back earlier and said "I'm too tired to cook" so we went out to eat :-) 01:58
jonathan Nice :-)
I've just spent the time catching up on personal emails.
pmichaud I need to do that also. My inbox runneth over. 01:59
jonathan They pile up after a while... :-S
The hardest bit is when you've got one to reply to in your not-first language.
pmichaud particle: ping
jonathan At 2am.
jonathan dreams of being fluent some day... 02:00
pmichaud beer should help with that :-)
(replying at 2am)
jonathan I'm sure she'll be able to decode my attempt at Slovak somehow. :-)
I'm pondering taking a copy of the synopses with me on the plane/train etc over the weekend's travelling and starting on getting S14 together. 02:03
particle pmichaud: pong 02:04
pmichaud (in privmsg) 02:05
Coke rants the googlecode commit messages are being marked as spam when going to google groups. 02:10
beerios.
pmichaud: ping 02:13
pmichaud: nevermind.
mugwump commit messages, how quaint
pmichaud pong.
particle oh, here comes the peanut gallery from down under 02:18
:P
Coke ? 02:19
mugwump yeah we'll have a few more people around later putting rakudo to the test
particle is it welly.pm night?
pmichaud test? test?
pmichaud didn't study for an exam. :-| 02:20
particle don't worry, it's a blood test.
Whiteknight: i need to apologize about messing up your release procedures.
little did i expect that the osuosl admins would pick the precise moment of our release to do trac admin work, i didn't see that coming when i requested it. 02:21
mugwump builds 0.8.2
particle turns out murphy was right.
02:21 TiMBuS joined
Whiteknight it's okay, I'm very adaptable 02:22
plus, I was planning for something to go wrong, would have been disappointed otherwise :)
Coke hurm, pmichaud, reping.
pmichaud repong.
Coke what is the current right way to handle exceptions?
pmichaud in Parrot?
purl in Parrot is probably false.
Coke I think that's the root cause of trac.parrot.org/parrot/ticket/48 02:23
pmichaud checks.
Coke purl, forget in parrot
purl Coke: I forgot in parrot
particle Whiteknight: i'll think of something else for the 1.0 release. something *truly* special >:) 02:24
pmichaud coke: to set an exception handler: push_eh
actually, push_eh label
chromatic Oh look, the same exception gets thrown multiple times.
pmichaud to remove the exception handler, pop_eh
removing the exception handler is also needed after catching the exception, assuming you no longer wish to catch it. 02:25
the "multiple exception handlers in backtrace" occurs (I think) when an exception handler is triggered multiple times in the same sub.
Coke ah. so the first line of each "normal" handler should be "pop_eh" ?
pmichaud first line of a handler is usually .get_results, if any. After that, pop_eh. 02:26
Coke danke.
pmichaud (multi exception handlers in backtrace) -- because if an exception handler isn't cleared or resumed, then it stays in the context "stack" until return exceptions are invoked.
mugwump so is 'make install' with parrot still a Bad Ideaā„¢ ? 02:27
pmichaud at least, I _think_ that might be what is going on -- I haven't looked deeply into the code to see if it short-circuits this somehow.
mugwump: yes, if you think you might want to do development
chromatic That's plausible, but ... hmm.
pmichaud the build version gets confused with the install version 02:28
mugwump well I just want to give people logins to this vserver to play with it
jonathan Not installing is probably safer. 02:30
mugwump I'll put it in a special prefix I think 02:31
eg /utsl/parrot
particle +1
purl 1
nopaste "pmichaud" at 72.181.176.220 pasted "example of exception handlers not being cleared" (183 lines) at nopaste.snit.ch/15007 02:34
pmichaud this is adapted from RT #59778
chromatic Oh, I really don't want to debug where the longjmp should go.
pmichaud what I _think_ happens is that we enter 'foo' the second time, put an exception handler on the handler stack, attempt to create a class 02:35
that throws an exception from C
which invokes the handler
Coke (&#$@#
jonathan This stack trace sounds...familiear...
pmichaud which creates a new continuation record
Coke thankfully, I have a .catch() macro where I can, for the most part, put this once.
that avoids TT #48. 02:36
pmichaud++
pmichaud that new continuation record then does the pop_eh and returns from foo
however, I don't know when the old continuation record gets collected.
i.e., the one that was in effect at the time we did the newclass op that triggered the exception in the first place 02:37
jonathan OK folks, sleep for me...night. 02:44
pmichaud good night. 02:45
Whiteknight me too. Good night all! 02:50
Coke (*&#$. 02:52
I have a bit of code that does BOTH push_eh foo and goto foo 02:53
and I've just been doing updates that ignored that fact, so I'm going to have to revert and go through again. :|
mugwump ffs, svn log svn.perl.org/perl6/pugs/trunk/t/spec returns an error 02:54
oh, have I got the path wrong
ah, svn.pugscode.org/pugs/t/spec/ 02:55
oh ... that's right, there were no branches in the pugs repo 02:56
Debolaz The joy of subversion. 03:02
mugwump well, maybe I should put this git-svn clone of the spectests on github
gist.github.com/37358 # perl6.pbc no worky for me :( 03:07
was there a version that wasn't incremented somewhere?
ok. calling it as ../../parrot worked 03:08
chromatic Whew. 03:09
mugwump tries rm -rf /utsl/parrot; make install to check it wasn't cruft 03:11
Coke partcl's interactive mode is borked. 03:13
Coke manages to bork another svn switch. 03:14
switching from the 0.8.1 tag to trunk makes svn unhappy.
mugwump yeah, this fresh 0.8.2 build thinks it's 0.8.1 03:28
chromatic I didn't 'make release' from a branch. Maybe I should have. 03:29
mugwump ok, with a fresh 'make install', I didn't need the ../../parrot 03:30
it would have been quite an old parrot install there beforehand. perhaps from when I came to your place chromatic :)
chromatic 0.3.x or something. 03:31
mugwump yeah
Coke chromatic: you should have made it from a tag, yes. 03:39
trunk didn't get the changes back until after the release.
is this the CPAN version?
or the parrot.org version? 03:40
(hopefully the latter.)
chromatic Both; same tarball.
Coke cries.
is a 0.8.2.1 in order? 03:41
chromatic Depends if what's in the CPAN tarball is missing anything we care about.
Coke If it's reporting 0.8.1, yes. 03:43
(since that means it's also missing any updates that occurred in the short lived release branch.
Coke wishes for a simple "what version was this file removed in" svn command. 03:46
(svn-bisect being a bit much for the entire branches hierarchy) 03:47
looks like r34005 contains the only commit in that branch. 03:50
Tene Coke: what file? 03:54
git log --diff-filter=D -- path/to/file 03:55
Coke tene: what what file? 03:57
mugwump sorry, was away sorting out nibbles for rakudo testfest :)
03:57 davidfetter joined
Coke 22:50 <@Coke> looks like r34005 contains the only commit in that branch. 03:57
Tene 20:46 * Coke wishes for a simple "what version was this file removed in" svn command 03:58
mugwump Coke: it was some tarball I got from ... ftp://ftp.parrot.org/pub/parrot/releases/devel/0.8.2/parrot-0.8.2.tar.gz
Coke found the version.
04:02 elmex_ joined
diakopter waits for dalek to start spewing 04:34
Coke hurm. parrotio had 'say' but filehandle doesn't? 04:38
04:45 vaidyagi joined
Coke wonders if trac is incredibly slow, or just coke. 04:45
msg pmichaud: it should be possible to update report #3; I could do it if it were oracle and not sqllite. 04:53
purl Message for pmichaud stored.
05:09 dalek joined
diakopter waits for someone to commit something 05:11
committers do it early and often.
Coke is committing... just to partcl. 05:13
"it's kind of serious"
05:13 mberends joined
Coke nearly has 'make test' working against trunk in both repos. whee. 05:14
diakopter waits for 34059 to appear in the rss
Coke it's midnight? oy.
diakopter Coke... 34059? 05:16
Coke must I? hang on.
diakopter hangs
mberends rakudo: my @a; my %h=('k1'=>undef); @a.push(%h); say @a.perl; 05:20
Coke there you go.
polyglotbot OUTPUT[[("k1" => undef)]ā¤]
Coke will hold off on re-basing partcl stable against parrot 0.8.2 until the tarball issue is resolved. 05:21
mberends rakudo: my @a; my %h=('k1'=>undef); @a.push(%h); @a[-1]<k1>='ok'; say @a.perl; # worked in r33839, dies in r34056
polyglotbot OUTPUT[get_pmc_keyed() not implemented in class 'Perl6Pair'ā¤current instr.: 'parrot;Associative;postcircumfix:{ }' pc 1593 (src/classes/Associative.pir:40)ā¤called from Sub '_block11' pc 140 (EVAL_12:42)ā¤called from Sub 'parrot;PCT;HLLCompiler;eval' pc 892 (src/PCT/HLLCompiler.pir:508)ā¤called from
..Sub 'parrot;PCT;HLLCompiler;evalfiles' pc 1216 (s...
pmichaud @a[-1] is invalid. 05:23
it's now @a[*-1] (and unimplemented at the moment)
diakopter Coke: ergh
05:23 Hadi joined, Hadi left
diakopter Infinoid: <sigh /> 05:24
mberends pmichaud, any workaround for @a[*-1]? the @a[-1] was ok in r33839 05:25
pmichaud mberends: I might be able to temporarily re-enable the [-1] form.
but it's not legal Perl 6.
rakudo: my @a = <a b c d>; say @a[-1]; 05:26
polyglotbot OUTPUT[dā¤]
pmichaud hmmmm.
diakopter dalek: be
pmichaud okay, maybe that's not it then.
rakudo: my @a; my %h = (k1 => undef); @a.push(%h); say @a.perl;
polyglotbot OUTPUT[[("k1" => undef)]ā¤]
pmichaud rakudo: my @a; my %h = (k1 => undef); @a.push(%h); say @a[0].perl; 05:27
polyglotbot OUTPUT[("k1" => undef)ā¤]
pmichaud rakudo: my @a; my %h = (k1 => undef); @a.push(%h); say @a[-1].perl;
polyglotbot OUTPUT[("k1" => undef)ā¤]
pmichaud ohhhhhh
I know what's happening. :-)
(checking S29)
mberends :-)
pmichaud .push on an array pushes values in list context 05:28
a hash in list context becomes a list of pairs
thus @a contains a single pair
(not a hash)
so the <k1> fails because the Pair doesn't know how to do postcircumfix:<{ }> 05:29
you can either push a reference to a hash, or you can treat @a as a hash
rakudo: my @a; my %h = (k1 => undef); @a.push(\\%h); say @a[-1]<k1>='ok'; say @a.perl; 05:30
polyglotbot OUTPUT[okā¤[{"k1" => "ok"}]ā¤]
pmichaud that's probably closest to what you want.
mberends pmichaud++ # push hash reference. Funny that my sloppy code got away without \\ 05:31
pmichaud before this week, hashes weren't interpolating in list context
so it really was treating %h as a single element. 05:32
mberends it's essential to the Pod6 Parser and emitters I'm still working on. 05:33
pmichaud okay. The [-1] will have to change to *-1 at some point, but not quite yet.
mberends no rush, it will be an easy correction when the time comes. 05:34
05:34 dalek joined
Coke starts up a new run of the tcl spec test. 05:36
MagnusShortwave :O 05:46
mugwump throws a golfball in
05:47 smrf joined 05:48 dalek joined
dalek r34057 | pmichaud++ | : 05:49
: [pct]: 'evalfiles' should pass command line arguments to 'eval'.
05:49 allison joined
dalek diff: www.parrotvm.org/svn/parrot/revision?rev=34057 05:49
r34058 | duff++ | :
: [ecmascript] Added relational ops <, >, <=, >=, and some tests
diff: www.parrotvm.org/svn/parrot/revision?rev=34058
r34059 | coke++ | :
: [codingstd] remove trailing spaces
diff: www.parrotvm.org/svn/parrot/revision?rev=34059
diakopter welcomes dalek back to the land of the living. or as living as irc is. 05:50
Infinoid: thanks for your help again..
Infinoid dalek is a heck of a lot more alive in here than I usually am :)
mugwump so, we're trying to figure out how to run individual tests using the perl6 binary 05:52
smrf found PERL6LIB
pmichaud make t/spec/path/to/test.t
mugwump smrf: MagnusShortwave: from /utsl/build/parrot-0.8.2/languages/perl6
05:56 petdance joined
Coke individual tests or test files? 05:58
mugwump well, individual files
pmichaud the "make" approach above is simplest.
mugwump yeah. what about PERL6LIB?
pmichaud it's just a list of directories to search for modules via 'use' or 'require' 05:59
mugwump right, so on this system I can set it to PERL6LIB=/utsl/build/parrot-0.8.2/languages/perl6 06:00
because the only library seems to be Test.pm
pmichaud yes. and hopefully someday soon even that won't be a library :-)
mugwump what will it be?
pmichaud testing functions are likely to become "builtins" 06:01
mugwump keeping with Perl as a kitchen sink language? ;)
pmichaud we'll still allow test modules, but some of the common test functions will be loaded by default.
smrf pmichaud: because you think it'll be easy to unify the Test:: namespace on CPAN?
mugwump <== channelling heckling
pmichaud oh, I suspect there will still be conflicts over Test:: 06:02
but at least the basics will be there (at least until someone writes a Test::* module that overloads them with something else :-)
mugwump I just don't understand why you'd want that..
pmichaud want what?
purl well, want is often mispronounced as 'need' or broken by design, but at least it's clever
mugwump what's wrong with writing 'use Test' at the top of test scripts?
pmichaud I think the sense was that testing out to be more integral than just test scripts. 06:03
s/out/ought/
i.e., that one could use tests as a form of (easily disabled) debugging, too.
we're also looking at building testing directly into the operators
diakopter purl, well? 06:04
purl Three holes in the ground!
diakopter purl, well?
purl Three holes in the ground!
pmichaud anyway, that's language design, and I tend to focus on implementation. :-)
mugwump right
smrf fair enough
mugwump the way I look at it, if you can't write those things as extensions it's broken
pmichaud we certainly _can_ 06:05
but in some sense testing shouldn't be "an extension"
mugwump except, I hate typing system. or os. all the time
pmichaud switching machines... brb
smrf hey, we could turn into php, with 1000+ entries in the default namespace :)
dalek r34060 | pmichaud++ | : 06:06
: [rakudo]: change '$def' lexical to 'def' register in actions.pm
diff: www.parrotvm.org/svn/parrot/revision?rev=34060
pmichaud we're not talking about _that_ many entries.
just the very basic ones. 06:07
smrf i was kidding
pmichaud sorry, missed the smiley :-)
sometimes I don't always recognize the difference between "." and ":)"
cotto diakopter++ #fixing dalek (or svn (or whatever)) 06:09
diakopter thanks to the patch posters on SVN::Web's RT 06:10
and Infinoid, as usual
mugwump gist.github.com/37408 # 'make spectest' output on utsl 06:11
pmichaud are you using the release version of rakudo, or the one from head? 06:12
mugwump it's the 0.8.2 tarball 06:13
pmichaud yes. the spectests track the head version.
mugwump I thought that being a day or so old it would be fine :)
pmichaud not the latest release.
mugwump hmm
06:14 Zaba joined
pmichaud rakudo development moves too fast for a monthly release schedule -- as you can see, 24 hours can make a huge difference. :-) 06:14
but yes, all of those tests were affected by changes within the last 24 hrs.
mugwump ok. so which svn rev should I use for that?
pmichaud I'd use the head from parrot/trunk 06:15
or do you mean "which svn rev of spectests?"
mugwump yeah, I can easily wind back because it's a git svn clone
smrf yeah, i'm using r34012 of parrot, and that was oooh, 24 hours ago.
pmichaud I'd use whatever svn rev was current as of yesterday 18:00 UTC 06:16
that should be pretty close.
mugwump actually in a maintained git conversion you could use submodules to make the t/spec directory track pugs nicely
so that any version you wound back to would have the right version of the spectests
pmichaud we're looking at moving rakudo to a git repo, yes.
mberends mugwump, take a look at trac.parrot.org/parrot/wiki/PerlSixTesting 06:17
cotto karma purl 06:18
purl purl has karma of 8342
cotto purl--
purl cotto: huh?
cotto karma purl
purl purl has karma of 8342
cotto cheater!!
mberends ratcheting 06:19
pmichaud karma pmichaud
purl pmichaud has karma of 2097
pmichaud oooh, I broke 2000. 06:20
been a long time since I checked. :-)
Coke karma coke?
purl coke has karma of 2343
pmichaud 2343-2097 06:21
purl 246
cotto I wonder if purl-- works in a sentance. 06:25
karma purl
purl purl has karma of 8341
mugwump so, a guy here at catalyst posted a program that worked in pugs but not in rakudo - svn.pugscode.org/pugs//examples/alg...il_call.pl
that's due to use of 'UInt' type
changing it to 'Int' stops the program from failing with a type error 06:26
pmichaud should work if we eliminate UInt, or switch to Int
or define subset UInt of Int where { $_ >= 0 }; :-) 06:27
cotto It's weird to see UInt in a script. 06:29
dalek r34061 | pmichaud++ | : 06:30
: [rakudo]: Convert some $?INIT to .loadinit on blocks.
diff: www.parrotvm.org/svn/parrot/revision?rev=34061
mugwump I see references to UInt in S02
pmichaud yes, in some of the hash types
we will probably implement UInt as I just described above, though :-)
mugwump cool 06:31
the more of Perl6 that is implemented in Perl6, the harder it will rock
cotto The script works with Rakudo with the UInt definition, but only if I *don't* pass anything to the script. 06:32
Otherwise it fails with "Parameter type check failed in call to fib"
pmichaud because arguments to the script are Str, most likely. 06:33
and Str is not UInt
try say fib(+$n)
cotto same 06:34
pmichaud maybe it's thinking that +$n is a Num.
hmmm. probably worth filing as a bug, or thinking about as one. 06:35
Tene That will be handled by type coercion eventually 06:37
pmichaud do we have automatic type coercion?
Tene I remember it being implied by S*, if not specced.
pmichaud I think there was some thinking in that direction, but I haven't seen anything like that in quite a while. 06:38
I suspect coercion might be done with 'as' 06:39
Tene Ah.
pmichaud fib($x as UInt)
but I haven't followed the details closely enough. I think that a type in a parameter specifies a constraint, not a coercion, though. 06:41
(but don't quote me on that.)
mberends Tene++ # very welcome given/when/default fixing: blogs.gurulabs.com/stephen/2008/12/...nwhen.html 06:47
dalek r34062 | pmichaud++ | :
: [rakudo]: Remove remaining pieces of $?INIT. All spectests pass.
diff: www.parrotvm.org/svn/parrot/revision?rev=34062
Tene mberends: are you using it? does it work well for you?
mberends yes, in a grammar based Pod6 parser. I was having to set a Bool $default and clear it inside each when{} as a workaround, quite tedious. 06:50
pmichaud I'm wondering if I want to try to get given/when into NQP 06:51
Tene mberends: I'm glad I could help. :)
06:52 iblechbot joined
mberends :) it has worked well so far. I'll be sure to call if it falls over ;-) 06:52
gaurav whois iftekhar 06:54
06:56 Alias joined
Alias Still working through trying to get Parrot to build without doing any --prefix stuff 06:56
Hopefully should be there tonight
06:59 Zaba joined 07:02 Theory joined 07:26 Zaba joined 07:40 Zaba joined
dalek r34063 | chromatic++ | : 08:26
: [PMC] Modified MultiSub PMC to use PMC ATTRs for storage and not to subclass
: ResizablePMCArray. This should allow us to add cache attribute as well as to
: enable better subclassing of this PMC.
diff: www.parrotvm.org/svn/parrot/revision?rev=34063
chromatic Er, maybe I shouldn't have checked that in. 08:29
08:29 ask_ joined
ask_ who was/is thgibbs? 08:29
I can't find anything matching in CREDITS
moritz purl: thgibbs? 08:30
purl moritz: i haven't a clue
ask_ purl, thgibbs is Tanton Gibbs 08:32
purl OK, ask_.
ask_ forgot he has the authentication database ...
cotto chromatic, do you think it'd be reasonable to make openssl a hard dependency so we can use RAND_bytes for system entropy? 08:33
chromatic It can be a real pain to install. How much entropy do you need? 08:34
justin isn't openssl's randomizer a pseudo-randomizer (which seeds itself a bit from /dev/random)? otherwise using /dev/urandom or /dev/random directly would be just as acceptable
cotto I'm thinking about initializing the hash seed, but the uuid code needs some too. 08:35
justin openssl's own randomization is purely done so as not to deplete the real random device during intensive key generation
(usefull for systems with a urandom equivalent)
chromatic I was about to say.
cotto *per-interp hash seed
C:\\dev\\random: no such file 08:36
lathos egd.sourceforge.net/ ? 08:40
cotto I'll file a bug report and bring it up at the next #ps if nothing happens before then. 08:44
08:46 ffwonko joined
cotto done 08:56
09:00 Aisling joined
dalek r34064 | chromatic++ | : 09:20
: [PMC] Reverting r34063 temporarily, as it slows down Rakudo tremendously and
: breaks Pheme. This patch needs a branch (and probably freeze/thaw).
diff: www.parrotvm.org/svn/parrot/revision?rev=34064
09:27 funktio joined 09:51 mberends joined 09:52 bacek joined
ask_ as a test I'm converting the parrot repository to git 10:09
moritz ask_: with git-svn? or something more elaborate? 10:10
ask_ moritzh: just with git svn
but I'm updating CREDITS with the missing email addresses and svn usernames (won't commit it for now though).
10:10 tomyan joined, ffwonko joined 10:20 iblechbot joined 10:36 AndyA joined 10:53 elmex joined 10:57 Theory joined 11:03 masak joined 11:17 AndyA joined 11:28 ruoso joined 11:29 ruoso joined
Alias hrm 11:37
Setting a --prefix to the same directory as you are building in seems problematic 11:38
11:40 gaz joined 11:53 AndyA joined 12:06 ruoso joined 12:20 AndyA joined 12:27 UltraDM joined 12:30 AndyA joined 12:37 ffwonko joined 12:53 Whiteknight joined 13:10 AndyA joined 13:29 tetragon joined 13:34 gmansi joined
particle emergency shutdown & 13:37
masak particle: hope everything is ok... 13:52
13:57 Lorn joined 14:14 cognominal joined 14:16 cognominal joined
moritz slow parrot day today, it seems :) 14:24
masak people are increasingly going on holiday vacation.
moritz aye
masak moritz: let's surprise them and finish the rakudoreg branch before they come back! :P 14:25
pmichaud some of us are committing refactors. :-)
14:25 ffwonko joined
masak pmichaud++ 14:25
dalek r34065 | pmichaud++ | :
: [rakudo]: Refactor MAIN and startup handling.
: Still reports "dubious" on a few spectests -- investigating this now.
diff: www.parrotvm.org/svn/parrot/revision?rev=34065
masak pmichaud: btw, I also get "Null PMC access in isa()" on trying to compile CGI.pm 14:26
pmichaud masak: in the rakudoreg branch?
moritz thought about implementing infix:<...>
masak pmichaud: aye
moritz dalek doesn't include the branch name in the commit message anymore :(
masak pmichaud: IIU{yesterday's discussion}C, that branch is waiting for another (quick) branch.
pmichaud masak: okay. jonathan and I decided we needed to clean up module and program initialization
masak aye.
pmichaud yes, it's waiting on the branch I just committed to.
I should finish it here shortly this morning. It's gone very smoothly. 14:27
masak excellent.
pmichaud and so far actions.pm is now 75 lines shorter.
(less code == win)
masak \\o/
pmichaud and there's fewer special cases 14:28
and... well, it's just better. :-)
need a morning break, though -- bbiab
14:31 particle joined
Coke horrible? 14:41
purl horrible is GateKeeper from wingate (will eat over 200mb of ram in 4 hours) or The Mummy or ouzo or PacMan joke book or being out of ice cream or STUCK AT WORK
Coke horrible is also a great "tv" show.
purl okay, Coke.
Coke finds some tests hanging in the partcl test suite he ran, which meant it was sitting about 30% of the way through the run all night. 14:42
Coke bets they are all EH re-throws.
Coke is interested in getting a clean run so he can see if it's any faster since 0.8.1
14:44 gryphon joined
Coke thinks out loud: when tcl sources code, it runs it through the tcl compiler; which generates PIR, which in turn generates more PIR. Is there a way in this scheme to get something I can just load? (similar to how test.pm for rakudo is prebuilt; except if I just prebuild the first run of PIR, I then still have the runtime overhead of re-invoking the PIR compiler again.) 14:45
14:46 Andy_ joined 14:48 wknight8111 joined 14:50 Andy joined 14:53 eric256 joined
Coke is there way to invoke parrot that says "don't bother trying to free memory when you're exiting?" 14:53
15:07 iblechbot joined
Coke (ah. I think that might be the default, if --leak-test is any indication.) 15:08
wknight8111 yeah, that is the default 15:13
GC will do a finalization run to call custom destructors on any PMCs, but it doesn't free any of the arenas
15:17 gryphon joined
Coke the delay I was seeing was probably EH related again. just grasping a straws, as debugging some of this takes an insane amount of time. 15:29
s/a /at/ 15:30
jonathan pmichaud: OH HAI 15:31
pmichaud: Haven't really had as much time as expected to do much today...
dalek r34066 | fperrad++ | : 16:03
: [win32] setup
: - update general info
diff: www.parrotvm.org/svn/parrot/revision?rev=34066
r34067 | Whiteknight++ | : 16:12
: [Book] a few small updates to chapter 3
diff: www.parrotvm.org/svn/parrot/revision?rev=34067
pmichaud jonathan: no problem, neither have I. I'm feeling slightly under the weather this morning. 16:13
jonathan :-(
That sucks. Hope you're feeling better soon
pmichaud however, the init/load branch is coming along very nicely when I'm not sleeping :-)
jonathan OK, great.
I'm going out with Slovak friends tonight and have a morning flights, so I'm afraid I'm not going to be around a great deal. :-( 16:14
pmichaud no problem
I may be able to solve 'use' here shortly as well.
jonathan OK. 16:15
pmichaud I've already gotten rid of the $?INIT stuff, and moved the main handling to builtins/guts.pir
so the generated 'main' always looks very short (nopaste coming) 16:16
nopaste "pmichaud" at 72.181.176.220 pasted "generated main" (10 lines) at nopaste.snit.ch/15016
16:16 iblechbot joined
Coke OOC, does "param_21" make it any easier to read than $P14 ? 16:17
pmichaud oops, that's not right. the tailcall is supposed to read '!UNIT_START'($P14, param_21)
anyway, that's the length.
param_21 is slightly easier than $P14, because we know it's a param. 16:18
it also helps when there's an optional param, because the :opt_flag register has a name similar to the :optional register.
16:19 AndyA joined
jonathan Looks nice. :) 16:22
Coke pmichaud: makes sense.
dalek r34068 | pmichaud++ | : 16:23
: [rakudo]: Fix 'main' generation to actually call !UNIT_START.
diff: www.parrotvm.org/svn/parrot/revision?rev=34068
pmichaud getting rid of $?INIT is nice -- one less global thing to be managing 16:24
I also changed all of the $def lexicals to be 'def' registers
particle now has a strawberry build environment for parrot, and can build regularly there 16:26
pmichaud yay!
Coke found it amazingly easier to setup than activeperl+msvc
particle it's not that much easier for me. 16:27
i have custom scripts for each of my environments, so i can switch perl versions, svn versions, etc all by modifying a config file
i should port that to perl someday, instead of using batch files
Coke wonders if adding a .try_catch() macro would clean up his code. 16:31
Coke guesses in about 80% of the cases he has. 16:32
yay. looks like all the tests that regressed in partcl's spec test were EH related. 16:46
17:00 Theory joined 17:06 Whiteknight joined 17:09 tomyan left
Coke thinks he has them all fixed now. 17:11
17:11 jhorwitz joined
nopaste "pmichaud" at 72.181.176.220 pasted "segfault backtrace" (41 lines) at nopaste.snit.ch/15018 17:11
17:11 xiaoyafeng joined
pmichaud note the _pmc_val in the state structure 17:12
Coke kicks off another spec test run for partcl, hopefully having caught all the regressions this time. 17:16
17:24 AndyA joined
pmichaud my segfault is again a tailcall issue :-( 17:28
but, it's easy enough to remove tailcalls, so.... done.
particle tailcall optimization is not supposed to mean we find segfaults faster :( 17:29
17:29 Zaba joined
pmichaud this does tell me that whenever we have segfaults that disappear with -G, perhaps start looking at tailcalls as the culprit. 17:32
Whiteknight you think the two are related? 17:36
17:39 ruoso joined 17:46 chromatic joined
Coke chromatic: yo 17:49
17:52 ffwonko joined
chromatic pong 18:00
pmichaud (tailcall and segfaults) I *know* there's a relationship. 18:10
using a tailcall in a vtable function generally leads to a segfault
I'm guessing that tailcalls in other situations may do so as well.
Coke tailcalls have traditionally not been well tested. 18:11
(that is, they work in a lot of situations, but fail in some; for some time, any tailcall into NCI was borked, as I recally.)
s/y//
chromatic: I hope to know in the next 2 hours if parrot has gotten any faster for partcl since 0.8.1 18:12
18:12 iblechbot joined
chromatic Good! 18:12
dalek r34069 | Whiteknight++ | :
: [Docs] Removed some old cruft from this file, and add a little bit of new information.
diff: www.parrotvm.org/svn/parrot/revision?rev=34069
pmichaud actually, I'm seeing that any tailcall _from_ an NCI is borked -- i.e., if an NCI sub calls something that does a tailcall, then *boom* 18:14
dalek r34070 | pmichaud++ | : 18:17
: [pct]: Don't remove the $0 argument from args when passing for execution.
diff: www.parrotvm.org/svn/parrot/revision?rev=34070
r34071 | pmichaud++ | :
: [pct]: Adjust interpretation of 'lexical' just a bit.
diff: www.parrotvm.org/svn/parrot/revision?rev=34071
r34072 | pmichaud++ | : 18:20
: [pct]: Adjust documentation of 'lexical' on PAST::Block nodes.
diff: www.parrotvm.org/svn/parrot/revision?rev=34072
r34073 | pmichaud++ | : 18:22
: [rakudo]: .tailcall doesn't work well from :load, so don't use it.
diff: www.parrotvm.org/svn/parrot/revision?rev=34073
Whiteknight NCI functions are called through an entirely different mechanism then ordinary subroutines are 18:24
I doubt they even create a return continuation to be used with tailcall
pmichaud yeah. part of the problem is that from PIR we often don't know that what we're calling (or what called us) is an NCI function. 18:25
Coke pmichaud: that tailcall you replaced; you're ignoring the return value now? 18:29
18:29 ffwonko joined
Coke (or is something automatically returning it?) 18:29
pmichaud ignoring.
purl ignoring is v-useful, coulda at least said "failed" or summat
pmichaud it's being called via a load_bytecode op, which doesn't take a return code anyway. 18:30
dalek r34074 | julianalbo++ | :
: fix deprecated syntax in string_cclass test
Coke oh. then you didn't want tailcall anyway. =-)
dalek diff: www.parrotvm.org/svn/parrot/revision?rev=34074
pmichaud was trying to free up a context. EPREMATUREOPTIMIZATION, I guess.
(but it still *should* work)
Coke mm. 18:31
18:38 ff-wonko joined
Coke chromatic: should we steal -Xmsn and -Xmxn from java? 18:44
(minimum and maximum sizes for memory allocation) 18:45
Whiteknight those values are current set with #defines inside the GC 18:46
chromatic Do they set heap allocation limits?
Coke I'd love to be able to run partcl with an -Xmx of 64M and watch it burn. 18:48
chromatic Wake up; time to die. 18:49
Coke chromatic: here's an older list of CL options for java:
java.sun.com/j2se/1.4.2/docs/tooldo.../java.html
I don't see hte one you're asking for.
unless you mean -Xmx 18:50
chromatic jhorwitz, ping 18:55
jhorwitz chromatic: pong
18:56 Hadi joined, Hadi left, alvar joined
chromatic How familiar are you with Apache httpd's request model? 18:58
(leading question? me?)
jhorwitz a little too intimately, some would say... 18:59
Coke chromatic: so far things look good for a speedup.
chromatic I just spoke to the mod_perlite developers, and they have some questions. Know anyone who can give them some tips? 19:01
jhorwitz aaron stone and byrne reese?
19:01 ask_ joined
chromatic Aaron in specific. 19:01
jhorwitz we talked at OSCON. he has my e-mail and can ping me anytime. 19:02
chromatic I'll let him know.
nopaste "pmichaud" at 72.181.176.220 pasted "Another wonderful Parrot treasure." (15 lines) at nopaste.snit.ch/15019
chromatic I've never seen that error message before. 19:03
pmichaud it only occurs in rindex
ascii_cs_rindex, to be precise.
chromatic You're using ISO-8859-1, however. That does make you a terrorist. Hey, the sun came out.
pmichaud and it occurs because the load_bytecode op makes calls to things that are looking for the final '/'
oh, it's even better if I use unicode
chromatic Now I know you hate America. 19:04
pmichaud of course I do, I'm from Texas. :-P
chromatic Is Mrs. Patrick from Texas?
nopaste "pmichaud" at 72.181.176.220 pasted "load_bytecode with unicode string" (15 lines) at nopaste.snit.ch/15020
Coke one does wonder why you need iso to specify "x1.pir".
(I know, I know, that's not the poitn.) 19:05
pmichaud Coke: that's the way the Perl 6 source comes in.
as iso-8859-1
chromatic Here's the question then: do we beg, borrow, steal, or forbid a Unicode library?
Coke I would try to downgrade anything you pass to load_bytecode to fixed8 until the strings PDD hits.
pmichaud yes, that's what I'm going to do.
Still, I'll file a ticket. :-) 19:06
Coke thank you! =-)
(trac, please! =-)
pmichaud now TT #65 19:07
now to remember how to downgrade to fixed8 19:08
Coke I'm wonder why the ascii* version is getting called on an iso-8859-1 string. 19:10
*ing
pmichaud the '/' is ascii
given a path like "path/to/filename.pir" it's searching for the final '/'
chromatic The '/' comes from C code.
pmichaud anyway, downgrading the string to ascii works. As long as we don't have non-ascii filenames. 19:11
NotFound Warning! You are opening a can of worms! 19:12
TimToady if it doesn't support unicode you can't call it Perl 6 :P 19:13
pmichaud TimToady: I can always convert the Unicode specifier into something that's valid ascii. As long as it's transparent to the P6 programmer :-P
(I think that's even mentioned explicitly in a synopsis somewhere) 19:14
NotFound File names encoding things can't be completely transparent, IMO.
pmichaud anyway, my response for today is to throw the can of worms into someone else's backyard. 19:15
NotFound You can have iso-8859-1 file names and be using an utf-8 locale
pmichaud If someone wants to add a spectest for non-ascii modules/filenames, I'll be glad to #?rakudo skip it until we're ready :-)
TimToady we'll need to figure some way of sorting filenames into Buf vs Str categories eventually... 19:16
s/sorting/classifying/ 19:17
NotFound Currently rakudo assumes iso-8859-1 in command line arguments, it isn't?
19:18 ff-wonko joined
TimToady and possibly have some mechanisms for guessing again when our first guess proves to be wrong :) 19:18
pmichaud rakudo doesn't assume anything about command line arguments. It just gets whatever Parrot gives it. 19:19
TimToady on my computer command line args better come in as utf8
pmichaud way-past-lunchtime... bbiaw 19:20
TimToady if there's a choice of default, we should always try to force utf8 wherever possible; iso-8859-1 is almost always going to be wrong in the future
pmichaud for program source rakudo assumes utf8
NotFound $ ack "iso-8859-1" languages/perl6/ 19:21
languages/perl6/perl6.pir
166: $P1 = $P0.'command_line'(args, 'encoding'=>'utf8', 'transcode'=>'iso-8859-1')
TimToady anything defaulting to 8859-1 is a design stench
dalek r34075 | pmichaud++ | :
: [rakudo]: Refactor 'use' statement.
diff: www.parrotvm.org/svn/parrot/revision?rev=34075
pmichaud rakudo assumes utf8, then tries to transcode the source to iso-8859-1 if it can to keep parsing from being interminably slow.
(if the transcode fails, it remains utf8 and parses that.)
TimToady we should make unicode fast :) 19:22
pmichaud that's another can I threw into someone else's backyard.
I suspect allison's backyard is full of these things by now. :-)
TimToady when is rakudo scheduled to get NFG support?
pmichaud checks roadmap.
Parrot gets nfg support by feb 2009 release. 19:23
so Rakudo shortly thereafter.
(if not simultaneously)
at least, I presume that's what the "pdd28-strings" task is intended to accomplish.
TimToady will that be with one integer repr or with multiple? 19:24
moritz has anyobody prototyped NFG things outside of parrot before?
pmichaud I don't know. That's somewhat transparent to Rakudo -- it's really a Parrot decision.
TimToady since UInt8 is an obvious way to store NFG for latin-1
pmichaud at least, at the moment it's transparent to Rakudo. 19:25
and not something I can spec until Parrot has its spec and implementation in place.
TimToady okay, hope they don't mess it up :)
Coke NFG?
purl NFG is No Fscking Good! or grapheme normalization form (see docs/pdds/draft/pdd28_character_sets.pod)
Coke agrees with purl.
purl agrees with Coke
pmichaud me too. Otherwise I get to add another layer of stuff on top of what Parrot provides. :-) 19:26
NotFound That's christmas spirit ;)
pmichaud okay, gone for real this time. bbiah
TimToady disagrees, since it should be and/or in English
19:26 ff-wonko joined
TimToady actually "or" is usually inclusive in English... 19:26
but not always 19:27
Infinoid There's an exception to every exception.
NotFound $ ./perl6 -e "say 'eƱe';"
eƃĀ±e
TimToady I can handle it.
NotFound Not really transparent.
TimToady biab & 19:28
Coke (exception to every exception) that's why partcl was hanging recently, I believe.
NotFound Exception: recursive exception found 19:29
Infinoid People who live in glass parrots shouldn't throw exceptions.
chromatic People who live in exceptions should tailcall exceptions. 19:30
Coke kegfaults. 19:31
cotto you don't make a very convincing Internet 19:41
19:43 Theory joined
Whiteknight kegfaulting can ruin a party if you're not careful 19:45
19:46 hercynium joined
Coke chromatic: spectest runtime has (with no changes in # of tests) dropped from 12550s to 9232s 19:55
12550/9232
purl 1.3594020797227
Coke 9232 / 3600
purl 2.56444444444444
Coke (12550-9232)/3600
purl 0.921666666666667
Coke that's nearly an hour shaved off.
nice.
only 2.5 more to go. ^_^ 19:56
Whiteknight Coke++ 19:57
how many tests are in the spec suite?
chromatic Do you use multidispatch? 19:58
Coke chromatic: a bit. 19:59
things like toNumber() which return a PMC with a number in it have braindead int and float versions, e.g. 20:00
(and more complicated string versions). Also, HEAVY use of PGE.
chromatic Okay. I just found two inefficiencies there.
Coke only two? =-)
chromatic Two easy ones. 20:01
Coke Whiteknight: code.google.com/p/partcl/wiki/SpecTestStatus 20:04
I can tell you how many test files and how many tests are running, but am not yet tracking how many more files or test there are in those files. 20:05
(currently, 57 files, 4001 tests, 2341 pass, 1059 fail, 601 skip)
particle out of ~20k tests total? 20:06
Coke code.google.com/p/partcl/wiki/SpecTestExplode points to some easy wins to get several hundred more tests passing.
s/easy wins/ways/
particle: I haven't counted. moment.
(running all.tcl with tclsh8.5) 20:08
dalek r34076 | chromatic++ | : 20:10
: [MMD] Removed two unnecessary clone operations on MultiSubs. As currently the
: candidate list doesn't get sorted in place, there's no reason to clone it.
Coke starts over with verbose PASS output.
dalek diff: www.parrotvm.org/svn/parrot/revision?rev=34076
Coke particle: some of the tests are slow. this could take some time. 20:12
particle sorry, i thought you knew.
Coke I so rarely get to run the suite. =-) 20:13
I have an old journal entry that says 20683.
no clue if that's right.
Coke doubts it.
purl doubts it is probably possible to offend this #perl
Whiteknight purl forget doubts it 20:14
purl Whiteknight: I forgot doubts it
Whiteknight damn straight you did
particle i remember timer alone was ~12k
chromatic AHA A MISSING DEPENDENCY 20:15
dalek r34077 | chromatic++ | : 20:17
: [config] Added a missing dependency for src/inter_create.c on its strings file.
diff: www.parrotvm.org/svn/parrot/revision?rev=34077
Whiteknight chromatic++
Coke particle: 13516 tests if I skip two that have hangs in them. 20:27
so, anywhere from 13-25K. =-) 20:28
Whiteknight has anybody tried compiling parrot with intel c? 20:30
20:30 rurban joined
dalek r34078 | pmichaud++ | : 20:30
: [rakudo]: Add some comments to use_statement handling.
diff: www.parrotvm.org/svn/parrot/revision?rev=34078
Coke Whiteknight: sounds faimilar. 20:31
Coke tries to get Whiteknight to remerge his GC work instead of chasing compilers. =-)
Whiteknight compilers I can do at work, because I can set the computer off downloading or installing things by itself 20:32
I can't do too much development or debugging on the company's dime
:)
plus, this laptop has a 320GB HDD. I'm never going to fill that all up unless I download all sorts of garbage 20:33
pmichaud wonders if filling up the 320GB HDD makes the laptop any heavier. 20:34
Whiteknight certainly makes my hands more tired
so it feels heavier
particle Whiteknight: we had folks building with icc in the past, but not recently. i haven't reinstalled it after rebuilding my pc 20:35
chromatic o/~ Never gonna fill it up / never gonna let it down o/~
particle comments++ pmichaud++ 20:36
jhorwitz did we just get rick-rolled? 20:37
particle man, seattle is SHUT DOWN today. 5" snow on top of an ice layer (due to rain last night) in a city with < 30 plows 20:38
pmichaud the towns 2 mi north of here were shut down on tuesday due to ice. We weren't, though. 20:39
afk, picking up kids from school so they don't freeze walking home.
Whiteknight only 5 inches of snow on the ground? I wouldn't even wear a jacket
Coke back in my day...
... I didn't live in dallas. 20:40
particle: (num of tests) note that I'm now targetting the 8.5.5 test suite; no clue what it was back then. 20:42
20:48 allison joined
chromatic Could huve been a duck roll. 20:51
o/~ Never gonna quack you up / never gonna quack you down o/~
dalek r34079 | chromatic++ | : 20:54
: [MMD] Tidied code; no functional changes.
diff: www.parrotvm.org/svn/parrot/revision?rev=34079
20:59 mj41 joined 21:02 kj joined
cotto snow-- 21:06
Coke o/~ I want a duck who can show me how to get down. Get It!? ~/o 21:07
ooh, a bus error. 21:08
purl Passengers dumped.
masak Coke: I've always loved that line, too.
Coke is it possible to share a screen session with another linux user? 21:09
(I'm wondering if I can run my long running code to bus error in gdb and then hand it off to someone like chromatic) 21:10
ooh. or I could get lucky and come up with a shorter test case. 21:11
21:13 gryphon joined
GeJ Good morning everyone 21:13
dalek r34080 | chromatic++ | : 21:14
: [MMD] Removed deprecated Parrot_mmd_add_function().
diff: www.parrotvm.org/svn/parrot/revision?rev=34080
masak GeJ: evenin'
GeJ hej 21:15
masak :)
cotto Coke, there is.
cotto runs man screen
NotFound Coke: you can make it dump core, and then gdb the core image. 21:17
dalek r34081 | util++ | : 21:18
: Typo - s/prevends/prevents/
diff: www.parrotvm.org/svn/parrot/revision?rev=34081
21:19 japhb joined
cotto Coke, www.linux.com/feature/56443 21:19
PacoLinux Coke>\tis it possible to share a screen session with another linux user? <-- you can use -kibitz- is a little program in expect : expect.nist.gov/example/kibitz.man.html 21:20
NotFound Current linux distros usually does not dump by default, but you can change that with: ulimit -c unlimited
PacoLinux Coke: and expect runs on top of tcl, I want to see expect in top of partcl :) 21:24
NotFound Do you expect it? 21:25
21:31 TimToady joined 21:33 Zaba_ joined
Coke PacoLinux: expect is more than just .tcl, though. 21:34
ah, kibitz, forgot about that. 21:35
been quite some time since I was paid to code in tcl/expect.
Coke gets a semi-reproducible test case.
davidfetter hello 21:36
anybody have contacts at MSFT USA?
Coke alias, particle, cotto... 21:38
why?
21:38 Zaba joined
Coke chromatic: the segf is, gleefully, in clone_key_arg 21:39
21:39 Hadi joined, Hadi left
chromatic Keys, my only nemesis. 21:39
davidfetter provides chromatic with several more nemeses 21:40
21:41 ElPenguin joined
masak "You win this time, Keys!" 21:42
Infinoid and your little dog Toto, too. 21:43
pmichaud Coke: that's the *same* segfault I was getting earlier. 21:44
dalek r34082 | chromatic++ | :
: [PIC] Removed the only reference to deprecated old MMD functions in src/pic.c.
: There are no consumers of this PIC function anyway. Let the chainsaw continue.
diff: www.parrotvm.org/svn/parrot/revision?rev=34082
r34083 | kjs++ | :
: [pirc] fix parser; some missing rules.
: + update a test
diff: www.parrotvm.org/svn/parrot/revision?rev=34083
pmichaud i.e., I was getting the segfault in clone_key_arg 21:45
21:45 pjcj joined
davidfetter Next time, Keys. Next time. 21:46
</klaw>
21:50 rob joined 21:51 Zaba_ joined
dalek r34084 | chromatic++ | : 21:53
: [MMD] Removed deprecated mmd_expand_x(), mmd_expand_y(), Parrot_mmd_register(),
: and Parrot_mmd_register_sub() functions.
diff: www.parrotvm.org/svn/parrot/revision?rev=34084
r34085 | chromatic++ | : 21:57
: [MMD] Removed deprecated mmd_destroy() function.
diff: www.parrotvm.org/svn/parrot/revision?rev=34085
Coke chromatic++ #ripping out more MMD stuff;
pmichaud: wow. for once, partcl and rakudo agree on something! ^_^
Coke tries to figure out how many passing spec tests aren't getting counted because of that. 21:58
GeJ is MMD something bad? 21:59
Coke nope.
just a lot of leftovers in the code from the old way of doing MMD. 22:00
chromatic I don't think we can get rid of mmd_cvt_to_types yet.
Coke ISTR that was the hard one.
chromatic The PIC mess wasn't much easier, but it's unused, so hooray.
Coke was PIC another optimization? 22:01
PIC? 22:03
purl PIC is a preprocessor for *roff, like tbl or eqn. or a language for drawing pictures and diagrams or pilot in command or MONKEY or Position Independant Code or Polymorphic Inline Cache or the popular line of Microcontrollers from Microchip or SHOW US THE PORN BITCH
22:03 apeiron joined
chromatic Yes, but some of the predereferenced cores use some of its features. 22:04
Coke I wonder again why we have so many cores. :|
chromatic To make Piethon go fast.
particle what's wrong with optimization, now?
Coke nothing, as long as everything /works/ first.
particle i don't like cgp, cause it's gcc-only
Coke (and after) 22:05
chromatic and the code is maintainable
particle pic is a perfectly fine optimization, if it's used
chromatic and gets tested
Coke pmichaud: do you have a smaller test case than mine? mine involves several KLO Tcl.
particle the only problem with our many runcores is that they're not tested regularly.
dalek r34086 | kjs++ | :
: [pirc] don't generate an instruction if it would be: set $P0, $P0. That's a no-op.
diff: www.parrotvm.org/svn/parrot/revision?rev=34086
r34087 | chromatic++ | :
: [MMD] Removed deprecated Parrot_MMD_search_default_infix() and
: Parrot_mmd_search_default() functions. 22:06
diff: www.parrotvm.org/svn/parrot/revision?rev=34087
Coke if they're never tested (by which we could say never used), why have it?
particle that's enough of a reason to remove them if it makes our releases more stable.
Coke there we go.
chromatic Yeah, you've almost convinced me.
particle however, better smoke infrastructure can also fix the problem
22:07 Whiteknight joined
Coke assuming they /work/, we can then assess if they're enough of a speed improvement to keep. 22:07
22:07 Zaba joined
particle wanna rip out -O also? 22:07
where do we draw the line? 22:08
Coke that would be an easy way to fix all the bugs reported against -O. =-)
particle: further away than we have so far.
254/2341 22:09
purl 0.108500640751815
dalek r34088 | kjs++ | :
: [pirc] remove commented-out code. it's no longer necessary.
diff: www.parrotvm.org/svn/parrot/revision?rev=34088
Coke chromatic; TT# 66 is blocking 254 spec tests from being accounted for, about 11% of my current total. 22:10
speaking of optimizations, is r34086 in the core pirc loop? is there an optimization stage there? 22:16
pmichaud Coke: I can make a relatively small test case, although it does involve loading perl6.pbc (which is a little large). But the actual code is small. 22:18
I can post an example once I get the rakudo 'main' branch merged to trunk.
Coke ok. 22:20
(thanks!)
GeJ pmichaud: the rakudo 'main' branch? 22:21
pmichaud it's a branch where I'm refactoring load/init/main handling. 22:22
I'm merging to trunk now..._except_ I've exposed a bug in NQP. :-(
no, I said that wrong.
I've exposed another Parrot segfault bug that manifests itself in NQP.
particle that's a smaller test case :)
pmichaud :load and :init are very fragile, it seems. 22:23
GeJ Time to resync my local repo.
Whiteknight everything related to sub calling is fragile
particle imcc is fragile
pmichaud good thing we don't call subs very oft..... oops.
Coke waits for someone to say, "your mama's so fragile, she..."
"... broke a hip sneezing." 22:24
Whiteknight "... she broke somebody else's hip sneezing"
masak "... broke a hip when somebody else sneezed." 22:25
chromatic "... she sits AROUND the house."
particle masak++
Whiteknight "... broke a hip cause she's so fat too!"
Coke karma dozens? 22:26
purl dozens has neutral karma
Coke heads out, hoping the segfault fairy fixes his bug.
Whiteknight out of sheer morbit curiosity, I am trying to update the gsoc_pdd09 branch to trunk. There are about a hundred conflicts
Coke ->
chromatic Coke, I'll look at it if pmichaud can give a small PIR example.
Whiteknight purl chromatic is the segfault fairy 22:27
purl ...but chromatic is <req><req>a lot of fun. For years I'd try to play stuff like `peter and the wolf', and then I'd be frustrated because it would use some note I didn't have. or the author of jellybean or mailto:chromatic@wgz.org or wgz.org/chromatic/ or the winner of the not-a-contest perl-bugathon. or best reached via email. or the guy who hit me in the eye....
Coke I can continue to try to strip it down on the tcl side, too.
pmichaud it's not too long, no. Very reproducible. Doesn't seem to be quite as sensitive to minor code changes.
Coke pmichaud: then I defer to you; the tcl variant seemed quite sensitive.
pmichaud yes, the lack of sensitivity is what surprised me about this one earlier today. 22:28
Coke (always in the same spot, mind you, but required a lot of apparently unnecessary preamble code.)
Coke runs another long-dead spec test to see if it's a related issue.
and now, really. -=>
22:39 Theory joined 22:41 funktio joined
dalek r34089 | pmichaud++ | : 22:55
: [nqp]: Clean up a test file for :load :init semantics.
diff: www.parrotvm.org/svn/parrot/revision?rev=34089
23:20 apeiron_ joined
nopaste "pmichaud" at 72.181.176.220 pasted "new code for use_statement" (22 lines) at nopaste.snit.ch/15028 23:22
dalek r34090 | pmichaud++ | : 23:39
: Merge Rakudo 'main' branch (refactor load/init/main) into trunk.
diff: www.parrotvm.org/svn/parrot/revision?rev=34090
23:42 Eevee joined
particle !UNIT_START could use documentation 23:50
23:55 Zaba_ joined

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!