Camelia, the Perl 6 bug

IRC log for #parrot, 2008-03-30

Parrot | source cross referenced

| Channels | #parrot index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:00 lichtkind pmichaud: currently finishinh just another perl 6 article :)
00:00 pmichaud the plan is that eventually rakudo and STD.pm will converge, yes.
00:00 Tene lichtkind: NQP is pmichaud's.  We'd like to use STD.pm eventually, and the current rakudo grammar is modeled after it.
00:00 pmichaud even STD.pm is still "just a draft"
00:00 lichtkind Tene: thanks
00:00 lichtkind yes
00:01 lichtkind raku do is also patricks brainchild?
00:01 pmichaud I don't know that I'd call it my "brainchild".  Several years ago I was asked to be the pumpking for developing a Perl 6 compiler on Parrot.  Rakudo is what has resulted.
00:01 lichtkind pmichaud: most pugs folks including larry dont know much about that plan :)
00:02 pmichaud about what plan?
00:02 lichtkind pmichaud>the plan is that eventually rakudo and STD.pm will converge, yes.
00:02 pmichaud I can't speak for pugs folks, but I highly doubt that Larry believes that we're purposefully avoiding convergence.
00:03 lichtkind pmichaud: but why then is rokudo just born last autumn?
00:03 lichtkind :)
00:03 pmichaud it wasn't born just last autumn.  It was created spring 2006
00:04 pmichaud in fact, one could say (and I believe Larry is one who has said it) that rakudo's early grammars informed much of the work on STD.pm
00:04 pmichaud (TimToady:  if I'm mis-speaking here, please feel free to publicly correct me.)
00:07 pmichaud of course, prior to January 2008 rakudo was known as "perl6".
00:07 lichtkind pmichaud: nono i think they ment it more that way that they dont expect to run std on pge next half year or so
00:07 lichtkind and what was that perl 6 capable of?
00:08 lichtkind they care more about to run stc in their own way
00:08 pmichaud I can't say for sure when STD.pm will be able to run on PGE.  But at the moment I'm more interested in getting something that can run Perl 6 programs than to make sure that we can run STD.pm
00:09 pmichaud that said, we're definitely following STD.pm and rely on it for all of our grammar additions
00:09 lichtkind pmichaud:  that more important since pugs is sleeping so i can say my listeners instead to use rakudo for the tutorial
00:10 pmichaud right -- at the moment my feeling is that having a platform that can run Perl 6 code is just a little bit more important than making sure we can parse/implement STD.pm.  But having rakudo be STD.pm based is a definite objective for us, and not just a "wish".
00:10 lichtkind but its fair to say that older version of perl 6 didn't made much?
00:10 Tene Hmm.  Thinking about working on rakudo this evening.
00:11 pmichaud the older version of perl6 was stalled because we needed a fair number of improvements to Parrot and to the Perl 6 spec, most of which happened in 2007
00:11 Tene Any requests?
00:11 pmichaud once those changes fell into place, work on rakudo sped ahead very quickly
00:12 lichtkind and its right when i wrote that the new oop modul of parrot gave birth to NQP
00:12 lichtkind during last year
00:12 pmichaud in fact, the existence of STD.pm (and the changes it caused to the spec) is part of the reason why work on rakudo itself stalled for much of 2007 -- we had to re-group to be able to meet the new spec
00:13 pmichaud the new OO module in parrot was a necessary condition for much of this (NQP, rakudo, PCT, etc.).  Before that we were severely hampered by problems in Parrot's object system.
00:14 pmichaud I came up with the idea for NQP in spring 2007, started building it in July 2007, but needed a few other tools in place before it could become a usable tool (November 2007)
00:15 lichtkind great
00:15 Tene pmichaud: in an actions.pm, it's possible to manually look backwards to see if a variable has already been referenced?
00:16 pmichaud not easily.  However, it's possible for a variable to store itself in a table that says "hey, I've been referenced" :-)
00:16 pmichaud and then actions.pm could look in that table
00:16 Tene Ooo, clever.
00:16 pmichaud but it depends on what you mean by "has already been referenced"
00:16 pmichaud I assume you're talking about undeclared variables?
00:16 Tene Right.
00:16 pmichaud consider code like
00:16 pmichaud sub foo() {
00:16 pmichaud if (cond) {
00:17 pmichaud $a = 5;
00:17 pmichaud }
00:17 guru joined #parrot
00:17 pmichaud //  # has $a been referenced here?
00:17 pmichaud }
00:17 Tene According to the current lolcode tests, it should be possible to say the equivalent of "my $a;" about an already-existing variable, without effect.
00:18 Tene The current implementation achieved that by only setting isdecl on declarations with an initial value.
00:18 Tene This, of course, resulted in declarations without an initial value failing.
00:19 pmichaud lichtkind: were my answers helpful?  I know you asked me these items in email, but I was on vacation last week (honeymoon, actually) and so I got backlogged on answering emails
00:19 guru I just tried to checkout a fresh copy of Parrot into a newly created directory on XP, and ran into a problem.
00:19 Tene I'm tempted to just solve it by putting 'isdecl' on every declaration, as I can't find anything in the spec to support requiring that behavior.
00:19 pmichaud Tene: variables are assumed to be local (lexical) ?
00:19 Tene Yes.
00:19 pmichaud isdecl on every variable (if they're lexical) will likely mess that up
00:19 pmichaud the isdecl needs to go only on the first instance
00:20 Tene Right, lolcode has explicit declarations.
00:20 guru everything went well until parrot\languages\hq9plus\hq9plus.pir
00:20 lichtkind pmichaud: happy for you (honeymoon) yeah great im currently sitting again on my editor figuring out the next textblock i thing were no finished for now :)
00:20 pmichaud guru:  I think you're running into the problem that others are having with a duplicate-named directory in the repo.
00:21 pmichaud tene:  a variable has to be declared before usage?
00:21 guru then the job aborted, claiming a duplicate hq9plus directory.  Any suggestions?
00:21 Tene pmichaud: yes.
00:21 pmichaud Tene: oh, then there should be no problem.  this is what the 'symbol' attribute is for in PAST::Block nodes
00:21 * Tene reads.
00:21 pmichaud rakudo does the same thing
00:22 pmichaud whenever a 'my $a' is encountered, rakudo checks the current block's symbol table to see if $a is already registered.  If so, it ignores the 'my' (actually it should check to make sure the decl is the same), if not then it sets the :isdecl flag on the PAST::Var node and places the variable into the symbol table.
00:22 diakopter guru: someone is working on that at the moment, I think
00:23 Tene Would it be easier for me to just check rakudo to see how to access the current block?
00:23 pmichaud well, it also requires keep track of "the current block"
00:23 guru OK, thanks.
00:23 pmichaud kjs' tutorial talks about it a bit
00:24 pmichaud in rakudo the code that handles it is the scope_declarator method
00:24 nopaste "pmichaud" at 76.183.97.54 pasted "scope_declarator from rakudo" (16 lines) at http://nopaste.snit.ch/12601
00:25 Tene Yeah, that's what I'm looking at right now.
00:25 pmichaud the unless statement says "unless we've seen this symbol already in the current block"
00:25 pmichaud then towards the end of the method there's a line that says
00:25 pmichaud $?BLOCK.symbol($name, :scope($scope));
00:26 pmichaud which tells the $?BLOCK that the symbol given by $name has been declared locally with the given $scope
00:26 lichtkind pmichaud: how big will parrot be , or its useless to estimate?
00:26 pmichaud lichtkind: I don't know.  If you're referring to memory footprint, I expect it will get much smaller at some point in the future.  Right now it's big because there's a lot of debugging and other items in place.
00:26 lichtkind pmichaud:  i had in erlangen some interesting rants with jonathan, something we maybe also could discuss
00:27 lichtkind pmichaud:  now im thinking more of size of sources, but yes also memory
00:27 lichtkind the plan was to make parrot into a heart of an next gen IDE with full flexed live introspection for dynamic languages
00:28 lichtkind since im maintainer of a editor we discussed some posibilities
00:28 pmichaud yes, that's one of the things I find particularly useful about Parrot
00:29 pmichaud I think it will have some _great_ IDE and introspection capabilities
00:29 pmichaud Paula's calling me for dinner -- bbl
01:03 lichtkind pmichaud: my point is that im mostly have not enough time to handle my editor project, not to think of such a big fuzz even if its more promising that anything
01:37 lichtkind is NQP the only reason why rakudo gained speed last year?
01:47 Tene One reason, not the only one.
01:47 lichtkind Tene do you know more?
02:04 jrt4 joined #parrot
02:32 lichtkind joined #parrot
03:35 rdice joined #parrot
03:36 Theory joined #parrot
03:49 tetragon joined #parrot
04:02 ask_ joined #parrot
04:16 AndyA joined #parrot
04:28 ask_ joined #parrot
04:28 Andy joined #parrot
04:55 Coke joined #parrot
05:16 Theory joined #parrot
05:24 lichtkind joined #parrot
05:47 Psyche^ joined #parrot
06:33 iblechbot joined #parrot
07:47 barney joined #parrot
08:01 rrando_ joined #parrot
08:31 iblechbot joined #parrot
08:33 Ademan joined #parrot
08:58 dalek r26639 | bernhard++ | trunk:
08:58 dalek : [HQ9+]
08:58 dalek : Clean up files generated during testing.
08:58 dalek : Remove DYNPMC support in Makefile.
08:58 dalek diff: http://www.parrotvm.org/svn​/parrot/revision?rev=26639
09:36 barney infinoid++ for getting rid of unneeded HQ9plus.pm
10:57 cognominal I am working on a fix for #52276
11:25 dalek r26640 | bernhard++ | trunk:
11:25 dalek : [HQ9+]
11:25 dalek : Remove obsolete directory languages/HQ9plus.
11:25 dalek diff: http://www.parrotvm.org/svn​/parrot/revision?rev=26640
11:30 dalek r26641 | bernhard++ | trunk:
11:30 dalek : #52264 (Bug: duplicate directory names in SVN)
11:30 dalek : Run mk_manifest_and_skip.pl after removing languages/HQ9plus
11:30 dalek diff: http://www.parrotvm.org/svn​/parrot/revision?rev=26641
11:56 dalek r26642 | bernhard++ | trunk:
11:56 dalek : [HQ9+]
11:56 dalek : Update NEWS.
11:56 dalek diff: http://www.parrotvm.org/svn​/parrot/revision?rev=26642
13:12 rdice joined #parrot
13:12 davidfetter joined #parrot
13:17 slavorg joined #parrot
14:29 guru joined #parrot
14:45 guru left #parrot
15:01 Theory joined #parrot
15:07 davidfetter joined #parrot
15:19 rrando joined #parrot
15:27 jan joined #parrot
15:47 Limbic_Region joined #parrot
15:53 guru joined #parrot
16:05 guru left #parrot
16:44 buchetc joined #parrot
16:51 guru joined #parrot
17:01 tetragon joined #parrot
17:08 darbelo joined #parrot
17:09 lightsey joined #parrot
17:11 Psyche^ joined #parrot
18:08 kj joined #parrot
18:12 jan joined #parrot
18:13 tsg joined #parrot
18:14 * Infinoid makes some progress in getting "make -j2" to work
18:18 Infinoid ("makes"!  get it?  ha ha.)
18:18 IllvilJa joined #parrot
18:22 cotto_home thanks for working on that!
18:24 dalek r26643 | infinoid++ | trunk:
18:24 dalek : [Makefile] The pdd17pmc branch merge broke "make -j2".
18:24 dalek : * Add some missing build interdependencies, so make knows what it can and can't
18:24 dalek :   build in parallel.
18:24 dalek : * Now "make -j2" seems to work pretty reliably for me... more testing is on the
18:24 dalek :   way.
18:24 dalek : * This change shouldn't have any effect on non-parallel builds.
18:24 dalek diff: http://www.parrotvm.org/svn​/parrot/revision?rev=26643
18:25 * Infinoid starts bashing the hell out of it with "make -j8"
18:26 Infinoid I'm focusing on GNU make, at the moment.  But I'm curious, does nmake support multithreaded building too?
18:29 leo Infinoid: -j3 started testing successfully here - looking good - thx
18:29 Infinoid -j8 causes a segfault in miniparrot
18:29 Infinoid thanks for the testing :)
18:29 leo welcome
18:31 Coke leo: hey!
18:31 leo Infinoid: indeed -j8 is b0rked
18:31 leo hi Coke
18:34 Infinoid (-j8 was busted before pdd17pmc, too)
18:36 leo Infinoid: -j5 did stop too - anyway, the amount of -j really doesn't matter if all dependencies are ok - so the succeeding -j2 was an accident ;)
18:37 Infinoid yeah, higher numbers just shake out bugs more I think :)
18:37 leo depends on load, whatever, but yes
18:37 Infinoid comprehensive tests are difficult.
18:37 leo watch out for missing .str file deps
18:38 Coke c has threatened twice now to redo how we do .str stuff to avoid issues.
18:38 * leo thinks this is error prone and should be automagicall happen
18:38 leo *ly
18:39 Infinoid leo: absolutely.  .str deps should be generated by the summary rules
18:40 guru left #parrot
18:40 leo indeed - that part of your patch: +$(SRC_DIR)/scheduler$(O) : $(SRC_DIR)/scheduler.str   \
18:41 leo should really be autogened by detecting usage of CONST_STRING inside .c
18:41 Infinoid is there a way to do that automatically, without grepping for CONST_STRING for each source file before building?
18:42 * Infinoid will let chromatic solve that problem :)
18:42 leo as -jn was working some time ago, I'd estimate, the evil is one of the newer .c files
18:42 leo Infinoid: greeping is fine, isn't it ;)
18:42 Infinoid it's just slow
18:43 Infinoid however, .str isn't the only issue here
18:43 leo doit it once, iff there is no .str file
18:43 Infinoid one of the bugs I fixed was: a source file in src/ wanted to #include a header (generated by pmc2c) in src/pmc/
18:43 Infinoid another of the bugs: compilers/pge/ was being run before $(GEN_LIBRARY) had completed
18:44 Infinoid so I'm trying to find and fix all of those things I can...
18:44 leo yep there are a log of possible errors
18:44 leo Infinoid++
18:45 Infinoid hmm
18:45 Infinoid [11:43] <@leo> doit it once, iff there is no .str file
18:46 Infinoid ... or if the .c timestamp shows a modification
18:46 leo well after a make, you have related .src files
18:46 leo .str - sorry
18:46 Infinoid true, just possibly out of date
18:47 leo doesn't  matter, check against the dep rules
18:47 Infinoid and if the .c file (after modifications) no longer has any CONST_STRINGs, that means you have to remove the .str and everything else
18:47 Infinoid I don't understand .str, but it seems pretty ugly :)
18:48 leo .str is a nice thing to have to preconstruct const STRINGs at compile time
18:48 Infinoid ah, that sounds useful
18:48 leo it is
18:49 Infinoid can you do what the linux kernel folks do, namely, drop it out into a constant-sized record in a separate linker section?
18:49 Infinoid that'd remove the need for any secondary files
18:49 leo the .str files are just doing that
18:50 Infinoid ok
18:51 leo well, not a separate linker section, but a separate const structure
18:51 Infinoid they use a linker section because ld concatenates everything for you in the final link, and that makes it easy to enumerate
18:51 Infinoid they use that for init() function pointers, for example
18:52 Infinoid I dunno.  maybe when make -j works, I'll start looking at breaking .str :)
18:52 leo see also:
18:52 leo # constant string support
18:52 leo .c.str : $(PERL) $(BUILD_TOOLS_DIR)/c2str.pl $< > $@
18:52 leo which has another list of deps
18:53 leo ... in the Makefile
18:57 Infinoid ok.  I see the problem caused by make -j8, now I have to figure out how to fix it.
18:57 Infinoid register_nci_method() was called with an Iterator PMC whose _namespace pointer was NULL
18:58 rrando_ joined #parrot
18:59 Infinoid I don't think the vtables[] array was fully built
18:59 leo Infinoid: that's very likely just a secondary error
19:11 Coke Infinoid: src/scheduler.c has a borked dependency.
19:11 Coke ... and it's my fault. =-)
19:11 Infinoid another one?
19:11 Coke I added 3 includes to pmc/*.h there to let it *build* on my machine, but didn't add those deps.
19:12 Coke ... did I miss an svn up? might have.
19:12 lichtkind pmichaud: thanks without your help i could not finish the article
19:12 Coke if you've already fixed that, danke.
19:12 Infinoid yeah, I added those 3 :)
19:12 Infinoid and a couple of other things
19:12 Coke yay.
19:14 darbelo Coke: I talked with you about a SoC proposal involving languages/c99 a few day ago. I have few more questions about that.
19:14 darbelo do you have a minute?
19:14 Coke darbelo: I have a real world thing that may interrupt at any second, but yes.
19:15 Coke if I run away, feel free to email particle and myself.
19:15 Coke (and I can reply within an hour or so.)
19:16 darbelo OK, sorry I was silent this few days but Life got briefly in the way. I'm trying to hammer out the timeline of the proposal.
19:16 Coke Infinoid: I just did a few high -j runs, and it either worked, or did the cygwin-y segfault when running miniparrot to gen the config.
19:16 Coke darbelo: happens.
19:17 Infinoid Coke: I am currently debugging a segfault in miniparrot, but it's happening on linux.  are we talking about the same thing, or is there an additioanl cygwin issue?
19:17 Coke I use it only as a point of reference. it's how cygwin used to die all the time when cygwin refused to build for me.
19:18 Coke I'm doing my current testing on feather because it's much faster than my desktop. =-)
19:18 Infinoid well, I'll ping you when I've found a fix.
19:19 Coke when I was debugging this the last time, I ended up doing a build in ../parrot with -j <big>, and moving that off. then getting a fresh copy in parrot/ and building with -j 1 and then doing a diff -rbu (using svn export to avoid diffing all the .svn files)
19:19 Coke which was a PITA but did help narrow down the .str issues.
19:20 Infinoid that's pretty similar to the setup I've got right now. :)  but the .str issues seem to be solved, for the moment.
19:20 Coke excellent.
19:20 Infinoid I think this is just a missing Makefile dep
19:20 Coke I feel like an idiot, I did a quick test after I saw your initial stuff and realized I'm the one that borked the deps on that file while fixing the -j1 build.
19:20 Infinoid seems likely that it's proceeding to the link phase with something half-built.
19:21 darbelo chromatic suggested testing c features by groups and using the tests as milestones.
19:21 darbelo I'm trying to find the right level of granularity.
19:21 Coke I've heard eric talk about "every 2 weeks"
19:22 Coke which from a real world project management standpoint, makes sense. anything longer than 2 weeks is hard to manage.
19:25 leo Coke: [BUG] in a subject is still enough to file a report?
19:26 Coke anything in the subject is fine.
19:26 Coke it all ends up in rt.
19:27 Coke BUG vs PATCH vs RFC vs RFE vs... is *mostly* just to help the humans looking at the tickets.
19:27 Coke (PATCH is slightly special)
19:27 darbelo I meant feature granularity. Think  'all looping constructs'  versus 'for' 'while' 'do-while' as separate entities.
19:28 darbelo Or, are you suggesting 'feature set that can be implemented in two weeks time' ?
19:32 Senaka joined #parrot
19:32 ewilhelm Coke: Senaka is applying to Apache for SoC
19:32 Coke real world was here and is now gone.
19:33 Coke darbelo: that.
19:33 Senaka Coke: hi
19:33 purl bonjour, Senaka.
19:33 Coke that's a good way to base your granularity, I think.
19:33 guru joined #parrot
19:33 Senaka need some help here
19:33 Coke ewilhelm: feel free to jump in, also. ^_^
19:33 Coke Senaka: Guten tag.
19:34 Senaka Coke: sorry ddn't get you.
19:34 * ewilhelm tries to keep fingers and toes away from GC's
19:34 Senaka :)
19:34 Coke Senaka: Good day. hello.
19:34 Senaka :)
19:34 Senaka hello, Good day you too
19:34 darbelo Now that I think about it, that makes more sense, from a management perspective,
19:34 Senaka I'm working on Harmony GC
19:35 Coke EEEEEEEK, garbage collection!
19:35 Senaka and the aim is to make it pluggable to Parrot
19:35 Senaka :)
19:35 Coke that would be nice, actually. =-)
19:35 Senaka now I spoke to Allison
19:35 Senaka i think they started some work on it
19:35 Coke which is something I would have recommended doing if time permitted (talking to allison).
19:36 Senaka anyway, I would like to know whether I could find someone involved in SoC to give me a hand on this end too?
19:36 Senaka is Allison into SoC?
19:36 Coke We have a few mentors lined up for the parrot end of summer of code, including myself and particle.
19:37 Coke (and chromatic)
19:37 Coke lemme check the list...
19:37 Senaka OK...
19:37 Coke (chromatic is one of our primary C hackers atm.)
19:37 Senaka Ok
19:37 Senaka would he like to give me a hand?
19:37 Senaka do u think so?
19:38 Senaka I mean on the Parrot end...
19:38 Coke hurm. let me back up a step: this would be an apache SOC project, but you want someone you can coordinate tech questions with once you're underway?
19:38 Senaka yes
19:38 Senaka exactly
19:38 Coke ah. that's a much lower bar that I am sure we can meet, yes. =-)
19:39 Coke seen chromatic?
19:39 purl chromatic was last seen on #parrot 1 day and 14 hours ago, saying: Most of them built for me anyway.  [Mar 28 22:22:18 2008]
19:39 Infinoid LIES
19:39 Senaka :)
19:39 Senaka need some info to start with if you don't mind
19:39 Coke what's your name/email? I can send out a solicitation to the list that will probably get you a few responses.
19:40 Senaka and BTW I ran Parrot on Gutsy
19:40 Senaka cheers...no issues at all
19:40 Coke excelllent.
19:40 Dave :)
19:40 Senaka name: Senaka Fernando, e-mail: senakafdo AT gmail
19:41 Senaka I went thru the Configure.pl
19:41 Senaka now, say for instance I want to change the GC? not use malloc or something, how to do that?
19:42 Senaka I mean is it possible 2 use the -gc option?
19:42 Coke The goal has been for some time to have pluggable GC models. I am uncertain of the current state.
19:42 Senaka hmm... ok
19:42 Coke I am not a C hacker, but I can probably find out the answer to that question.
19:42 Senaka perhaps I can contribute to that fact
19:42 Coke darbelo: haven't forgotten you: any more questions? =-)
19:43 ewilhelm Coke: you must have heard me wrong
19:43 ewilhelm darbelo: I've always said "every 1 week"
19:43 Coke ewilhelm: regarding the 2 weeks thing? very likely. Fire hose, you know.
19:44 Infinoid Senaka: there's a --gc=<foo> option to Configure.pl
19:44 Coke ewilhelm: I'm used to the "industry standard" "no more than 2 weeks", and probably combined them in my head when reading.
19:44 Infinoid possible options of <foo> are: gc, libc, malloc, malloc-trace
19:44 ewilhelm mainly a matter of challenging the student to think through things at a granularity that reveals potential roadblocks
19:44 Coke so, slightly smaller chunks... which makes sense for a project with a mentor. =-)
19:44 Senaka Infinoid: yes I saw that too
19:44 Infinoid config/auto/gc.pm sets up the Makefile to build for each of those
19:45 * Coke asks Infinoid to volunteer so I don't have to send to the list!
19:45 Infinoid what am I being volunteered for?
19:45 Senaka the issue is how could i change the gc altogether?
19:45 Coke ... giving senaka GC related information. Oh look, you've already started. =-)
19:45 Infinoid I think you'd want to add a 5th option
19:45 Senaka like say -gc=bar?
19:45 Infinoid rather than replacing the current ones
19:46 Coke (we already have 4 options? Is anyone testing those! Aigh. =-)
19:46 Infinoid Coke: adding options is easy.  making them *work* isn't. :)
19:47 Infinoid config/auto/gc.pm claims only 1 of the 4 is working
19:47 Senaka Coke: ok I'll take the second
19:47 Senaka aha looks like....yes
19:47 darbelo Ok, so "every seven days a new test shall pass, or I will be kicked in the ass". Sounds good.
19:47 Infinoid Senaka: so to add another option, you'd add another if/else clause to runstep() in config/auto/gc.pm.
19:47 Infinoid then, create a source subdirectory and start fleshing out the API
19:47 Senaka reading
19:48 Senaka reading gc.pm
19:48 Coke looks like src/gc/gc_*.c has 2 different ones there atm.
19:48 Senaka yes
19:48 Senaka got it
19:48 Senaka Infinoid
19:48 purl Infinoid is mailto:mark@glines.org or Mark Glines
19:49 Infinoid or just put it in src/gc/ with the others...
19:49 Senaka purl thanks
19:49 purl pas de quoi Senaka
19:49 Infinoid purl++
19:49 Senaka purl: ddn't get u
19:49 purl Senaka: i'm not following you...
19:49 Senaka :)
19:49 Infinoid purl?
19:49 purl Infinoid?
19:50 Senaka oh!
19:50 Infinoid purl: what are you?
19:50 purl i am probably going alone. or almost an anagram of Donaudampfschiffahrtskapitaensmuetzenkordel or a perv or an auto-triage bot or a she or so stupid or a smartass or the sixth beatle. or http://www.infobot.org or dumb
19:50 Senaka purl: are you a bot?
19:50 purl a bot? yeah right.
19:50 Infinoid (yes)
19:50 Coke bad bot.
19:50 purl :(
19:50 Senaka purl: wow u do good
19:50 purl Senaka: excuse me?
19:51 Senaka :)
19:51 Senaka purl: at least u r faster when it comes to typing
19:51 purl Senaka: sorry...
19:51 Senaka :)
19:51 Senaka Infinoid: where were we?
19:51 purl were we are on stage together
19:52 Infinoid reading gc.pm I believe
19:52 Coke purl, forget were we
19:52 purl Coke: I forgot were we
19:52 Coke so, yes, I think Infinfoid can be your contact. =-)
19:52 Senaka Coke: thanks
19:52 Infinoid I'm a cage cleaner, but I can whack more intelligent people as needed
19:52 Senaka will let the list @ Harmony
19:53 Coke Infinoid++ #for letting himself be volunteered.
19:53 Senaka Infinoid what's your name btw
19:53 Infinoid Mark Glines
19:53 Coke it's in scrollback!
19:53 Coke Infinoid?
19:53 purl Infinoid is mailto:mark@glines.org or Mark Glines
19:53 Senaka ok got it
19:59 Senaka ewihelm: literal is you?
20:00 Senaka ewihelm: I meant are you literal on the #gsoc on freenode?
20:00 Senaka I just want to thank
20:04 ewilhelm Senaka: no, I don't know who that is
20:05 ewilhelm oh, that's Hinrik -- a student
20:05 Senaka ewilhelm: thanks
20:05 Senaka he gave me details about this list
20:10 darbelo Can someone comment on the sanity of separating de c99 preprocessor into it's own grammar.
20:12 Senaka hi anyone here who knows about SoC proposals n stuff?
20:12 darbelo It could make it (almost) testable on it's own and unclutter the c99 grammar, which is good.
20:14 darbelo But it might be more effort than it's worth.
20:15 ewilhelm Senaka: what about them?
20:30 guru I'm trying to track down a windows-specific test failure. t/op/arithmetics.t shows that -0 is not being produced.  Where should I start looking?
20:41 Infinoid guru: I'd start with a simple C application, figure out what hoops you need to go through, to get a float variable set to -0
20:42 Infinoid the code in question is the neg_n op; it boils down to just a function that takes a float and returns its negative
20:42 Infinoid grep src/ops/math.ops for inline op neg(inout NUM)
20:44 Infinoid (assuming its even possible on MSVC.  I have no idea, negative 0 is one of those bizarre C99 spec things I've never tried to understand)
20:53 guru Is src/ops/math.ops in PIR?
20:53 Infinoid its in pseudo-C
20:54 Infinoid it is preprocessed to form core_ops*.c
20:54 guru No wonder I couldn't figure out what I was reading.
20:56 guru left #parrot
20:58 darbelo ewilhelm: SoC has 13 weeks for coding, right?
20:58 darbelo Or am I missing somethig from the schedule?
20:59 Infinoid ewilhelm: Senaka had a question about SoC stuff... its possible that the GC work will expose bugs in Parrot that we'll need to fix up, and was wondering whether/how to mention that
21:00 Infinoid (its an apache SoC project)
21:03 chromatic joined #parrot
21:03 chromatic We won't let any Parrot bugs block successful completion of a project.
21:04 chromatic That is, any bugs we can fix or work around, we will.
21:04 Senaka chromatic: thanks
21:04 chromatic You're welcome.
21:06 Senaka g2g... bye channel... will be back 2moro...
21:06 Senaka thanks for all the help I got from everybody
21:07 Senaka left #parrot
21:17 moha joined #parrot
21:17 moha left #parrot
21:30 Infinoid do flock bits follow inodes, or pathnames?
21:31 Infinoid I hacked c2str.pl to just lock all_cstring.str, instead of creating a .lck file and unlinking it afterwards.  that seems to have fixed "make -j8".
21:32 Infinoid you have 2 c2str.pl processes, one of which holds a lock, the second of which is waiting for the lock, then the first one exits (unlinking the lockfile as it goes), a third one comes up and creates a new lockfile and (of course) obtains the lock, so now the 2nd and 3rd are writing simultaneously.
21:33 Infinoid at least, I think that's what was happening.
21:34 dalek r26644 | infinoid++ | trunk:
21:34 dalek : [c2str] Fix another race, now "make -j8" works reliably for me.
21:34 dalek diff: http://www.parrotvm.org/svn​/parrot/revision?rev=26644
21:34 Auzon joined #parrot
21:35 * Infinoid hopes he didn't just break everyone else.
21:46 chromatic Works for me.
21:49 chromatic Ah, nice work on the seek.
21:51 Infinoid thanks.  That's straight from "perldoc -f flock".
21:56 kj joined #parrot
22:07 wknight8111 joined #parrot
22:09 wknight8111 hey, can somebody fix the problem with the languages/HQ9Plus directory?
22:10 Infinoid wknight8111: looks like bernhard cleaned that up this morning.  are you having an additional issue?
22:10 wknight8111 oh, it's still saying that I can't update my SVN
22:10 wknight8111 I'll try cleaning it all up and trying again
22:11 Infinoid hmm.  is the old directory getting in the way of the update?
22:11 Infinoid maybe just delete all of languages/ and retry the update
22:11 wknight8111 that's what I'm doing now
22:11 wknight8111 stupid windows and case-insensitive path names...
22:12 wknight8111 yay! it's working! now I can update and get back to hacking!
22:12 Infinoid great!
22:14 Infinoid how familiar are you with nmake?  do you know if there's an option to build with multiple processes in parallel?
22:14 Infinoid (something analogous to GNU make's "-jN" option, for N jobs)
22:14 cotto_home Thanks for fixing make -j, Infinoid.  That had been bugging me for some time.
22:15 Infinoid cotto_home: me too.
22:17 cotto_home Have you spent much previous time looking at the build system, or did you just get annoyed and dive in to fix that?
22:17 cotto_home *just to fix
22:17 Infinoid hmm.  I've fought against these Makefiles once before, trying to get splint running.
22:18 wknight8111 let me look around at nmake, see if i can find something like that
22:19 wknight8111 no, it doesnt appear that nmake has a parallel build option
22:19 Infinoid well, great.  that means I haven't broken anything.
22:20 wknight8111 that reassurance is key
22:20 Infinoid I appreciate it. :)
22:21 chromatic joined #parrot
22:22 leo Infinoid: -j5 is currently testing i.e. it seems to have passed build troubles
22:23 Infinoid great, thanks!
22:24 leo just that one again:
22:24 leo Failed Test    Stat Wstat Total Fail  Failed  List of Failed
22:24 leo ----------------------------------------​---------------------------------------
22:24 leo t/library/pg.t               43    1   2.33%  13
22:25 leo but if nobody has fixed that code since, I'm the culprit anyway
22:28 mmcleric joined #parrot
22:29 chromatic Which Pg test is that?
22:38 particle joined #parrot
23:02 dalek r26645 | chromatic++ | trunk:
23:02 dalek : [runtime] Finished moving Test::More and friends to hierarchical namespaces.
23:02 dalek : All tests still pass.
23:02 dalek diff: http://www.parrotvm.org/svn​/parrot/revision?rev=26645
23:03 rrando joined #parrot
23:06 peepsalot joined #parrot
23:08 darbelo particle: I'm a SoC applicant, Coke told me you were the one to talk to regarding my project. It's about improving languages/c99.
23:20 wknight8111 hello darbelo, I'm an SoC applicant too
23:24 wknight8111 I haven't seen particle around today
23:24 wknight8111 see particle?
23:24 wknight8111 seen particle?
23:24 purl particle was last seen on #parrot 2 days and 11 hours ago, saying: harumph, my nicks switched  [Mar 28 05:32:43 2008]
23:24 Auzon particle is here now :P
23:34 Auzon left #parrot
23:35 ewilhelm darbelo: you just missed particle by a few minutes
23:36 ewilhelm probably be ~30 before he's back
23:37 darbelo oh well.
23:37 ewilhelm you have general SoC questions or parrot-specifiC?
23:38 darbelo Kind of both. It was about geting a mentor for my project.
23:39 ewilhelm well, we sort that out if you get in, but it is good if you talk with someone ahead of time
23:39 ewilhelm my guess is that I'm not going to allow either of coke or particle to be (directly) mentors since we have lots of parrot projects for SoC
23:40 ewilhelm so they'll be somewhat busy managing mentors
23:40 darbelo I was without a working net connection for the last few days. Just got it back now. Last time I was here, Coke referred me to particle.
23:42 darbelo I think he was the one who had the most knowledge in the areas I'd have to work with during the project.
23:42 darbelo If I get in, o course ;)
23:42 * ewilhelm also refers you to particle
23:42 ewilhelm but he's at the store atm
23:43 ewilhelm chromatic will be mentoring about 8 GC implementations
23:43 ewilhelm Tene: you know anything about c99?
23:45 darbelo Hm. Is the MAINTAINERS file up to date for c99?
23:45 ewilhelm tewk: ping
23:46 ewilhelm well, tewk is a student, so I hope somebody else has knowledge there
23:47 wknight8111 darbelo, what are you trying to do with the c99 implementation?
23:47 ewilhelm otherwise the two of you will have to have a cage match to determine who is the student and who is the mentor
23:48 darbelo wknight8111: Get it to work, first. Then, do some Sparse-like things with it.
23:49 wknight8111 oh, okay. I heard somebody talking about implementing NCI signatures for it, but I thought that would be a lot of work
23:49 ewilhelm hmm, perhaps we could get josh siglet(?) to mentor
23:49 ewilhelm wknight8111: that would probably be tewk
23:50 wknight8111 ok
23:50 darbelo Compiling c99 seems like too much for the summer. So I'm sticking with parsing, and then Doing Things to the parse trees.
23:52 darbelo I've already submitted it to the web-app, Some of the details are there. But it would benefit from some review, and sanity checking of the schedule.
23:53 ewilhelm triplett that is
23:54 ewilhelm darbelo: it looks pretty good, but I'm just the proposal gestapo
23:57 darbelo Well, my disconnectedness forced me to write it without any kind of reference. I just got back here and transcribed it in a hurry to get in before the 'soft deadline'
23:57 kid51 joined #parrot

| Channels | #parrot index | Today | | Search | Google Search | Plain-Text | summary

Parrot | source cross referenced