Camelia, the Perl 6 bug

IRC log for #parrot, 2010-01-08

Parrot | source cross referenced

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

All times shown according to UTC.

Time Nick Message
00:02 chromatic Makes sense.
00:14 Whiteknight in make whats the difference between = and :=
00:15 chromatic One has a colon and the other... needs a machine.
00:16 darbelo := is nonstandard, but expands the RHS before assigning.
00:16 Coke http://www.gnu.org/software/m​ake/manual/make.html#Flavors
00:17 Coke := ?
00:17 purl rumour has it := is pascal or U+2254
00:17 Coke := is also http://www.gnu.org/software/m​ake/manual/make.html#Flavors
00:17 purl okay, Coke.
00:21 Coke I just did a big ole cut'n'paste in one_make branch if anyone is looking to clean up some perl.
00:29 dukeleto the parrot github repo should be updating regularly again. My machine died while I was on vacation
00:31 dalek parrot: r43408 | coke++ | branches/one_make/tools/dev/checkdepend.pl:
00:31 dalek parrot: Search through .pir files for .include's and try to match them to real files.
00:31 dalek parrot: Don't report errors on this yet.
00:31 dalek parrot: review: http://trac.parrot.org/parrot/changeset/43408/
00:31 dalek parrot: r43409 | coke++ | branches/one_make/tools/dev/checkdepend.pl:
00:31 dalek parrot: Actually report on missing .include dependencies.
00:31 dalek parrot: review: http://trac.parrot.org/parrot/changeset/43409/
00:35 plobsing joined #parrot
00:37 lucian joined #parrot
00:44 bacek joined #parrot
00:53 cotto_work chromatic, is that paper more about how to do it if we could re-do Parrot or are there practical implications that could be applied to Parrot as-is?
00:55 abqar joined #parrot
00:56 chromatic I haven't read it, but it seemed interesting to read and steal.
00:57 cotto_work Yes.  I definitely hope there's something that could be useful to Parrot in there.  It sounds shiny.
00:57 cotto_work Where do you find that stuff?
00:58 chromatic I don't recall.
01:00 cotto_work It's a different concept of vtables than I'm used to with Parrot.
01:04 chromatic Exactly.
01:05 cotto_work very amenable to a "yo dawg"
01:17 Whiteknight which paper?
01:17 cotto_work http://piumarta.com/software/id-objmodel/
01:20 dalek parrot: r43410 | coke++ | branches/one_make/tools/dev/checkdepend.pl:
01:20 dalek parrot: report on missing deps from load_bytecode; look in runtime/parrot/library
01:20 dalek parrot: review: http://trac.parrot.org/parrot/changeset/43410/
01:30 kid51 joined #parrot
01:30 kid51 Coke ping
02:04 mtk joined #parrot
02:04 mtk left #parrot
02:04 jhelwig joined #parrot
03:12 bacek joined #parrot
03:41 bacek joined #parrot
04:07 eternaleye joined #parrot
04:19 theory joined #parrot
04:27 theory joined #parrot
04:47 patspam joined #parrot
04:48 plobsing pmc_freeze_cleanup tests fine everywhere AFAICT, but merging into trunk reveals a threads+GC problem
04:49 plobsing I assume this prevents merge prior to 2.0
04:49 chromatic Unless it's easy to fix.
04:50 plobsing I can change 2 lines to fix it - un-comment the body of stopGC in src/pmc/parrotthread.pmc
04:51 plobsing seems a little like passing the buck though
04:53 chromatic I don't see it in there on trunk.
04:54 chromatic Oh, stop_GC.
04:54 chromatic It's #ifdef 0 on trunk.
04:54 chromatic sorry, #if 0.
04:57 patspam1 joined #parrot
04:58 plobsing yeah, that commenting out was added when STM landed
04:58 chromatic If you remove the #if, does the branch work?
04:59 plobsing I have to regenerate some native_pbc test pbcs to run a couple of tests, but other than that, it appears fine
05:02 * dukeleto is hacking on PL/Parrot at the pdx hackathon
05:04 dalek plparrot: d4cb26f | dukeleto++ | ROADMAP:
05:04 dalek plparrot: Add a ROADMAP. Needs a lot of work
05:04 dalek plparrot: review: http://github.com/leto/plparrot/commit/d​4cb26f3d32efc1dcae583630d378dc7a00a38d8
05:14 eternaleye joined #parrot
05:14 cotto_w0rk joined #parrot
05:25 chromatic Sounds like we need to debug that thread problem then.
05:26 chromatic Maybe something used to be constant but now isn't.
05:27 plobsing Constants in packfiles haven't been created as constants wrt GC since before this branch. it's a workaround for non-constant subs or something like that.
05:30 plobsing running gdb, it appears that a PObj in compact_pool is pointing to an invalid memory location (freed last GC run)
05:34 chromatic Disable your system's stack randomization.
05:35 chromatic Put a breakpoint on whichever function returns a new PMC header.
05:35 chromatic Check that it returns that particular PMC.
05:35 chromatic Then run a backtrace and see what allocates it.
05:59 plobsing The PObj itself is very short lived - created as a copy in String.get_string and handed off for freezing.
06:00 ZeroForce joined #parrot
06:01 nopaste "plobsing" at 67.204.44.186 pasted "Allocation Backtrace" (47 lines) at http://nopaste.snit.ch/19236
06:12 dalek plparrot: c138e6b | dukeleto++ | ROADMAP:
06:12 dalek plparrot: Add some notes from eggyknap
06:12 dalek plparrot: review: http://github.com/leto/plparrot/commit/c​138e6baa8d032fd636c69c02eb7ffc626e4bffb
06:18 plobsing I think I have a grasp on what's going on now
06:18 plobsing A PObj is being COWed across threads that don't seem to share memory pools, so when one compacts, the other gets hosed.
06:19 plobsing This doesn't show up in trunk because pmc_freeze_cleanup generates larger images, therefor creating more GC pressure
06:19 plobsing does that make sense?
06:34 dalek TT #1404 created by aardvark++: [PATCH] eliminate ambiguous term PIR book, chapter 3
06:41 chromatic That makes sense.
06:42 chromatic We should check ownership when COWing across threads.
06:45 Coke .
06:49 * plobsing sleeps
06:52 dalek plparrot: 0fb3bd3 | dukeleto++ | ROADMAP:
06:52 dalek plparrot: More ROADMAP updates
06:52 dalek plparrot: review: http://github.com/leto/plparrot/commit/0​fb3bd380e897348c167193be297a781b4136146
06:52 dalek plparrot: e55bf0c | dukeleto++ | ROADMAP:
06:52 dalek plparrot: More maps of roads
06:52 dalek plparrot: review: http://github.com/leto/plparrot/commit/e​55bf0c83d4768947ad000db6b7e405d44f4b796
07:00 uniejo joined #parrot
07:09 dalek plparrot: 81dc5bc | (Joshua Tolley)++ | ROADMAP:
07:09 dalek plparrot: Add notes to roadmap
07:09 dalek plparrot: review: http://github.com/leto/plparrot/commit/8​1dc5bc41b026698ed28cf2477ce575e39402e8f
07:12 bacek joined #parrot
07:12 eternaleye joined #parrot
07:13 theory joined #parrot
07:19 fperrad joined #parrot
07:55 abqar joined #parrot
07:59 cognominal joined #parrot
08:12 dalek plparrot: 821e790 | dukeleto++ | src/handler/plparrot.c:
08:12 dalek plparrot: Add the beginnings of something that does SPI
08:12 dalek plparrot: review: http://github.com/leto/plparrot/commit/8​21e790e9f14828382c787ca907ef4c7c86ff1de
08:15 eternaleye joined #parrot
08:37 iblechbot joined #parrot
08:40 bacek joined #parrot
09:38 dalek plparrot: c9755c5 | dukeleto++ | src/handler/plparrot.c:
09:38 dalek plparrot: Add some error checking for SPI_connect
09:38 dalek plparrot: review: http://github.com/leto/plparrot/commit/c​9755c5c1595ee3c726efb4221beb21dd2572ac9
09:38 dalek plparrot: 6a5a19c | dukeleto++ | src/handler/plparrot.c:
09:38 dalek plparrot: Improved error handling for SPI_connect
09:38 dalek plparrot: review: http://github.com/leto/plparrot/commit/6​a5a19c1c400d70a557b04ed952fddbe33a6b0c7
09:38 dalek plparrot: ae90d19 | dukeleto++ | src/handler/plparrot.c:
09:38 dalek plparrot: Add some more SPI boilerplate
09:38 dalek plparrot: review: http://github.com/leto/plparrot/commit/a​e90d1955d276a0b00f13bbce926f9670683bc4a
10:01 cognominal joined #parrot
10:10 payload joined #parrot
10:38 bacek joined #parrot
11:03 bacek joined #parrot
11:36 patspam joined #parrot
12:25 ruoso joined #parrot
12:34 payload joined #parrot
12:50 mariano__ joined #parrot
12:57 tetragon joined #parrot
13:27 cognominal joined #parrot
13:43 fperrad_ joined #parrot
13:52 payload joined #parrot
14:12 dalek TT #1404 closed by coke++: [PATCH] eliminate ambiguous term PIR book, chapter 3
14:13 ttbot Parrot trunk/ r43411 i386-linux-thread-multi make error http://tt.ro.vutbr.cz/file/cmdout/180795.txt ( http://tt.ro.vutbr.cz//buil​dstatus/pr-Parrot/rp-trunk/ )
14:21 dalek parrot: r43411 | coke++ | trunk/docs/book/pir/ch03_basic_syntax.pod:
14:21 dalek parrot: Documention cleanup courtesy aardvark++ (TT#1404)
14:21 dalek parrot: review: http://trac.parrot.org/parrot/changeset/43411/
14:34 payload joined #parrot
15:12 payload joined #parrot
15:21 mariano__ joined #parrot
15:39 iblechbot joined #parrot
15:41 ash_ joined #parrot
15:45 Psyche^ joined #parrot
15:48 theory joined #parrot
15:50 whiteknight joined #parrot
16:00 payload joined #parrot
16:13 cognominal joined #parrot
16:21 hercynium joined #parrot
16:29 whiteknight Parrot builds with icc on Linux32, but fails a few random tests. Mostly math stuff
16:29 whiteknight Parrot builds with clang on Linux32, passes all tests I think
16:30 whiteknight also passes all tests with gcc and g++, at least as of the last time I ran those tests
16:31 whiteknight I haven't tried any of these with --optimize
16:33 whiteknight smolder is stupid slow today
16:34 whiteknight most ICC failures I can see look like -0 problems, so those are minor (annoying, but minor)
16:35 whiteknight It also has two failed GC assertions in t/compilers/pct/complete_workflow.t, which are troubling
16:43 Coke that test has been failing on and off for some time.
16:43 Coke (in one_make)
16:50 ZeroForce joined #parrot
17:06 ash_ is there a target for parrot that only builds the libraries and not the parrot executable?
17:07 PerlJam ash_: make help # should tell you
17:08 darbelo ash_: Not as such, but it'd be pretty easy to add.
17:08 darbelo look in the makefile for LIBPARROT
17:09 ash_ i see there is  a PARROT_LIBS
17:10 ash_ if you were going to embed parrot into another program, all you would need is the PARROT_LIBS right? (it builds static and dynamic librarys)
17:13 darbelo Yeah, but you'll want to look at the embedding docs too.
17:13 darbelo docs/embed.pod should be helpful.
17:13 Coke there is no single target for that, no.
17:14 Coke if you happen to know the name of the library file you wish to generate, you can of course just make that one file.
17:14 ash_ kk, thanks, still trying to figure out if I can make parrot as a linkable library for an iphone app
17:18 ash_ Coke: so i shouldn't be doing make PARROT_LIBS, but rather make blib/lib/libparrot.a ?
17:19 Coke I do not see that target in root.in, building makefile...
17:20 Coke oh, you can make a varname?
17:20 Coke huh.
17:20 ash_ it worked, i don't know if its the right way or not
17:24 ash_ if i want to link to the library from an iphone app it has to be statically linked for whatever reason, so i only really need the LIBPARROT_STATIC target
17:35 Essobi joined #parrot
17:51 Coke so, does gnat write perl anymore or just send links to boingboing? =-)
18:04 Tene cognominal: Oh, you're here already!
18:05 theory Coke: He's a professional maven.
18:06 cognominal yea, but I think my question is more appropriate to #perl6 :)
18:11 cotto_w0rk ignore piper?
18:14 Coke our usage of .include makes reporting on dependencies annoying. :|
18:22 dukeleto 'ello
18:40 mj41 joined #parrot
18:40 dukeleto mj41: hola
18:40 purl privet, dukeleto.
18:45 mj41 dukeleto: ahoj
19:25 whiteknight joined #parrot
19:28 whiteknight Coke: how do I check the output of the config step, --verbose?
19:31 Coke --verbose-step       Comma-delimited string of configuration steps
19:31 Coke where the step is prolly auto::neg_0
19:31 whiteknight okay, I'll have to try that later
19:32 whiteknight OpenSolaris/gcc3.4.5 was failing the neg0 tests, and Ubuntu/icc11.1 was too this morning
19:32 whiteknight and I seem to remember we were seeing some kind of similar issues on a windows configuration
19:33 Coke it's possible that the config step is not dealing with some aspect of -0 that you need.
19:33 Coke I'm pretty sure it tries to generate a -0 and then just print it to see if it comes out 0 or -0
19:33 Coke I'm not sure what aspect the test failures you're seeing is dealing with.
19:34 whiteknight yeah, probably need to expand that to checking for -0 results from operations
19:34 whiteknight if you look through the smolder reports, mine are the ones that use either ICC or OpenSolaris/i386
19:34 whiteknight I'll dig out some links if you need (or I'll just submit new reports in an hour or so)
19:35 whiteknight op/arithmetics.t, op/trans.t and op/sprintf.t were all failing if I remember correctly
19:36 chromatic joined #parrot
19:36 dukeleto chromatic: good moroning
19:36 chromatic morning
19:37 dukeleto chromatic: thanks for the mentoring link. i need to get on that
19:38 chromatic We could find a few good students.
19:39 dukeleto chromatic: definitely. I was talking to a prospective gsoc student at the hackathon last night
19:41 ash_ i am a student, well, i graduate in spring, and then i am applying for grad school, but i think that still counts over the summer
19:41 PerlJam ash_: it does IIRC
19:46 whiteknight joined #parrot
20:04 joeri joined #parrot
20:07 cognominal joined #parrot
20:10 dukeleto ash_: sounds good! start thinking about projects
20:10 ash_ i already know of 1 i am interested in, lorito and compiler optimizations
20:11 PerlJam ash_++
20:11 cotto_work Hmm.  We'll have to get hacking on Lorito so you've got something to work on then. ;)
20:11 dukeleto ash_: sounds good. just make sure you have a well-defined plan that can be accomplished in 3-4 months
20:13 ash_ that, or building a stack frame builder that uses the llvm
20:13 ash_ i know someone made one that uses libjit at one point
20:14 ash_ i think it would be interesting to see a comparision of the two
20:14 Zak joined #parrot
20:16 zak_ joined #parrot
20:17 dukeleto ash_: we really need a frame builder
20:17 dukeleto ash_: plobsing is working on that, I think
20:18 ash_ yeah, thats the libjit one, although the github repo i was watching hasn't been updated since november
20:20 ash_ thats weird. when building in xcode for the iPhone, the MACOSX_DEPLOYMENT_TARGET environmental variable is set to 10.5 even though I am on 10.6, wonder why that is...
20:22 patspam joined #parrot
20:23 hercynium joined #parrot
20:24 Coke ... because the iphone is only running 10.5?
20:27 bacek joined #parrot
20:27 ash_ no, i found it, its a project preference
20:40 ash_ well, i have it building for the simulator now... but targeting the actual device is not working, i am going to have to figure out some way to fudge the makefile or something
20:44 zak__ joined #parrot
21:01 zak_ joined #parrot
21:08 ash_ So... my plan's kind of this, build parrot for the host system (so you can make pbc files if needed) and then have a second build phase that builds for the target system and supplements any uses of parrot with the host system's parrot instead of the target system's parrot
21:13 darbelo ash_: You need to watch out for host information bleeding into the target files.
21:14 ash_ maybe, i could setup some way for it to build the pbc's on the iphone after its run?
21:15 dukeleto ash_: sounds cool. this could be a gsoc project
21:16 zak__ joined #parrot
21:16 ash_ dukeleto: thats no fun, i'm already working on it
21:16 ash_ :-d
21:21 darbelo Starting early is the key to success ;)
21:23 ash_ what exactly are the requirements for a gsoc anyway? i have been interested in them but i don't know the details
21:24 darbelo Be a student.
21:24 darbelo Write a proposal.
21:24 darbelo Get accepted.
21:24 darbelo Profit!
21:24 ash_ you forgot the ... step
21:25 darbelo I did.
21:25 darbelo http://socghop.appspot.com/ has more data.
21:26 darbelo The 'Elelgibility' section of the FAQ is probably what you want.
21:26 darbelo s/Elelgibility/Eligibility/
21:29 ash_ i don't see why i wouldn't meet the requirements for that then
21:30 darbelo ash_: Go for it!
21:30 darbelo Well, when they announce the '10 edition.
21:34 ash_ i made a profile on that site
21:34 ash_ but yeah, lots of that is still 2009 stuff
21:35 ash_ so i can't apply till march then? it says student applications don't start till then
21:38 ash_ are there any other student developer things if I wanted to participate in those? or should I just keep toying with things, like trying to run parrot on the iPhone,
21:41 * darbelo back
21:42 darbelo ash_: Actaully there hasn't been any official word of a GSoC'10.
21:42 darbelo They could just not do it, but that's rather unlikely.
21:43 dukeleto ash_: this is the time to work on your project ideas
21:43 dukeleto ash_: what do you mean by "student developer things" "?
21:43 ash_ things like the gsoc, i was just wondering if there are other things like that out there that i may not be aware of
21:44 darbelo If there are, I'm not aware of the either.
21:44 dukeleto ash_: there are some $season of code things from a few places, but they are more specialized
21:46 ash_ one of my professors asked if i wanted to write a proposal on http://eng.auburn.edu/admin/marketing/news​room/2009/october/mark-spencer-award.html last semester, so I should be finding out about that soon (end of january) i was going to do llvm stuff, but now I am kinda thinking of working with parrot more (or merging the two if i can every figure out how the stack frame builder works)
21:47 darbelo plobsing did a lot of work on the frame builder, he might be of help.
21:48 darbelo s/did a lot of work/rewrote/
22:01 ash_ if I wanted to try working on the stack frame builder for that grant, do you think anyone would mind?
22:02 darbelo All of us would mind. We are all against the improvement of parrot.
22:02 darbelo ;)
22:03 darbelo Rule of thimb: If it makes parrot better, we're all for it.
22:03 darbelo s/thimb/thumb/
22:06 ash_ well, looks like i am going to revise my original proposal for my professor and see what he says
22:07 japhb joined #parrot
22:15 zak_ joined #parrot
22:29 ash_ does nci world against objective c libraries?
22:29 integral joined #parrot
22:30 Tene ash_: it works against anything you can dlopen
22:31 ash_ okay, cool, just curious
22:58 redbrain joined #parrot
23:28 mariano__ joined #parrot
23:56 zak__ joined #parrot

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

Parrot | source cross referenced