Camelia, the Perl 6 bug

IRC log for #parrot, 2009-09-11

Parrot | source cross referenced

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

All times shown according to UTC.

Time Nick Message
00:07 Whiteknight joined #parrot
00:09 darbelo Whiteknight: I been waiting for you. I have a ton of patches today. ;)
00:12 cotto smolder?
00:12 purl smolder is http://sourceforge.net/projects/smolder or web-based smoke test aggregator used by developers and testers to upload (automated or manually) and view smoke/regression tests using the Test Anything Protocol (TAP). or http://smolder.plusthree.com/app​/public_projects/smoke_reports/8
00:12 * duk3leto watches as darbelo dumps a truck-load of patches onto Whiteknight's front lawn
00:12 Whiteknight yay! patches
00:12 chromatic Ding and dash, ding and dash!
00:13 ruoso joined #parrot
00:13 ruoso_ joined #parrot
00:13 darbelo Whiteknight: Actually, I was just messing with you.
00:14 * darbelo got his bit today, he can commit his own patches now.
00:14 Whiteknight yay! darbelo++
00:14 * kid51 looks through old TTs to see what he can dump on darbelo ;-)
00:14 duk3leto darbelo++ # congrats on your bit
00:16 * darbelo preemptively dumps TT#986 on kid51
00:16 cotto kid51, ping
00:16 kid51 darbelo:  I'm not sure I know how to solve anything about that one.
00:16 kid51 cotto pong
00:17 cotto nm.  I was going to ask for a specific smolder failure, but I remember that you put one in your message to the list.
00:18 darbelo kid51: I'm about to commit that, actaully. But I had to pull a number from somewhere :)
00:18 kid51 go to smolder recent reports and filter on tag "darwin"
00:18 kid51 You'll see that the issue is probably more PPC than it is darwin.
00:19 cotto the thing is that the code in that file shouldn't even be run except as part of the profiling runcore.
00:21 cotto I suspect the merge and hires_timer stuff is a red herring.
00:22 cotto or a more indirect cause
00:23 kid51 cotto:  I'm getting git-trained in another channel now, so I can't chat very much about that now.
00:23 cotto np
00:23 kid51 Can u post to list or in appropriate TT?  thx
00:26 Whiteknight what is hires_timer?
00:26 dalek parrot: r41197 | darbelo++ | trunk/compilers (2 files):
00:26 dalek parrot: Remove more strstart uses, this time from pirc and imcc.
00:26 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41197/
00:26 Whiteknight oh, high-resolution
00:29 Whiteknight what is sig10 on darwin?
00:30 Coke SIGBUS
00:30 Whiteknight oh, ouch
00:31 mikehh rakudo (5960161) builds on parrot r411796 - make test / make spectest (up to r28216) PASS - Ubuntu 9.04 amd64 (g++)
00:31 mikehh rakudo - t/spec/S03-operators/arith.rakudo - TODO passed:   120, 131-132
00:31 mikehh rakudo - t/spec/S02-names_and_variables/perl - Non-zero wait status: 11 (Segfault after passing tests)
00:31 dalek TT #986 closed by darbelo++: Further ->strstart removals
00:32 Coke mikehh: I am getting a segfault on tclsh when running in interactive mode; I wonder if that's the segfault you're seeing on sporadic test files.
00:36 * kid51 got run over by SIGBUS many times this week
00:36 mikehh coke: they seem relatively random to me - I run test make test a few times and I get some repeats but some one offs and noting that fails every time
00:37 Coke kid51: Welcome to my world!
00:37 Whiteknight can you try it out in gdb and try to get a backtrace?
00:37 * Whiteknight needs to get a darwin box to develop on, eventually
00:37 cotto kid51, I hate to seem like I'm dumping this off, but I really don't see a way the profiling runcore could be causing those problems.
00:38 NotFound I'm working in some problems with PackFile_Segments, specifically in PackFile_Directory that may be the culprit of most segfaults, but I'm too tired now :O
00:38 cotto chromatic++ #running rakudo hello world with -Rprofiling under Valgrind got ~25% faster.
00:39 Whiteknight wow, nice
00:39 mikehh NotFound: feel a little that way myself - a lot of $work poroblems today
00:39 Whiteknight I can't wait to start seeing some of the data from that
00:39 chromatic That's not even the faster part.
00:40 mikehh NotFound: well yesterday for an hour or so
00:40 RobertLJ joined #parrot
00:40 nopaste "NotFound" at 213.96.228.50 pasted "patch: Working on packfile problems" (42 lines) at http://nopaste.snit.ch/17915
00:40 cotto I still need to replace Parrot_Context_get_info.  It's taking ~86% of the time.
00:40 Coke cotto: what version of cg_annotate are you using?
00:40 chromatic I'm working on that now, cotto.
00:40 NotFound Here is, if someone want to look at it.
00:40 Whiteknight NotFound: me!
00:41 NotFound Still has problems, but shows the sources of problems.
00:41 cotto Coke, I haven't tried the output with cg_annotate.
00:42 Whiteknight so many of these slowdowns would disappear if GCC could optimize across compilation units
00:42 NotFound Whiteknight: good luck, I'm going to bed.
00:42 Whiteknight because all those context API functions could easily be inlined
00:42 cotto Coke, Also, it's likely to be goofy because I'm stuck getting my line and file information from imcc.
00:43 Coke at this point, cg_annotate fails for me entirely.
00:43 Coke what are you using?
00:43 cotto kcachegrind
00:43 purl kcachegrind is GUI? or "me and the kache be bumpin' and grindin'"
00:43 hercynium_ joined #parrot
00:45 cotto Coke, it seems to work for me with callgrind_annotate-3.4.1-Debian
00:45 Whiteknight Also, internally we should be able to break encapsulation on the Context PMC. The API is only really necessary for external poking
00:46 Whiteknight in src/sub.c, I don't see any reason why we should be accessing important context information through API calls
00:46 Coke now that I'm back on my os x box, I see I tried to get kcachegrind working at one point.
00:56 Coke won't start up, tough.
00:56 chromatic cotto, I know what to do in src/sub.c.  I'm less sure what we need to do in src/runcore/profiling.c.
00:56 Coke anyone familiar with open's new modes that could map all of http://www.tcl.tk/man/tcl8.5/TclCmd/open.htm#M6 to them?
00:56 chromatic Can I give you Parrot_sub_get_line_from_pc and Parrot_sub_get_filename_from_pc?
00:57 Coke the + modes from tcl don't seem to have direct analoges.
00:57 Coke "analogues"
00:57 cotto chromatic, sure
00:58 * darbelo departs, driven to the hunt for food.
00:58 darbelo See y'all later.
00:58 darbelo left #parrot
00:58 chromatic Okay, hang on.
00:58 purl hang on. is this actually "session is still there but user has been deleted" ?
01:03 Whiteknight purl forget hang on
01:03 purl Whiteknight, I didn't have anything matching hang on
01:03 Whiteknight purl forget hang on.
01:03 purl Whiteknight, I didn't have anything matching hang on
01:03 Whiteknight purl you're retarded
01:03 purl Whiteknight: what?
01:03 Whiteknight you heard me
01:03 ZeroForce joined #parrot
01:04 chromatic purl, forget hang on.
01:04 purl chromatic, I didn't have anything matching hang on
01:05 Coke forget okay, hang on.
01:05 purl Coke, I didn't have anything matching okay, hang on
01:06 Austin hang on
01:06 Austin okay, hang on
01:06 Austin Okay, hang on.
01:06 purl it has been said that hang on. is this actually "session is still there but user has been deleted" ?
01:06 Austin purl, forget 'hang on.'
01:06 purl Austin, I didn't have anything matching 'hang on.'
01:06 Austin purl forget hang on.
01:06 purl Austin, I didn't have anything matching hang on
01:06 Austin sigh
01:07 Coke cotto: if I run kcachegrind /tmp/parrot.out.32261 on OSX, it just hangs. ah well.
01:08 dalek partcl: r722 | coke++ | trunk/runtime/builtin/open.pir:
01:08 dalek partcl: cleanup PIR
01:08 dalek partcl: review: http://code.google.com/p/p​artcl/source/detail?r=722
01:08 dalek partcl: r723 | coke++ | trunk/runtime/builtin/open.pir:
01:08 dalek partcl: add initial support for more short access modes.
01:08 dalek partcl: review: http://code.google.com/p/p​artcl/source/detail?r=723
01:15 chromatic Oh joy, I pushed a bunch of .gitignore files.
01:15 dalek parrot: r41198 | chromatic++ | trunk/src/runcore/profiling.c:
01:15 dalek parrot: [runcore] Tidied code in the profiling runcore; no functional changes.
01:15 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41198/
01:15 dalek parrot: r41199 | chromatic++ | trunk (109 files):
01:15 dalek parrot: [sub] Added helper functions Parrot_Sub_get_line_from_pc() and
01:15 dalek parrot: Parrot_sub_get_filename_from_pc(), intended for a lighter weight version of
01:15 dalek parrot: Parrot_Context_get_info().
01:15 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41199/
01:19 dalek partcl: r724 | coke++ | trunk/runtime/builtin/open.pir:
01:19 dalek partcl: fix [open] with no access mode specified.
01:19 dalek partcl: review: http://code.google.com/p/p​artcl/source/detail?r=724
01:23 TiMBuS joined #parrot
01:23 chromatic cotto, r41199 should get you where you need to go.
01:44 JimmyZ joined #parrot
01:48 sri_ joined #parrot
01:50 cotto chromatic, great.  I'll use that as soon sa I have some tuits
02:00 bobke joined #parrot
02:11 payload joined #parrot
02:21 dalek partcl: r725 | coke++ | trunk/ (4 files):
02:21 dalek partcl: make tcl_error opcodes more like [error] syntax.
02:21 dalek partcl: Simplify [error], and update all usages of the 2-arg version to be 3-arg.
02:21 dalek partcl: review: http://code.google.com/p/p​artcl/source/detail?r=725
02:22 JimmyZ joined #parrot
02:35 Coke I am getting an error "object must be created by class" trying to do a VTABLE_assign_pmc .
02:37 Coke looks like Parrot_scalar_assign_pmc
02:38 dukeleto joined #parrot
02:42 janus joined #parrot
02:58 dalek TT #999 created by coke++: convert some base_type to VTABLE lookups.
03:27 sri joined #parrot
03:41 Tene ONE MORE TO ONE THOUSAND!
03:53 jdv79 Coke: http://nopaste.com/p/ay5paS2Lgb
04:03 Coke jdv79: nifty; can you open a ticket at partcl.googlecode.com, i'll get it applied tomorrow. zzzz.
04:05 jdv79 k,nite
04:20 dukeleto chromatic: you can use .git/info/exclude as a place to store .gitignore stuff without worrying about it getting committed
04:26 bacek "git commit -a" considered harmful :)
04:28 Ryan52 or add .gitignore to .gitignore (iirc that works)
04:29 chromatic I didn't git commit -a... I think git svn git-ignore did it automatically.
04:34 Tene bacek: -a won't add new files.
04:40 kyle_l5l joined #parrot
04:48 bacek Oookey. "*svn*" considered harmful :)
04:50 Zak joined #parrot
04:50 Tene bacek: I'm reminded of http://meyerweb.com/eric/comment/chech.html
04:55 bacek Tene: yeah. I'm totally agree that "Considered Harmful" considered harmful :)
04:56 Tene :)
05:01 iblechbot joined #parrot
05:13 jdv79 gmake[1]: *** [PGE.pbc] Segmentation fault
05:13 jdv79 anyone else getting that?
05:14 shockwave joined #parrot
05:16 dukeleto jdv79: did you do a realclean?
05:16 _shockwave joined #parrot
05:20 jdv79 yeah.  its a vigin box.  maybe something random is off...
05:20 jdv79 *virgin
05:26 jdv79 hmm, got me.  any ideas?
05:34 dukeleto jdv79: full nopaste output would be useful
05:43 jdv79 http://nopaste.com/p/apw2xEb1ib
05:49 sundar joined #parrot
05:50 sundar Hi.. I'm having some issues compiling parrot under 64-bit Windows. Is there any guide or anything like that specifically for Windows?
05:52 shockwave sundar, I was able to do it flawlessly after installing strawberry perl.
05:53 jdv79 joined #parrot
05:54 sundar shockwave: Oh. I'm using Activestate's 64 bit perl (since mine is a 64 bit machine).
05:55 sundar Specifically, when I try to do 'perl Configure.pl', I get 'test_4644.exe has stopped working' error for most of the lines.
05:56 shockwave sundar, I just started using Parrot myself, so I can't really help troubleshoot your problem other than answering your initial questions.
05:56 shockwave question*.
05:57 sundar shockwave: Ok... I was under the impression that this is some problem with Configure.pl detecting my C compiler. Will give Strawberry perl a try anyway.
05:58 shockwave cool
05:59 uniejo joined #parrot
06:05 dalek rakudo: a0494f8 | pmichaud++ | build/PARROT_REVISION:
06:05 dalek rakudo: Bump PARROT_REVISION.
06:06 dalek rakudo: review: http://github.com/rakudo/rakudo/commit/a​0494f8a4576694437cc8aa933a8f5f6ff29c5e2
06:07 jdv79 http://nopaste.com/p/ahhiotQTb - there's a larger dump
06:11 japhb joined #parrot
06:39 fperrad joined #parrot
06:54 HG` joined #parrot
07:28 mj41 joined #parrot
07:40 mikehh joined #parrot
07:40 mikehh back on i386
07:53 kjeldahl joined #parrot
08:06 chromatic NotFound, bacek, you might find r41200 interesting.
08:07 dalek parrot: r41200 | chromatic++ | trunk/src (3 files):
08:07 dalek parrot: [pcc] Added context struct initialization to Parrot_pcc_set_sub(), specifically
08:07 dalek parrot: the program counter, current HLL, and current namespace.  The Sub and Coroutine
08:07 dalek parrot: PMCs already did this directly in their invokes, so moving this code here gets
08:07 dalek parrot: rid of a few function calls.  There's a modest (0.733%) performance improvement
08:07 dalek parrot: in the oofib PIR benchmark, but the biggest benefit is removing duplicate code.
08:07 dalek parrot: There's more to go; perhaps the lex pad initialization code can move here too.
08:07 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41200/
08:38 dukeleto joined #parrot
08:53 dalek parrot: r41201 | mikehh++ | trunk/MANIFEST:
08:53 dalek parrot: manifest_tests failure - run tools/dev/mk_manifest_and_skip.pl
08:53 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41201/
08:53 dukeleto jdv79: do you have an installed parrot 1.5 ? sometimes odd things happen when you are building a newer version of the same parrot release that you have installed
08:53 dukeleto jdv79: there should be a Configure.pl-time warning about it. if not, that is a bug
08:53 dukeleto jdv79: which OS is this happening on?
09:00 dalek parrot: r41202 | mikehh++ | trunk (106 files):
09:00 dalek parrot: set svn properties on .gitignore files
09:00 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41202/
09:00 dukeleto oy vey
09:03 mikehh I am not sure that was a good idea - I am now getting all those files want more properties set
09:03 moritz we have .gitignore files?
09:03 moritz why?
09:05 mikehh presumably for the git svn folk
09:05 moritz they can use git-svn create-ignore or .git/info/exclude
09:06 moritz anyway, I was just wandering, not complaining :-)
09:07 mikehh I was just trying to fix some codetest/manifest+tests failures and I think another approach is needed here
09:16 mikehh the got added by chromatic at r41199, I am not sure how we should deal with manifest/svn properties on them
09:17 payload joined #parrot
09:24 mokurai left #parrot
09:27 mikehh I somehow don't think those files should actually be in the MANIFEST
09:32 MoC joined #parrot
09:39 masak joined #parrot
09:44 mikehh I think I am going to undo those changes and we need to discuss this
09:52 bacek_ joined #parrot
09:56 bacek_ o hai
09:56 masak o/
09:58 Whiteknight joined #parrot
09:59 dalek parrot: r41203 | mikehh++ | trunk (107 files):
09:59 dalek parrot: undo the last two commits - I think this needs discussion
09:59 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41203/
10:01 * Whiteknight pokes his head in for 5 minutes
10:05 mikehh Ok - anybody any ideas on what should be done with the .gitignore files introduced in r41199
10:06 bacek_ mikehh: they should be removed.
10:12 mikehh bacek_: we need some mechanism to either ignore them or prevent them from being re-introduced
10:13 bacek_ mikehh: I know only one good mechanism for this. But "The Architect" against it.
10:15 dalek TT #863 closed by fperrad++: GC Segfault
10:16 mikehh bacek_: if chromatic has them set up on his system we need for svn to ignore them or something
10:17 mikehh and I would be reasonably happy with git with some reservations
10:17 bacek_ "some reservations"?
10:18 mikehh actually I really don't care which version control system we use as long as it is consistent
10:19 mikehh and NOT cvs
10:20 bacek_ Why not CVS? It's "mature, robust and consistent"! Consistently bad actually :)
10:20 mikehh I have used hg, bzr, git and svn on a lot of projects - it is just remembering all the in-and-outs
10:20 bacek_ Yay! I finally helped fperrad++ with something broken in parrot :)
10:21 mikehh I personally would opt for bazaar - but as long as it does what we need
10:22 bacek_ mikehh: there is one main git feature that beats everything. It's bloody fast.
10:23 bacek_ I usually commit every few lines of code and then rebase it to provide "perfect history". But it's my workflow. I don't try to force everyone else to use it.
10:24 mikehh I still think it's a lot more difficult to keep track of where you are in git, but that's just me
10:26 mikehh I have dealt in the past with mostly centralised systems and distributed ones sometimes get me confused
10:26 bacek_ They are... just different.
10:27 bacek_ It's like using Perl after BASIC :)
10:27 mikehh actually I have no major problems with perl or rakudo
10:28 mikehh I'll have you know I taught a few courses in BASIC once upon a time - before Perl even existed :-}
10:29 bacek_ I did them too :)
10:29 bacek_ Ok, time to make some dinner.
10:29 bacek_ See you soon.
11:01 ttbot joined #parrot
11:02 cognominal joined #parrot
11:25 bacek_ joined #parrot
11:35 bacek_ msg chromatic I think "clone" Context for Sub/Continuation should be properly encapsulated into some subroutine.
11:35 purl Message for chromatic stored.
11:40 szabgab joined #parrot
11:42 szabgab joined #parrot
11:47 mj41 joined #parrot
11:57 payload joined #parrot
12:22 dalek parrot: r41204 | bacek++ | trunk/src/pmc/scalar.pmc:
12:22 dalek parrot: [cage] Apply patch from TT#999 to use more VTABLE_type. Coke++
12:22 dalek parrot: Also drop explicit store of const strings.
12:22 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41204/
12:24 dalek TT #999 closed by bacek++: convert some base_type to VTABLE lookups.
12:29 dalek parrot: r41205 | coke++ | trunk/docs/project/release_manager_guide.pod:
12:29 dalek parrot: Attempt to make the google calendar the canonical calendar.
12:29 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41205/
12:29 dalek parrot: r41206 | coke++ | trunk (107 files):
12:29 dalek parrot: Remove the .gitignore files.
12:29 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41206/
12:31 Coke there. we can think about how we want to make life easier for the git folk later.
12:31 Coke (in partcl, I just went ahead and committed them, but I have no coding standards tests nor a large dev base. =-)
12:40 bacek_ Coke: so you can just migrate to git easily :)
12:48 quek joined #parrot
12:52 Coke bacek_: yup. if googlecode ever offers it, I'll probably switch.
12:53 dalek partcl: r726 | coke++ | trunk/ (2 files):
12:53 dalek partcl: Apply patch from jdv to enable smolder reporting for partcl!
12:53 dalek partcl: Update issue #109
12:53 dalek partcl: review: http://code.google.com/p/p​artcl/source/detail?r=726
12:53 Coke IWBNI 'parrot_config' reported the parrot version.
12:53 Coke (you can figure it out from versiondir, but ew.)
12:55 * Coke hopes NotFound++ works more magic and fixes all these segfaults. :|
12:57 Coke we currently have 21 open tickets that mention a segfault in trac alone.
12:57 Coke (more in RT)
12:57 bacek_ Coke: he didn't fixed all of them yet? Bad NotFound :)
12:58 NotFound Coke: I have a version now that pass make test without any segfault, maybe leaking memory though
13:02 ruoso joined #parrot
13:05 NotFound That's really funny, a packfile segment destoy function pointing to Parrot_default_subtract :o
13:07 dalek TT #154 closed by coke++: subclassed ExceptionHandlers can't handle exceptions
13:08 whiteknight joined #parrot
13:11 dalek TT #832 closed by coke++: segfault in _int_malloc
13:11 dalek TT #321 closed by coke++: Segfault overriding init sub in PIR
13:14 dalek TT #479 closed by coke++: Segfaulting in PGE;Perl6Grammar;Compiler;compile ...
13:14 dalek TT #729 closed by coke++: Segfault on printing a String PMC assigned a null string
13:17 Coke there, now only real segfaults (or invalid ones that are too complex to verify) are left.
13:19 dalek TT #737 closed by coke++: imcc segfault when assigning to undefined retval
13:25 NotFound Coke: I left open some of that waiting for confirmation. I suppose that several months without answer is enough ;)
13:26 Coke NotFound: ayup.
13:26 Coke NotFound++
13:27 Coke is https://trac.parrot.org/parrot/ticket/995 the one you're working on now?
13:27 Coke that's the biggest one for partcl atm.
13:27 Coke (and might also be the root cause of rakudo's issues.)
13:27 Coke I'm happy to test out a patch. =-)
13:28 dalek parrot: r41207 | mikehh++ | trunk/src/pmc/scalar.pmc:
13:28 dalek parrot: codetest failure - there should be one space or a newline after a comma
13:28 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41207/
13:28 NotFound Coke: yes, but I think the problem is the same for lots of segfaults.
13:29 NotFound The PackFile structures is a little nightmare, that's the root of the problems.
13:29 Coke if there's a significant memory leak, I'm sure partcl will trip over it.
13:29 NotFound No idea on how significant can be, I left that for valgrind knowers
13:30 Coke k. ping me when It's in or there's a patch I can try. =-)
13:31 Coke (I'll try out all my segfaults and then kick off a spectest run)
13:31 NotFound Coke: I'll commit it if a test I'm doing goes well. If not, I'll put the patch in TT #995
13:44 JimmyZ joined #parrot
13:47 dalek TT #1000 created by flh++: t/pmc/resizablefloatarray.t in PIR
13:51 fperrad first ticket wth 4 digits
13:52 dalek parrot: r41208 | mikehh++ | trunk (2 files):
13:52 dalek parrot: codetest failure - run headerizer
13:52 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41208/
13:56 Coke flh++ # ticket #1k
13:57 Andy joined #parrot
13:59 Coke it's my tiny pet dancer!
14:00 jdv79 dukeleto: no - there is no installed parrot as far as i know.  i've never built parrot on this box.
14:01 jdv79 its fedora linux 9
14:02 krunen joined #parrot
14:02 particle why am i so happy that tt#1000 is a patch and not a bug?
14:02 NotFound A signal from the Parrot Gods?
14:03 NotFound (Hope is not a SIGKILL)
14:08 Coke particle: I was tempted to port over some bletcherous RT for #1000, but thought better of it. =-)
14:16 dalek parrot: r41209 | coke++ | trunk/tools/build/pmc2c.pl:
14:16 dalek parrot: Convert RT #39142 to a documentation bug and fix it.
14:16 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41209/
14:21 Coke NotFound: http://rt.perl.org/rt3/Tic​ket/Display.html?id=46687
14:23 NotFound Coke: "This problem will go away once packfiles are garbage collectable" Yeah, that's the reason I'm trying ugly and leaky workarounds
14:24 theory joined #parrot
14:29 NotFound BTW the PackFile structures uses lots of mem_sys allocations
14:31 dalek parrot: r41210 | NotFound++ | trunk/src (3 files):
14:31 dalek parrot: [core] several fixes and ugly and maybe leaky workarounds to avoid segfault during destruction, TT #995 and others
14:31 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41210/
14:43 payload joined #parrot
14:45 NotFound Coke: forgot to ping you. Commited in r41210
14:46 quek left #parrot
14:48 Psyche^ joined #parrot
14:51 mikehh NotFound: I am getting some codetest errors - Failed test 'no unused assert macros' in include/parrot/sub.h
14:52 mikehh NotFound: I ran headerizer and clleared some other errors - but I still get those - help
14:53 Coke mikehh: those can typically be fixed by adding an ASSERT_ARGS at the top of the function.
14:53 Coke (you can look at another function for the format)
14:53 Coke something like ASSERT_ARGS(function_name)
14:53 NotFound mikehh: not my fault, but I'll take a look
14:53 Coke no, let mike! =-)
14:53 NotFound Ok
14:53 mikehh NotFound: I know - just asking - but Coke replied
14:54 quek joined #parrot
14:55 mikehh ok got it - trying now
14:56 NotFound whiteknight: What's the point of TT #656. I think we must let the garbage collector do his work.
14:56 whiteknight let me look
14:58 whiteknight chromatic says there is a performance win by allowing the PMC to destroy itself early
14:58 whiteknight we should test it both ways
15:01 NotFound Tell me a way to ensure that no one has an active pointer to that pmc, and I'll buy the argument.
15:01 Coke NotFound: verified, the ^D segfault is gone.
15:02 Coke mikehh: please check that revision of parrot to see if your segfaults go away.
15:03 Coke NotFound: oooh. history.test is not immediately segfaulting anymore...
15:03 Coke completes...
15:05 mikehh Coke: will try it
15:06 Coke NotFound: https://trac.parrot.org/parrot/ticket/966 is still live.
15:06 dalek TT #961 closed by coke++: segfault in Parrot_oo_find_vtable_override
15:07 dalek TT #962 closed by coke++: segfault in Parrot_Object_assign_pmc
15:07 dalek parrot: r41211 | mikehh++ | trunk/src/sub.c:
15:07 dalek parrot: codetest failure - no unused assert macros
15:07 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41211/
15:08 Coke mikehh++ # woot.
15:09 Coke NotFound_++ #super woot
15:09 davidfetter joined #parrot
15:09 Coke NotFound: checking #965 now.
15:11 Coke (mathop.test is big, though.)
15:11 quek left #parrot
15:12 Coke We can leave 995 open; I've closed the ancillary tickets.
15:14 Coke (that is, the ones that no longer occur after your fix for #995)
15:14 Coke sadly, 965 still segfaults.
15:14 dalek parrot: r41212 | mikehh++ | trunk/src (2 files):
15:14 dalek parrot: codetest failure - tabs in leading whitespace
15:14 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41212/
15:15 Coke still, that's 3 segfaults gone out of 5 remaining. =-)
15:17 davidfetter joined #parrot
15:17 NotFound Coke: Congratulations, you've been able to segfault gdb
15:18 Coke it's not me, it's parrot! =-)
15:18 dalek parrot: r41213 | cotto++ | trunk (3 files):
15:18 dalek parrot: [profiling] start using Parrot_Sub_get_x_from_pc, clean some code and add notes on what doesn't dtrt yet
15:18 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41213/
15:27 payload1 joined #parrot
15:27 jdv79 is parrot segfaulting on build anywhere else?
15:29 jdv79 if getting on the box would help let me know
15:33 mokurai joined #parrot
15:34 Coke NotFound: yah, memory usage seems to be up. (expr.test dying sooner.)
15:34 Coke (but it was already dying before)
15:34 NotFound Coke: The TclString get bool problem is just that vtable function calling himself
15:34 Coke NotFound: yes, but it's not explicitly calling itself. :|
15:35 NotFound Coke: is autoboxed in the pir part
15:36 Coke NotFound: autoboxing to TclString only happens in Tcl and _Tcl HLL's, not parrot.
15:36 Coke (I think I already had this conversation with someone.)
15:36 NotFound Coke: you must have it with PGE ;)
15:36 Coke PGE is also in the parrot HLL.
15:36 purl okay, Coke.
15:37 Coke so, sure, the backtrace says it's calling itself (I assume that's where you're seeing this), but where in that PIR is causing that, or is it a bug?
15:38 mikehh All tests PASS (pre/post-config, smoke, nqp_test, fulltest) at r41212 -Ubuntu 9.04 i386 (g++)
15:38 Coke I bet it's because we're throwing an exception inside a vtable.
15:39 NotFound Coke: autoboxing seems to happen here: match = true_s(self_s, 'grammar'=>'TclExpr::Grammar')
15:40 NotFound tclstring.pir get_bool
15:40 Coke ok. true_s is defined in parrot HLL.
15:40 Coke so that should not be autoboxing.
15:40 Coke ... to a TclString
15:41 Coke (and match is a PGE::Match, supposedly.)
15:43 Coke let me see if I can find out what self_s is at point to see if I can get a smaller test case.
15:43 Coke (this was on the mathop.test failure, eys?)
15:43 NotFound Let me check a stupid idea...
15:44 dalek partcl: r727 | coke++ | wiki/SpecTestStatus.wiki:
15:44 dalek partcl: NotFound++ # this segfaults are gone.
15:44 dalek partcl: review: http://code.google.com/p/p​artcl/source/detail?r=727
15:48 NotFound Coke: Who may tests has t_tcl/mathop.test ?
15:48 pmichaud (.gitignore files)  --- fwiw, only one .gitignore file should be needed, not one-per-directory
15:48 mikehh Coke - no segfaults in make test (that's on i386 mind)
15:48 Coke NotFound: tcl doesn't count their tests, they name them.
15:49 Coke it was segfaulting after 24.2
15:49 Coke sorry, after 'mathop-24.2'
15:49 NotFound Is still in 8-... X-)
15:49 Coke I'm running with a local mod to tclstring's get_bool to see if I can find out what boolean value is making it go crazy.
15:49 Coke s/boolean/string/
15:50 mikehh partcl r727 builds on parrot r41212 - make test PASS - Ubuntu 9.04 i386 (g++)
15:50 Coke NotFound: what's your stupid idea?
15:51 NotFound Coke: Pass a 'String' PMC built with the string
15:51 Coke gah.
15:52 Coke (only because that's another pmc created.)
15:52 NotFound I'll call it stupid for some reason ;)
15:53 Coke last two lines before segfault:
15:53 Coke CHECKING BOOL: '1'
15:53 Coke CHECKING BOOL: '1'
15:53 Coke so I don't think it's recursing.
15:53 Coke (before that is the output about what test just failed.)
15:55 NotFound That's new: src/call/context.c:1534: failed assertion 'Parrot_pcc_get_regs_used(interp, ctx, REGNO_PMC) > idx'
15:55 NotFound With just 59 stack frames.
16:03 whiteknight NotFound++
16:07 dalek parrot: r41214 | dukeleto++ | trunk/t/pmc/resizablefloatarray.t:
16:07 dalek parrot: [TT #1000][t] Convert t/pmc/resizablefloatarray.t to PIR, flh++
16:07 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41214/
16:15 NotFound Coke: my atempt of workaround gives a failure in the onscurities of PCC, I'll give up.
16:16 theory joined #parrot
16:17 dukeleto holy .gitignore churn, batman!
16:18 mikehh Coke I am getting random segfaults in rakudo now (on exit - PASS tests, except 1))
16:20 darbelo joined #parrot
16:20 mikehh rakudo - FAILS t/spec/S12-introspection/walk.rakudo - test 9 - set_pmc() not implemented in class 'Failure'
16:21 dalek TT #1000 closed by dukeleto++: t/pmc/resizablefloatarray.t in PIR
16:27 dukeleto when is the commit deadline for 1.6 ?
16:27 NotFound mikehh: test or spectest?
16:35 whiteknight joined #parrot
16:36 whiteknight internet at work is shitty
16:50 einstein joined #parrot
16:51 darbelo Is it just me, or is op_append() a half assed, encapsulation breaking, imitation of Parrot_str_append()
16:55 whiteknight I don't think I've ever seen that function
16:56 pmichaud src/pmc_freeze.c
16:56 purl well, src/pmc_freeze.c is full of scary.
16:56 darbelo what pmichaud said. It's only used twice there.
16:57 NotFound purl++
16:59 darbelo Apparently, so. I've just replaced it with that and make test is runing fine so far.
17:01 darbelo I've come to expect explosions everytime I touch the freeze code. Looks like I got it right this time.
17:01 NotFound darbelo: How? One of the arguments isn't a String
17:01 rindolf joined #parrot
17:03 nopaste "darbelo" at 200.49.154.172 pasted "[PATCH] replace op_append(), by use of Parrot_str_append()" (101 lines) at http://nopaste.snit.ch/17923
17:04 darbelo NotFound: It converts it to a char* and tacks it at the end of a string. That sounds like string_appending to me :)
17:05 darbelo I made the conversion to string explicit, and used the correct API for the appending.
17:06 chromatic joined #parrot
17:07 darbelo Mind, you. That Parrot_str_new() call is not as optimal as poking into the string's guts, but it was aiming for cleaner, not faster.
17:09 NotFound darbelo: I'm not sure that assuming that the default charset and encoding are compatible with plain binary usage is clean, but that file is doing it anyway...
17:09 nopaste "darbelo" at 200.49.154.172 pasted "[PATCH] replace op_append(), by use of Parrot_str_append()" (84 lines) at http://nopaste.snit.ch/17924
17:10 darbelo There was an extra hunk at the end of the patch. Repasted.
17:11 darbelo NotFound: Hmm, is there a binary-clean encoding availble? I could just Parrot_str_new_init() and use that.
17:14 whiteknight_ joined #parrot
17:14 NotFound darbelo: there is binary charset whose prefered encoding is fixed_8
17:16 darbelo Yes, looking inside Parrot_str_new_init I see that Parrot_fixed_8_encoding_ptr is the only one that will do what freeze wants, I'll revise the patch.
17:16 dalek parrot: r41215 | allison++ | trunk/docs/pdds/pdd15_objects.pod:
17:16 dalek parrot: [pdd] Some cleanups while reviewing object PDD, particularly:
17:16 dalek parrot: - Changed the spec to match current design/implementation where
17:16 dalek parrot:   attributes and parents can only be added before the first object is
17:16 dalek parrot:   instantiated from the class.
17:16 dalek parrot: - Clarified the differentiation between class methods and instance
17:16 dalek parrot:   methods for 'can' and 'does'.
17:16 dalek parrot: - Removed the comment about Class and Role PMC's methods being supplied
17:16 dalek parrot:   by a role. Unnecessary complexity.
17:16 dalek parrot: - Removed optional parameters to 'add_method' for selecting vtable
17:16 dalek parrot:   override (that has a separate method now).
17:16 dalek parrot: - Added information on the 'add_vtable_override' method in the Class
17:17 dalek parrot:   PMC.
17:17 dalek parrot: - Removed the mention of special handling for MultiSubs in 'add_method'
17:17 dalek parrot:   for Class and Role, since there is none.
17:17 dalek parrot: - Modified the description of 'find_method' in Object and Class,
17:17 dalek parrot:   matching current behavior (respecting the difference between class
17:17 dalek parrot:   methods and instance methods).
17:17 dalek parrot: - Deleted (sketchy) information about object "container types". The base
17:17 dalek parrot:   set of classes don't (and won't) store or respect that information,
17:17 dalek parrot:   languages that need it will need more extensive support (which may
17:17 dalek parrot:   eventually be merged into the base set).
17:17 dalek parrot: - Deleted 'remove_role' from Class and Role for consistency (the spec
17:17 dalek parrot:   already said removing roles wasn't allowed).
17:17 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41215/
17:19 NotFound "It makes no sense to try and search for a substring in raw binary data." What?
17:21 darbelo NotFound: You don't want a sub-string you want a sub-'raw binary data' :)
17:22 NotFound darbelo: And a 'binary' string isn't supposed to be that?
17:23 NotFound If not, kill the binary encoding and provide a raw byte buffer that isn't a string.
17:23 NotFound s/encoding/charset
17:35 pmichaud note that the word "string" really means "sequence of something"
17:35 pmichaud as in "string of pearls" or "string of lights"
17:36 pmichaud a string that has "fixed_8" encoding should be perfectly valid for "raw binary data".
17:36 pmichaud Searching such a string makes perfect sense.
17:37 darbelo pmichaud: C programmers will want to put a NULL after those pearls :)
17:37 Tene and pascal programmers?
17:37 pmichaud darbelo: only if they're talking about "character string"
17:37 dalek rakudo: 84296e4 | (Solomon Foster)++ | src/setting/Num.pm:
17:37 dalek rakudo: Fix implementations of the "forward" trig functions.
17:37 dalek rakudo: Other than sin, cos, and tan, most of the forward trig functions were broken.  This patch fixes all the broken ones I have spotted thus far.  Corresponding massive overhaul to trig.t, which just appears to have been flat out wrong for many of the functions.
17:37 dalek rakudo: Note that many if not most of the inverse trig functions remain broken.
17:37 dalek rakudo: review: http://github.com/rakudo/rakudo/commit/8​4296e43d89488073be442486b66e38abf2164ef
17:38 darbelo typical C library functions assume that, if Parrot's string subsystem is built on top of that, the Parrot does at well.
17:39 NotFound pmichaud: provided that is in the 'binary' charset. Otherwise is not supposed to be raw at all.
17:39 darbelo It's the only excuse I can think of for not allowing substring searches.
17:39 pmichaud I'm simply saying that searching for a subsequence in binary data makes sense.
17:40 NotFound darbelo: typical C functions are inheritances from the '70. Supposedly we are trying to build a modern system.
17:42 NotFound pmichaud: agreed
17:42 pmichaud afk, lunch & errands
17:43 nopaste "darbelo" at 200.49.154.172 pasted "[PATCH] kill op_append, replaced by Parrot_str_append(). All tests successful." (85 lines) at http://nopaste.snit.ch/17925
17:44 darbelo NotFound: does http://nopaste.snit.ch/17925 look sane to you?
17:44 chromatic +1, darbelo
17:45 NotFound darbelo: for consistency, I'll better also change the string creation in create_image
17:46 NotFound BTW, that enum_stringrep_one isn't yet deprecated?
17:47 darbelo NotFound: That one is _noinit, not _init
17:48 pmichaud chromatic: I tried the recycling of ResizableIntegerArray's in PGE, but didn't see _any_ performance improvement in PGE's spectests.
17:48 pmichaud sorry, in Rakudo's spectests
17:48 pmichaud (didn't see an improvement in PGE's spectests either... but)
17:49 darbelo But I can probably convert it too. Let me run some tests.
17:50 chromatic pmichaud, I didn't either... but let me do more profiling.
17:50 pmichaud okay
17:50 pmichaud I have a slightly different patch to offer... just a sec
17:51 NotFound Parrot_str_new_init: "Don't call this directly." What????
17:52 nopaste "pmichaud" at 72.181.176.220 pasted "updated recycle pool for PGE (for chromatic)" (91 lines) at http://nopaste.snit.ch/17926
17:52 pmichaud the #17926 version will also recycle cstack upon success when it's known that we don't backtrack into the regex
17:52 pmichaud (it also adds a counter to see how many times we get RIAs from the pool instead of creating a new one)
17:53 Coke it makes me cry that we'd have to track that ourselves. :|
17:53 pmichaud however, masak++ asked a question earlier that points out to me a place where PGE is suboptimal for token/rule, so I'm going to see about fixing that.  It should reduce a ton of GC pressure by avoiding the RIAs in the first place
17:53 iblechbot joined #parrot
17:53 pmichaud but I need to do lunch and shopping first
17:53 pmichaud bbiaw
17:55 chromatic That's a better place to start.
17:57 Coke #17926 actually slows while.test down from 45.1s to 47.6s
17:57 Coke (just one run each.)
17:58 Coke avoiding Integers for ints is probably very helpful; what about Strings for strings?
17:58 rindolf left #parrot
17:59 chromatic That can help too, Coke.
17:59 Coke I'm thinking avoiding CodeString and just doing a string concat would be a win.
18:02 chromatic CodeString can be expensive.
18:04 whiteknight What exactly does CodeString do?
18:04 whiteknight I've never looked at it closely enough to figure that out
18:05 chromatic It helps keep track of HLL annotations.
18:06 whiteknight oh. That doesn't sound like much
18:06 kyle joined #parrot
18:09 Coke also gives you something like printf without having to wrap your args in an array.
18:09 Coke which makes the embedded PIR code /much/ more reasonable.
18:09 Coke (readable)
18:13 Coke 45.1/44.8
18:13 purl 1.00669642857143
18:16 dalek parrot: r41216 | darbelo++ | trunk/src/pmc_freeze.c:
18:16 dalek parrot: Minor src/pmc_freeze.c cleanup: replace the op_append() function with calls to Parrot_str_append()
18:16 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41216/
18:21 ttbot Parrot trunk/ r41216 MSWin32-x86-multi-thread make error http://tt.ro.vutbr.cz/file/cmdout/92322.txt ( http://tt.ro.vutbr.cz//buil​dstatus/pr-Parrot/rp-trunk/ )
18:25 Coke darbelo: you broke the build. you are SO in trouble.
18:26 NotFound Off with his head!
18:27 Coke src/pmc_freeze.c:336: warning: ISO C90 forbids mixed declarations and code
18:28 * Coke wonders what compiler darbelo is using that he didn't get that error.
18:28 chromatic Anything built in this millennium, presumably.
18:28 chromatic Except Visual Studio.
18:28 Coke chromatic: I get that error on gcc.
18:29 Coke because we turned on that waring.
18:29 chromatic I get the warning too, but it's not an error.
18:29 NotFound That has happened to me some times, while building with C++ and forget to try with C before commit.
18:29 Coke bah. misread my output, you are correct sir.
18:29 Coke I blame gcc for lying and telling us we could make that an error.
18:29 chromatic The *best kind* of correct: technically correct.
18:32 Coke darbelo: anyway, move that PARROT_ASSERT down after the declarations. (or split up the declarations from the setting if you need it to happen sooner.)
18:34 NotFound Mmmm... we are already requiring that condition?
18:36 Coke TimToady, that's spelled cdr.
18:40 TimToady no, it's not; see the Steel slides
18:40 TimToady he's slanging Lisp up and down for that
18:42 TimToady ref is  http://research.sun.com/projects/plrg/​Publications/ICFPAugust2009Steele.pdf
18:43 TimToady and I do think he's latched onto some fundamental ideas on parallelizability
18:45 fperrad_ joined #parrot
18:45 einstein can it be that the trunk does not compile at the moment?
18:45 kjeldahl joined #parrot
18:45 whiteknight fun slides! Makes me wish I paid more attention to Lisp in school
18:46 whiteknight einstein: I think it might be broken right now, yes
18:46 einstein hmm ok
18:47 dalek parrot: r41217 | darbelo++ | trunk/src/pmc_freeze.c:
18:47 dalek parrot: Move a PARROT_ASSERT to prevent a "mixed declarations and code" warning.
18:47 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41217/
18:47 whiteknight I once wrote my own Perl5-based expert system shell from the ground up to avoid the pain of learning a lisp-like expert system language
18:47 Coke especially if you're on windows.
18:47 Coke does that help?
18:47 Coke whiteknight: ah, exactly the kind of misplaced laziness that parrot has come to rely on.
18:47 Coke ;)
18:50 darbelo Coke: Your build unbroken now?
18:52 Coke darbelo: my build was not; I was pointing at the buildbot output.
18:53 Coke einstein: can you try again?
18:53 Coke (since you reported a problem pre 41217?)
18:54 dalek parrot: r41218 | chromatic++ | trunk/src (2 files):
18:54 dalek parrot: [profiling] Made Parrot_Sub_get_line_from_pc() more correct for the profiling
18:54 dalek parrot: runcore.
18:54 purl i think runcore is the part that actually executes the byte code
18:54 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41218/
18:54 Coke TimToady: I imagine that would make a lot more sense to me if I were listening to the lecture or high on caffiene or back in college. :|
18:56 joeri joined #parrot
18:57 Coke ah. page 69 onward is in english. =-)
19:01 Coke chromatic, cotto: does your version of cg_annotate work with the generated files?
19:01 chromatic yes.
19:02 szabgab joined #parrot
19:02 einstein ok its now building on my machine
19:02 Coke (can someone with sudo privs on feather update the callgrind tools?)
19:05 darbelo Wait, MSVC, a c++ compiler, failed to build parrot because *ISO C 90* forbids something. WTF?
19:06 whiteknight darbelo: you're running it in C mode
19:06 whiteknight which means it travels back in time 20 years and acts stupid for the time
19:07 Coke hey, valgrind 3.5.0 claims to work on OSX 10.5
19:07 Coke I have 10.4 :|
19:07 chromatic It's only been a decade since C99; won't you give Microsoft a break?  They're still trying to figure out the PNG and CSS specifications.
19:10 * Coke just builds his own vg on feather.
19:12 Coke chromatic: I just updated cg_annotate to 3.5.0 - it still fails to deal with the .out file.
19:13 chromatic I don't know what to tell you, Coke.
19:14 darbelo Is there a way to make MSVC notice that we are being 100% compatible with c++ and not behave like we're still in the eighties?
19:15 nopaste "coke" at 72.228.52.192 pasted "Do my command lines look right, at least?" (10 lines) at http://nopaste.snit.ch/17927
19:15 chromatic That looks fine to me, Coke.
19:16 Coke k. thanks.
19:18 pmichaud 18:04 <whiteknight> What exactly does CodeString do?
19:18 pmichaud 18:05 <chromatic> It helps keep track of HLL annotations.
19:18 pmichaud False.
19:18 pmichaud I mean, that's not really the purpose of CodeString.
19:18 chromatic It's not the only purpose, I agree.
19:18 pmichaud It's not even the primary one.  The primary method in CodeString is 'emit'
19:19 pmichaud We used to have that in PIR, but switched it to its own PMC so it would be faster.
19:19 pmichaud There's a significan't speedup by having it in C versus PIR.
19:19 pmichaud *significant
19:20 pmichaud but ultimately a CodeString concat should be exactly the same as a String concat
19:20 pmichaud (since CodeString is just a subclass of String)
19:25 whoppix joined #parrot
19:30 theory joined #parrot
19:31 pmichaud .oO( Why are my comments on CodeString always met with silence...? )
19:32 chromatic What else is there to say?  You wrote it and you know what it does better than anyone.
19:32 darbelo pmichaud: If you understand them enough to explain that much, it means we don't have to :)
19:32 pmichaud I dunno.  I keep having to say it, though.  (See TT #448.)
19:32 pmichaud also various threads in parrot-dev
19:33 pmichaud I'm just curious, not miffed.
19:34 chromatic Yeah, I agree with all of your comments in that ticket.
19:36 pmichaud anyway, if we could convert CodeString into an efficient StringBuffer of some sort, that would likely be a Win.
19:36 pmichaud gotta pick up kids... bbiab
20:01 Coke I think "emit" in opcode form is "sprintf"
20:04 Coke (not quite the same, but similar.)
20:09 pmichaud I tried using sprintf for what emit does.  It was.... horrible.
20:13 dalek rakudo: 3b63817 | pmichaud++ | docs/spectest-progress.csv:
20:13 dalek rakudo: spectest-progress.csv update: 436 files, 14267 (69.2% of 20603) pass, 5 fail
20:13 dalek rakudo: Failure summary:
20:13 dalek rakudo: S32-str/samecase.rakudo aborted 5 test(s)
20:13 dalek rakudo: review: http://github.com/rakudo/rakudo/commit/3​b6381761a44cb1865c8063a9e57bd0fe0ba4171
20:30 particle joined #parrot
20:33 Coke list of BoD should be updated on wikipedia.
20:41 payload joined #parrot
20:53 ash_ joined #parrot
21:01 dalek parrot: r41219 | allison++ | trunk/docs/pdds/pdd15_objects.pod:
21:01 dalek parrot: [pdd] Stray comma in code example, caught by Pod tests.
21:01 fperrad_ joined #parrot
21:01 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41219/
21:09 Tene chromatic: I have a segfault reproducable in 10 lines of pir
21:13 chromatic Is it a known segfault?
21:14 nopaste "tene" at 97.117.70.208 pasted "loadlib/thread segfault for chromatic++" (10 lines) at http://nopaste.snit.ch/17930
21:16 NotFound Tene: doesn't segfault for me
21:16 Tene NotFound: :(
21:16 NotFound linux amd64
21:16 * Tene also linux amd64
21:17 ash_ Tene: that might be related to your OS or something, that doesn't segfualt for me...  Parrot build 41181
21:17 NotFound r41218
21:17 * Tene 41219
21:17 NotFound Only difference is a .pod
21:17 dalek parrot: r41220 | darbelo++ | trunk/docs/pdds/pdd15_objects.pod:
21:17 dalek parrot: Reformat a paragraph to avoid lines longer that 78 characters and please t/codingstd/pdd_format.t
21:17 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41220/
21:18 NotFound Tene: --optimize ?
21:18 Tene NotFound: no.
21:18 chromatic No segfault here.
21:18 Tene ><
21:19 NotFound I think I know the difference: no libsqlite here
21:19 Tene NotFound: replace it with any other .so
21:19 chromatic True, no libsqlite here either.
21:20 Tene libz
21:20 ash_ I have libsqlite3 and it works for me, i'll try a different lib
21:20 Tene libz should be available pretty widely.
21:20 NotFound Bingo
21:21 ash_ Did yours crash NotFound?
21:21 NotFound With libz, yes
21:22 chromatic Same here.
21:22 chromatic The backtrace says things go wrong starting in src/dynext.c in run_init_lib.
21:23 ash_ hmmm mine isn't crashing with both libz and libsqlite3 in mine, i am on OS X 10.6.1 though
21:23 Tene Over the past few days, I've been able to get a lot of errors from clone_interpreter, but they've all been exceptions, NPMCA, get_string not implemented, etc.  This is the first segfault.
21:23 ash_ i'll try updating parrot and see if it still happens
21:23 NotFound I think the problem may be related to library handles
21:24 chromatic #7  0xb7e1a7c4 in Parrot_dlsym (handle=0x80d96e0, symbol=0x81c1478 "Parrot_lib_libz_load")
21:24 Tene This came out of trying to evaluate the patches on tt757, but I don't really understand what's going on in there.
21:24 chromatic Do we have a libz with that symbol?
21:25 Tene chromatic: with sqlite, it says: #6  0x00007ffff7b5d493 in Parrot_dlsym (handle=0x6ab940,  symbol=0x7c87e0 "Parrot_lib_libsqlite3_load")
21:25 chromatic My guess is no such symbol.
21:26 ash_ just update to r41220 and it still doesn't segfault
21:28 chromatic Random patch.
21:28 nopaste "chromatic" at 72.87.39.97 pasted "Silly workaround for Tene's segfault" (16 lines) at http://nopaste.snit.ch/17931
21:28 chromatic This isn't the most correct solution, but it gets around the problem for me.
21:29 Whiteknight joined #parrot
21:31 chromatic Whatever we do, we can't assume that we can dlopen random shared libraries, rummage around for a Parrot_* prefixed symbol, and call that as a function.
21:31 chromatic That's definitely a recipe for crashes.
21:33 nopaste "tene" at 97.117.70.208 pasted "chromatic: think I could get you to look at this case too?" (14 lines) at http://nopaste.snit.ch/17932
21:35 chromatic No crash here.
21:35 Tene ... oh. :(
21:36 chromatic No difference with or without that patch.
21:36 Tene chromatic: what if you change it to: load_language 'perl6' ?
21:37 chromatic Change what and where?
21:37 purl chromatic: that doesn't look right
21:37 nopaste "tene" at 97.117.70.208 pasted "Like this" (14 lines) at http://nopaste.snit.ch/17933
21:37 NotFound I think there is a problem with Parrot_clone_lib_into, doesn't look like getting the handle from the right place.
21:38 fperrad left #parrot
21:38 chromatic I don't have perl6 installed.
21:39 Tene you could load_bytecode 'perl6.pbc' from rakudo's build dir.
21:39 chromatic Will do.
21:41 Tene There are some patches on tt757, but I don't understand what's going on well enough to understand them.
21:46 NotFound That was.
21:46 NotFound Some objection to add vtable get_pointer to ParrotLibrary ?
21:46 chromatic Makes sense.
21:46 NotFound With that, the fix is easy.
21:54 mikehh All tests PASS (pre/post-config, smoke, nqp_test, fulltest) at r41220 -Ubuntu 9.04 i386 (g++)
21:55 dalek parrot: r41221 | NotFound++ | trunk/src (2 files):
21:55 dalek parrot: [core] add get_pointer vtable to ParrotLibrary and use it in Parrot_clone_lib_into, solving a problem reported by Tene++ on irc
21:55 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41221/
22:00 rdice joined #parrot
22:02 mikehh partcl r727 builds on parrot r41220 - make test PASS - Ubuntu 9.04 i386 (g++)
22:12 theory joined #parrot
22:20 kid51 joined #parrot
22:25 davidfetter joined #parrot
22:27 davidfetter joined #parrot
22:28 mikehh rakudo (3b63817) builds on parrot r41220 - make test PASS / make spectest (up to 28220) FAIL - Ubuntu 9,04 i386 (g++)
22:28 jdv79 anyone one this build segfault?
22:28 jdv79 i don't know enough to triage it.
22:29 darbelo jdv79: Build segfault? Where?
22:29 jdv79 something to do with PGE.pbc
22:29 mikehh rakudo - 1 direct fail -  t/spec/S12-introspection/walk.rakudo - Non-zero exit status: 1 - Parse errors: Bad plan.  You planned 10 tests but ran 8.
22:30 darbelo jdv79: Can you point me to a ticket or nopaste describing it?
22:30 jdv79 i guess i'll push a ticket
22:30 Whiteknight doesn't fail when I build
22:30 Whiteknight jdv79: what system?
22:30 purl system is down is at http://www.homestarrunner.com/sbemail45.html
22:30 jdv79 it doesn't fail with the same OS on another box of mine either
22:30 mikehh rakudo - ./perl6 t/spec/S12-introspection/walk.rakudo - fails with (after ok 8):
22:30 mikehh rakudo - set_pmc() not implemented in class 'Failure'
22:30 mikehh rakudo - in regex _block424 (t/spec/S12-introspection/walk.rakudo:92)
22:30 mikehh rakudo - called from method Perl6Object::WALK (src/gen_setting.pm:1871)
22:30 mikehh rakudo - called from Main (t/spec/S12-introspection/walk.rakudo:94)
22:30 mikehh rakudo - t/spec/S03-operators/arith.rakudo - TODO passed:   131
22:30 mikehh rakudo - I got 112 tests PASSing but exiting with Non-zero wait status: 11
22:31 jdv79 http://nopaste.com/p/ahhiotQTb
22:32 chromatic The Rakudo problem looks like accessing the wrong constants.
22:32 Whiteknight when I check out rakudo, do I have to do it in /languages/rakudo from parrot trunk?
22:32 chromatic src/call/pcc.c:609 or so
22:33 Whiteknight urg, that whole pcc.c file needs to burn and die
22:34 chromatic Yep.
22:34 Whiteknight I keep hoping our lady of blessed PCC refactoring will hear our prayers and smite it
22:35 darbelo jdv79: Are you by any chance on Fedora or Red Hat?
22:35 jdv79 yes, fedora 9
22:36 mikehh I've got rakudo checked out in $HOME/rakudo and parrot in $HOME/parrot
22:36 jdv79 is there a known issue?
22:36 sri joined #parrot
22:36 Whiteknight holy crap, I just coretested in 36s
22:36 Whiteknight personal best!
22:36 chromatic My best is 28.
22:36 Whiteknight chromatic: show-off
22:37 jdv79 rakudo spectest is the worst
22:37 mikehh after building parrot I run sudo make install-dev and in rakudo perl Configure.pl --parrot-config=/usr/local/bin/parrot_config
22:37 Whiteknight either way, my time has been steadily trending downward as we've made fixes following context_pmc3
22:37 darbelo jdv79: Then it's probably SELinux interference here, I think setting that to 'premissive' was the standard workarround.
22:38 jdv79 oh wow.  i didn't notice that was on.  thanks.
22:38 darbelo auto::jit -           Determine JIT capability...p = 0x8940000  PAGE_SIZE = 4096 (0x1000)
22:38 darbelo failure: Permission denied
22:38 purl Permission denied is sort of self-explanatory
22:38 darbelo .........................yes.
22:39 darbelo The configure probe fails with permission denied, but still repports you platform as JIT-capable.
22:39 darbelo Segfaulting ensues.
22:40 Whiteknight mikehh: How do you have the different directories? my rakudo Configure.pl can't find parrot_config
22:41 mikehh what you guys usin' - my best is about 50 seconds
22:41 darbelo Whiteknight: you need --parrot-config=/path/to/parrot_config OR --gen-parrot
22:41 chromatic TEST_JOBS=5 make -j5 coretest
22:42 mikehh Whiteknight: I insatll parrot - sudo make install-dev (default) and in rakudo - perl Configure.pl --parrot-config=/usr/local/bin/parrot_config
22:42 rg1 joined #parrot
22:42 jdv79 hmm, i didn't notice that
22:42 jdv79 is there a way to warn about sellinux?
22:42 jdv79 that'd be swell
22:43 jdv79 *selinux
22:43 Whiteknight mikehh, darbelo: Okay, I think I have it
22:44 darbelo jdv79: I have no idea, I'm on BSD. I just remembered the SELinux issue when I saw that on the nopaste. If it makes you feel better the current JIT is on it's way out :)
22:46 jdv79 i'll know in a minute if i feel better when it builds
22:48 mikehh still takes me about 60 seconds
22:48 darbelo mikehh: You probably need more cores then :)
22:49 darbelo The rule of thumb for make -j is number of cores + 1
22:50 mikehh ok down to 46 if I use 9
22:51 kid51 jdv:  Try:  perl Configure.pl --verbose-step=auto::jit
22:52 MoC joined #parrot
22:53 mikehh I don't think the -j has any effect running under t/harness
22:53 Coke mikehh: that's TEST_JOBS
22:53 mikehh just the TEST_CORE=n
22:53 mikehh sorry - just the TEST_JOBS=n
22:54 Coke you could do "make TEST_JOBS=3 -j3 test" to build and test in || in one step.
22:57 mikehh ha - make coretest TEST_JOBS=20 - Files=226, Tests=6423, 28 wallclock secs ( 2.00 usr  0.49 sys + 34.86 cusr 16.48 csys = 53.83 CPU)
22:58 jdv79 darbelo: that was it - thanks
23:01 Whiteknight mikehh: how many cores does yourmachine have?
23:02 Coke hey, cg_annotate is a perl script.
23:02 mikehh Whiteknight - 4 AMD Phenom II 940 3.0 GHz with 8GB ram
23:02 Whiteknight oh yeah, so you're rocking that
23:03 mikehh built it myself a couple of months ago
23:03 Zak joined #parrot
23:04 mikehh the motherboard allows for 16GB but that is far too expensive
23:05 mikehh so I got 4x 2GB DDR2 1033
23:06 Coke chromatic, cotto: can one of you post a .out. file that works for you?
23:12 Coke aha. it's not us. this version of callgrind can't deal with this verison of valgrind, either. :|
23:12 Coke (installed version of perl6 on feather is borked)
23:15 Coke (val grinned? who knew.)
23:16 jdv79 treed: http://nopaste.com/p/aEi9Ims9nb
23:21 mikehh managed to get it down to 22 seconds - make coretest TEST_JOBS=40 - Files=226, Tests=6423, 22 wallclock secs ( 1.66 usr  0.41 sys + 28.06 cusr 12.50 csys = 42.63 CPU)
23:21 treed Huh?
23:22 treed jdv79: How old is that?
23:22 treed I don't run ?0 at all.
23:26 jdv79 what do you mean how old
23:26 jdv79 ?
23:26 jdv79 how would i know?
23:27 jdv79 rake, version 0.8.99.3
23:28 treed Recent revision of cardinal?
23:28 jdv79 latest
23:28 treed k
23:28 treed I won't have much time to look into this kind of stuff for a couple days.
23:28 Coke chromatic: AIIIIIIIIIIIIIIIIIIIIIIIIIIIGH
23:29 treed I'm starting a new job on Monday.
23:29 treed And the new job necessitates a move.
23:29 Coke msg chromatic it's callgrind_annotate. not cg_annotate.
23:29 purl Message for chromatic stored.
23:29 Coke works fine.
23:29 Coke (*#&$#
23:29 chromatic Good to know.
23:30 jdv79 might be file perms, treed  - retrying
23:31 treed k
23:35 jdv79 nope
23:36 mikehh decnum_dynpmcs r181 builds on parrot r41220 - make test PASS - Ubuntu 9.04 i386 (g++)
23:41 jdv79 treed: i guess i'll wait til you're back
23:41 jdv79 thanks
23:46 Coke http://feather.perl6.nl/~coke/power.profile - from  tcl.pbc -q examples/power.tcl
23:48 s1n joined #parrot
23:50 chromatic Lots of parsing.
23:51 Coke ayup.
23:52 Coke that's http://code.google.com/p/partcl/sou​rce/browse/trunk/examples/power.tcl
23:56 Coke pmichaud: hey, did you have a change to poke at tcl?
23:58 jdv79 Coke: my first 100% partcl smolder run. nice.

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

Parrot | source cross referenced