Camelia, the Perl 6 bug

IRC log for #parrot, 2009-02-21

Parrot | source cross referenced

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

All times shown according to UTC.

Time Nick Message
00:03 Andy joined #parrot
00:03 rurban typedef INTVAL (*nci_jit_sub_t)(PARROT_INTERP, PMC *, char *); is not needed?
00:04 rurban oops, i see
00:07 rurban - PMC_struct_val(ret) = PMC_struct_val(SELF); smells fishy to me.
00:07 Limbic_Region joined #parrot
00:08 rurban deleting SET_ATTR_multi_sig(INTERP, ret, PMCNULL); also
00:08 cotto the PMC_struct_val assignment is taken care of by assigning nci_info_ret->orig_func
00:09 AndyA joined #parrot
00:09 cotto the SET_ATTR_multi_sig is unnecessary because PMC_data is zeroed when allocated
00:11 rg rurban: where would you want to use the union trick for alignment?
00:13 rurban I used explicit alignment on unaligned malloced ptrs
00:14 rurban but the problem comes from freeze, where the internal strcutures suddenly are unaligned.
00:14 rurban so it's unpack to blame
00:15 rurban Most likely a string
00:16 rurban I'd really need access to such a machine. I turned up my internal PARROT_PTR_ALIGNMENT to 8, to see if it makes a difference. of course not.
00:19 rurban solaris has dtrace so theoretically we could write a dtrace function to check the addr during pmc freeze.
00:19 rurban => check the addr alignment
00:20 rg i wonder if we really need to fix this. after all, gcc doesn't give us trouble and we have a compiler option for the sun compiler that makes things work.
00:20 Theory_ joined #parrot
00:20 rurban its the entry before the integer for sure. so stepping the backtrace up should help a bit.
00:21 rurban sure. we will commit a fix to makes thinks work for 1.0
00:21 rurban But I would prefer properly aligned addresses for speed on 64 bit. it's there for a reason.
00:22 rurban fix: ccflags hint for sunpro cc on 64bit
00:29 purl joined #parrot
00:29 rurban PackFile_ConstTable_unpack fails on 64bit big-endian. otherwise fine.
00:30 rurban good, my first bug is fixed!
00:31 rurban endianess is not the issue, just 64bit. good
00:35 gryphon joined #parrot
00:35 bacek joined #parrot
00:35 TimToady joined #parrot
00:36 rurban rg: you could compile src/gc/memory.c with -DDETAIL_MEMORY_DEBUG if you want to, but it's very verbose.
00:37 rurban and its most likely not the cause. my numbers all look aligned
00:38 rg well it's not getting very far before it hits the misaligned access
00:39 rurban I suppose it fails after the very first string, which can be the name of a pmc method.
00:42 rurban okay, my big 32-64 but bug is fixed at least. I can properly ready your and the other 64 bit pbcs on 32 bit.
00:42 zostay joined #parrot
00:42 rurban enough for today!
00:43 rg thanks for all your help and hard work. rurban++
00:43 rurban rg++
00:48 zostay joined #parrot
01:01 Tene joined #parrot
01:05 dalek parrot: r36914 | coke++ | trunk:
01:05 dalek parrot: TT # 188 - Remove deprecatd Closure PMC
01:05 dalek parrot: review: https://trac.parrot.org/parrot/changeset/36914/
01:12 TiMBuS joined #parrot
01:23 Whiteknight coke++
01:27 PacoLinux joined #parrot
01:30 kj joined #parrot
01:32 PacoLinux joined #parrot
01:37 dalek parrot: r36915 | whiteknight++ | trunk/src/pmc/class.pmc:
01:37 dalek parrot: [Core] Fix TT #321 by replacing an instance of Parrot_run_meth_fromc_args with Parrot_pcc_invoke_method_from_c_args, which is actually MORE stable in this case
01:37 dalek parrot: review: https://trac.parrot.org/parrot/changeset/36915/
01:39 samlh joined #parrot
01:51 Whiteknight it's funny to me because Parrot_pcc_invoke_method_from_c_args seems to break everything else. But in this case it's actually the solution
01:53 leto_ joined #parrot
01:56 dalek parrot: r36916 | whiteknight++ | branches/rename_pccinvoke:
01:56 dalek parrot: [rename_pccinvoke] update src/pmc/class.pmc to use the new function, revert changes the src/io/api.c that I'm not ready to update yet
01:56 dalek parrot: review: https://trac.parrot.org/parrot/changeset/36916/
01:57 Andy joined #parrot
02:15 samlh joined #parrot
02:27 Andy joined #parrot
02:28 dalek parrot: r36917 | cotto++ | trunk/src/pmc/lexpad.pmc:
02:28 dalek parrot: [PMC] use ATTRs in LexPad PMC
02:28 dalek parrot: review: https://trac.parrot.org/parrot/changeset/36917/
02:29 ron joined #parrot
02:39 ron left #parrot
03:00 GeJ Good morning everyone.
03:08 leto_ GeJ: mornin'
03:29 jimmy joined #parrot
03:30 jimmy morning
03:42 janus joined #parrot
03:45 gerd joined #parrot
04:11 jimmy joined #parrot
04:17 Infinoid hi, morning people
04:20 rurban_ joined #parrot
05:18 samlh joined #parrot
05:21 klapperl joined #parrot
05:36 dalek parrot: r36918 | chromatic++ | trunk:
05:36 dalek parrot: [PMC] Made nested UnManagedStructs independent of each other, per RT #31292.
05:36 dalek parrot: This unTODOes one test.
05:36 dalek parrot: review: https://trac.parrot.org/parrot/changeset/36918/
06:12 mberends joined #parrot
06:27 Theory joined #parrot
06:31 masak joined #parrot
06:53 contingencyplan joined #parrot
07:04 jimmy joined #parrot
08:37 dalek parrot: r36919 | cotto++ | trunk:
08:37 dalek parrot: [PMC] make NCI PMCs use ATTRs, with one temporary workaround for i386 jit
08:37 dalek parrot: review: https://trac.parrot.org/parrot/changeset/36919/
08:41 iblechbot joined #parrot
08:49 dalek parrot: r36920 | cotto++ | trunk/src/pmc/nci.pmc:
08:49 dalek parrot: [PMC] add TT to previous commit
08:49 dalek parrot: review: https://trac.parrot.org/parrot/changeset/36920/
08:56 dalek tracwiki: v21 | cotto++ | PMCUnionDeprecationTasklist
08:56 dalek tracwiki: https://trac.parrot.org/parrot/wiki/PMCUnionD​eprecationTasklist?version=21&action=diff
08:56 shorten dalek's url is at http://xrl.us/begz63
09:00 bacek joined #parrot
09:16 barney joined #parrot
09:26 iblechbot joined #parrot
09:43 iblechbot joined #parrot
09:55 bacek bacek@icering:~/src/parrot$ git bisect bad
09:55 bacek Bisecting: 1 revisions left to test after this
09:55 bacek [9c6457d3393113c48f47707f596ac35167b2e1cc] [PMC] convert String PMC to use ATTRs
09:56 bacek This commit broke Rakudo...
10:27 TiMBuS bacek, what broke?
10:32 masak TiMBuS: try and build Rakudo. I get a segfault over here.
10:33 TiMBuS oh i see. i was confused there because bacek used git so i thought it was a rakudo commit
10:34 masak probably git-svn.
10:34 TiMBuS there we go, segfault on an updated parrot
10:34 masak I probably don't need to say this, but _please_ someone fix the segfault.
10:34 masak :/
10:36 TiMBuS is it possible for actions.pm to access the PCT namespace?
10:37 masak I don't know, but it sounds likely.
10:37 masak if I were to check, what would I be looking for?
10:38 rurban msg cotto your string.pmc update broke rakudo
10:38 purl Message for cotto stored.
10:38 TiMBuS i'm just wondering if PCT could set some variables for actions.pm to grab, like filename, and the context the code is being run in (interactive mode, oneliner eval, etc)
10:39 rurban But I'm pretty sure rakudo has to use the new string.pmc
10:50 dalek parrot: r36921 | fperrad++ | trunk/t/pmc/os.t:
10:50 dalek parrot: [t] skip new test on Windows
10:50 dalek parrot: review: https://trac.parrot.org/parrot/changeset/36921/
11:26 jimmy joined #parrot
12:00 mikehh I built rakudo but over half the coretests fail
12:01 mikehh most of the spectests skip or fail
12:06 mikehh parrot seemed to build ok - r36921 - http://smolder.plusthree.com/app/pu​blic_projects/report_details/18224
12:06 shorten mikehh's url is at http://xrl.us/beg2cp
12:12 jimmy joined #parrot
12:17 dalek parrot: r36922 | NotFound++ | trunk/src/pmc/lexpad.pmc:
12:17 dalek parrot: [cage] fix c++ build in lexpad.pmc
12:17 dalek parrot: review: https://trac.parrot.org/parrot/changeset/36922/
12:19 Themeruta joined #parrot
12:20 NotFundgf hi
12:21 rurban_ joined #parrot
12:25 jimmy hello
12:25 purl hola, jimmy.
12:31 baest joined #parrot
12:36 dalek parrot: r36923 | rurban++ | trunk/tools/dev/pbc_header.pl:
12:36 dalek parrot: [tools] fix wrong optimization logic from r36890.
12:36 dalek parrot: review: https://trac.parrot.org/parrot/changeset/36923/
12:41 TiMBuS so whats it mean when the value of s->strstart is 0xdeadbeef :/
12:41 TiMBuS because thats why rakudo is crashing
12:41 TiMBuS s being a STRING
12:47 jimmy dead beef :(
12:56 jan joined #parrot
13:38 dalek parrot: r36924 | rurban++ | trunk/tools/dev/ops_not_tested.pl:
13:38 dalek parrot: [tools] skip t/native_pbc from the pbc_disassemble steps
13:38 dalek parrot: review: https://trac.parrot.org/parrot/changeset/36924/
14:01 muixirt joined #parrot
14:02 muixirt hi
14:02 purl hola, muixirt.
14:02 jan joined #parrot
14:07 muixirt is it save to assume that in function signatures for NCI that size_t equals int?
14:23 NotFundgf muixirt: no
14:24 muixirt NotFound, so what can I use instead?
14:27 muixirt NotFundgf, of course
14:28 rurban I would say long on 64bit, int on 32bit
14:30 NotFundgf We must add an option for it. In the meantime I think long may be the most appropiate
14:51 muixirt ok, thanks
15:04 TiMBuS rurban, long is 32 bits on 32bit platforms
15:05 TiMBuS and 64 bit on most 64bit platforms, except windows
15:05 rurban and that's what size_t is normally, isn't it?
15:07 TiMBuS i think its the maximum supported int size yeah
15:13 rurban another string padding bug found with 64bit
15:14 rurban only one 32/64bit bug missing...
15:22 Whiteknight joined #parrot
15:25 rurban Regarding reading older or newer pbc's: what our policy there? I think parrot should be strict (as currently), and the utils non-strict. So the utils will only warn, parrot will fail.
15:40 rdice joined #parrot
16:11 cotto rurban, do you mean that there's a bug in the code or that rakudo needs to be updated?
16:12 cotto (string.pmc)
16:12 rurban parrot is fine obviously, but rakudo crashes , bisected to your string.pmc update
16:13 cotto I have to leave soon, but I'll look at it later today.
16:13 rurban See the irclog, bacek and TiMBus reported
16:16 TiMBuS hm? oh yeah, rakudo is crashing inside perl6str's get_int, for some reason the string isnt init'd so s->strstart is 0xdeadbeef
16:17 TiMBuS correction, it's in get_number
16:17 Limbic_Region joined #parrot
16:28 masak joined #parrot
16:36 dalek lua: cdc0dc3 | (Francois Perrad)++ |  (4 files):
16:36 dalek lua: luap is an installable
16:36 dalek lua: review: http://github.com/fperrad/lua/commit/cd​c0dc3bd61b76e637d567de6a34767e06102a9d
16:36 shorten dalek's url is at http://xrl.us/beg2q8
16:57 Theory joined #parrot
17:53 bacek joined #parrot
18:00 geof joined #parrot
18:04 Whiteknight any PCT experts online?
18:06 jonathan Depends what bit of PCT you need expertise on...
18:46 masak purl: rakudobug
18:46 purl rakudobug is probably mailto:rakudobug@perl.org
18:46 dalek wmlscript: 2071b41 | (Francois Perrad)++ |  (9 files):
18:46 dalek wmlscript: move *.pbc in ./library
18:46 dalek wmlscript: review: http://github.com/fperrad/wmlscript/commit​/2071b419bb8a56356be6601628aa1b37f6aae39c
18:46 shorten dalek's url is at http://xrl.us/beg22f
18:46 dalek wmlscript: ba26db5 | (Francois Perrad)++ |  (2 files):
18:47 dalek wmlscript: add generation of executables
18:47 dalek wmlscript: review: http://github.com/fperrad/wmlscript/commit​/ba26db59891828475cc5ad2a7b4ea124b5eb7e20
18:47 shorten dalek's url is at http://xrl.us/beg22h
18:47 dalek wmlscript: c2387c3 | (Francois Perrad)++ |  (2 files):
18:47 dalek wmlscript: add target 'install'
18:47 dalek wmlscript: review: http://github.com/fperrad/wmlscript/commit​/c2387c3a04e3b7584cfb09482908ff65c84db72e
18:47 shorten dalek's url is at http://xrl.us/beg22w
18:49 particle joined #parrot
18:53 dalek parrot: r36925 | fperrad++ | trunk/tools/dev/mk_language_shell.pl:
18:53 dalek parrot: [install] typo
18:53 dalek parrot: review: https://trac.parrot.org/parrot/changeset/36925/
19:07 dalek wmlscript: 497a299 | (Francois Perrad)++ | config/makefiles/root.in:
19:07 dalek wmlscript: fix Makefile
19:07 dalek wmlscript: review: http://github.com/fperrad/wmlscript/commit​/497a2990dcd79a0da627d049dab70e272a465723
19:07 shorten dalek's url is at http://xrl.us/beg24m
19:07 dalek wmlscript: 8cac149 | (Francois Perrad)++ | wmls (2 files):
19:07 dalek wmlscript: update doc
19:08 dalek wmlscript: review: http://github.com/fperrad/wmlscript/commit​/8cac14990c83d0258c879074c079cb9860281c41
19:08 shorten dalek's url is at http://xrl.us/beg24q
19:42 rblasch joined #parrot
19:43 rblasch left #parrot
19:43 Ron joined #parrot
19:46 Ron left #parrot
19:47 Ron_ joined #parrot
19:48 Ron joined #parrot
19:55 rblasch joined #parrot
20:03 dalek parrot: r36926 | whiteknight++ | branches/rename_pccinvoke:
20:03 dalek parrot: [rename_pccinvoke] update all calls to Parrot_run_meth_fromc_args in src/pmc/object.pmc. Also update one weird test that I think was doing the wrong thing, and a few other changes I had to make for cleaning. two weird failing tests persist, working on that later
20:03 dalek parrot: review: https://trac.parrot.org/parrot/changeset/36926/
20:10 clunker3 joined #parrot
20:21 rurban_ joined #parrot
20:23 Tene joined #parrot
20:43 rurban http://use.perl.org/~rurban/journal/38522 about ptr alignment
20:57 rblasch1 joined #parrot
21:04 jan joined #parrot
21:13 rblasch joined #parrot
21:49 Casan joined #parrot
21:54 jan joined #parrot
22:03 gravity joined #parrot
22:15 afbach joined #parrot
22:30 perlsyntax joined #parrot
22:31 perlsyntax left #parrot
23:25 GeJ Good morning everyone
23:55 Infinoid hi GeJ
23:56 Infinoid busy in here, I see :)

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

Parrot | source cross referenced