Parrot 6.9.0 "Rosy-faced Lovebird" | parrot.org/ | Log: irclog.perlgeek.de/parrot | #parrotsketch meeting Fridays 13:30 UTC
Set by moderator on 22 October 2014.
00:49 kid51 joined 03:27 bighugedog joined 04:00 bighugedog joined 06:41 bighugedog joined 07:32 FROGGS joined 08:45 bighugedog joined 09:18 rurban joined
rurban I think I found the GC bug. Coro->caller_seg 09:19
FROGGS rurban++ # \\o/ 09:20
rurban Nope, one more 09:22
but rlimit and better gc debugging tracers really helps 09:23
Yes, it's definitely a Coro, coming from Parrot_pcc_reuse_continuation. Nice 09:29
FROGGS that might solve a handful issues we have in rakudo (which shows up in rakudo star, some bundled modules fail) 09:37
rurban it's only on low memory, now with rlimit very easily repro. 09:38
It's someone pointing to a Coro, need to find it 09:39
09:51 atrodo joined 09:52 ivan joined, bighugedog joined 09:59 mtj- joined 10:00 slavorg joined 10:33 kurahaupo joined
dalek rrot/rurban/gcdebug-gh1108: 2de8aa5 | rurban++ | / (8 files):
[gc] add -D2xx MEM_DETAIL_DEBUG_FLAG

pass -D for parrot also down to gc_initialize, not only parrot_old undo --gc-dynamic-threshold=2 for gms, use the standard --gc-nursery-size as documented. add seperate GC_DEBUG_DETAIL_2 tracer to seperate it from MEM_DETAIL_DEBUG_FLAG tracing for every single alloc/free.
10:38
rrot/rurban/gcdebug-gh1108: 4d5ae76 | rurban++ | src/pmc/ (2 files):
[pmc] add Coroutine.mark with caller_seg

Coroutine.mark was just calling Sub.mark and missed to mark the caller_seg. Do it now properly with inheritance. Note that this does not fix GH #1109, which is a Coro in a wrong gen
10:41 kid51 joined 10:47 bighugedog joined
dalek rrot/rurban/gcdebug-gh1108: 2b3e739 | rurban++ | / (2 files):
[gc] Better -D1 printout. ChangeLog #1108

  $ ulimit -Sv 640
  $ ./parrot -D1
Free Memory: 8589934592 Memory via rlimit restricted to: 655360 GC nursery size: 2.0% GMS GC threshold: 13107
10:49
rrot: fe01f34 | rurban++ | ChangeLog:
ChangeLog: actually fixed make smoke, now stable
11:04
11:05 kurahaupo joined
dalek rrot/rurban/gcdebug-gh1108: 8afd192 | rurban++ | / (9 files):
[gc] decruft -DMEMORY_DEBUG and DETAIL_MEMORY_DEBUG, add -D100

Check -D100 instead of DETAIL_MEMORY_DEBUG when -DMEMORY_DEBUG is enabled. This way we can run -D101 and -D1 to get verbose and non-verbose GC tracings.
Add new MEMORY_DEBUG_DETAIL_2 macro for fprintf helpers all over. Move common panic_failed_allocation into gc_private.h, maybe the other common gc memory funcs also to save space. Print avail sysmem on startup with -D1. GH #1108
11:08
rrot/rurban/gcdebug-gh1108: bc7ae97 | rurban++ | / (4 files):
[core] enable Interp_debug_TEST and flags with flag combinations

Now you can check for multiple flags, e.g. if (Interp_debug_TEST(interp,
   PARROT_MEM_STAT_DEBUG_FLAG | PARROT_GC_DETAIL_DEBUG_FLAG))
Was used only for single flags, so no changes for existing code. Document the single flag limitation for the left-over trace_flags test. There's no -t flag combination, like verbose yet.
rrot/rurban/gcdebug-gh1108: 9cfd6f2 | rurban++ | src/io/ (2 files):
[cage] unconst IO_VTABLE * vtables, -Wcast-qual cannot be suppressed

see e.g. stackoverflow.com/questions/132497...7_13253997
2 remaining -Wcast-qual warnings for legacy API.
rrot/rurban/gcdebug-gh1108: e56decf | rurban++ | / (13 files):
[gc] check -D1 in startup, add -D2xx MEM_DETAIL_DEBUG_FLAG

Print GS and Memory info on -D1. Add seperate GC_DEBUG_DETAIL_2 tracer to seperate it from MEM_DETAIL_DEBUG_FLAG tracing for every single alloc/free. Document the two new -D100 and -D200 flags with --ccflags=-DMEMORY_DEBUG
rrot/rurban/gcdebug-gh1108: 1896d4e | rurban++ | src/pmc/ (2 files):
[pmc] add Coroutine.mark with caller_seg

Coroutine.mark was just calling Sub.mark and missed to mark the caller_seg. Do it now properly with inheritance. Note that this does not fix GH #1109, which is a Coro in a wrong gen
rrot/rurban/gcdebug-gh1108: 79d8950 | rurban++ | / (2 files):
[gc] Better -D1 printout. ChangeLog #1108

  $ ulimit -Sv 640
  $ ./parrot -D1
Free Memory: 8589934592 Memory via rlimit restricted to: 655360 GC nursery size: 2.0% GMS GC threshold: 13107
11:20 travis-ci joined
travis-ci parrot/parrot#1201 (master - fe01f34 : Reini Urban): The build has errored. 11:20
Change view : github.com/parrot/parrot/compare/c...01f34f815c
Build details : travis-ci.org/parrot/parrot/builds/39139853
11:20 travis-ci left
rurban travis internal problem, timeout, killed the test 11:26
11:35 travis-ci joined
travis-ci parrot/parrot#1201 (master - fe01f34 : Reini Urban): The build passed. 11:35
Change view : github.com/parrot/parrot/compare/c...01f34f815c
Build details : travis-ci.org/parrot/parrot/builds/39139853
11:35 travis-ci left 12:51 rurban joined
dalek rrot/rurban/gcdebug-gh1108: f51d164 | rurban++ | / (7 files):
[gc] add trace_pmc_flags_dump

and use it within trace_pmc_dump and #1108, coro in wrong gen
12:58
13:01 TonyC joined
dalek rrot/rurban/gcdebug-gh1108: e26cf72 | rurban++ | / (2 files):
[cage] fixup 2 test errors with pmc flags
13:13
rrot/rurban/gcdebug-gh1108: eed079d | rurban++ | src/ (2 files):
[gc] dont stress the GC when tracing the GC

printf directly to stderr in trace_pmc_flags_dump, when not in the debugger
14:08
rurban $ tail -n 100 log.det2 14:20
tail: error reading ‘log.det2’: Result too large
WTF
1.9G only
16:46 Chirag joined, FROGGS joined 17:49 FROGGS_ joined 18:14 kurahaupo joined 19:32 prammer joined 20:23 ggherdov_____ joined 20:44 prammer left 23:00 kid51 joined 23:36 jsut_ joined