github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
00:24 patrickz left 00:34 harrow left 00:39 harrow joined 00:58 Kaiepi left 01:01 Kaiepi joined 01:50 Kaiepi left, Kaiepi joined
Geth MoarVM: Kaiepi++ created pull request #1138:
Fix pointer arithmetic in inlined CArrays in CStruct/CPPStruct REPRs		 01:52
07:32 brrt joined
brrt \o 07:32
08:20 sena_kun joined
timotimo o/ 08:43
Kaiepi \o 08:59
09:04 patrickb joined 09:29 brrt left 09:35 brrt joined
brrt pamplemousse++ nice post, nice progres 10:11
11:04 Kaiepi left, Kaiepi joined 11:26 brrt left 12:05 patrickb left 12:25 bartolin left 12:33 bartolin joined
Geth MoarVM: MasterDuke17++ created pull request #1139:
Add context to error messages		 12:46
13:21 lucasb joined 13:42 pamplemousse joined 13:49 releasable6 left 13:51 releasable6 joined 14:02 pamplemousse left 14:10 rfold joined
MasterDuke i think i've asked this before, but if so, don't remember the answer. anyone have an idea why a module that's just a single large hash literal (~50k lines) won't pre-compile? 14:10
and wow, tons of different `JIT: bailed completely because of` 14:13
lizmat MasterDuke: congrats! please make an issue! 14:16
MasterDuke lizmat: heh, there might be one already. someone else discovered this, i've just had the file lying around and try again every once in a while 14:21
14:26 pamplemousse joined
timotimo oooooh my gtk_cairo_shooter runs with barely a modification, and it performs well 14:54
15:59 MasterDuke left
Geth MoarVM/nativecall_errors_give_argument_index: 7a5d41ba94 | (Timo Paulssen)++ | 4 files
NativeCall: output index of argument with repr mismatch

also properly says "native cast" or "return type" for those.		 16:47
MoarVM: timo++ created pull request #1141:
NativeCall: output index of argument with repr mismatch		 16:49
timotimo cheque it out
16:53 brrt joined 16:58 pamplemousse left
japhb timotimo: If you're referring to github.com/timo/gtk_cairo_shooter , that has got to be one of the most minimalist repos I've seen. The code looks 5 years old -- can you push your mods? 18:04
timotimo ha
japhb: just pushed 18:06
japhb Woah, that's quite a change at the top there. Rebased on a new lower layer? 18:09
timotimo just stole some stuff that GDK at that point didn't have i think? 18:14
can perhaps throw that out again
too tired now to try it
those changes are oooold, though
the only change necessary to get it to run was to remove the arguments to gtk_simple_use_cairo
18:59 sena_kun left 19:02 brrt left
nine finds it a bit worrying that it's so easy to cause segfaults and GC panics by mixing sockets and threads. I guess that will affect Cro, too, which we are using more and more at $work 19:10
nwc10 that's not great (the SEGVs, not the "more and more") 19:12
if you build MoarVM with ASAN can you provoke it into ging vaguely useful backtraces?
nine That...could actually be useful. ==31239==ERROR: AddressSanitizer: heap-use-after-free on address 0x6180002f48a0 at pc 0x7f558f6a8c48 bp 0x7f558405ec00 sp 0x7f558405ebf8 19:18
It's using an already destroyed thread context?! 19:21
Why would a thread even get destroyed after the whole program handled just a couple of connections? 19:23
19:26 sena_kun joined
nwc10 I really don't know. You seem to be asking the right questions, but I fear that you're going to have to go digging to try to answer that. 19:26
(I'm giong to step away from the keyboard to get a beer, but I might not return until I've finished it.) 19:27
nine Oooh it's because the example posted in the bug report uses Thread.start
So we really do start a lot of threads that aren't used very long. 19:28
Replacing Thread.start with just start fixes the issues. Probably because we keep those threads around
Apparently it breaks when the spesh worker thread is doing GC and marks e.g. static frames while the thread in question has just been freed by it's owning thread. 19:32
nwc10++ # suggesting the right tool for the job 19:37
timotimo huh, our code is supposed to be just fine with threads ending 19:39
or do you explicitly have to join them?
nine joining doesn't seem to change anything 19:42
Well it does make it fail even faster 19:43
timotimo odd.
Cro almost exclusively uses IO::Socket::Async fwiw
nine I'm not sure the use of sockets is even important in that example
Need to get off the train 19:44
nwc10 :-) 19:50
else you'll end up in Salzburg? Or in Vienna?
(whereas had I not got off the sofa, I would have still been there in the morning) 19:51
as is, I'm going to bed.
timotimo bon appesleep 19:52
20:06 Kaiepi left 20:07 Kaiepi joined 20:08 Kaiepi left, Kaiepi joined 20:09 Kaypie joined, Kaypie left 20:10 lucasb left 20:21 MasterDuke joined
Geth MoarVM: MasterDuke17++ created pull request #1142:
Jit getuniprop_(int|bool|str)		 20:22
20:31 zakharyas joined 20:58 Kaiepi left, Kaiepi joined
MasterDuke does it make sense to jit getcomp and/or freshcoderef and/or scsetobj and/or sp_rebless? 21:12
jnthn sp_rebless is probably the most worthwhile 21:17
MasterDuke they all showed up as `jit bailed completely` in a spesh log of running that file that's too big to pre-compile 21:25
jnthn: speaking of, how would i find why said file doesn't pre-compile? 21:27
jnthn "too big" is a bit under-specified :)
MasterDuke 807kb, 50k lines (all but a couple of those lines are a hash literal) 21:29
jnthn Yeah but...what makes it too big to precompile? How does it fail? 21:30
21:30 zakharyas left
MasterDuke that's what i don't know 21:30
it always takes ~28s to parse, even subsequent runs 21:31
jnthn Oh, so there's no failure, just no speedup? 21:32
Maybe running RAKUDO_MODULE_DEBUG=1 will shed some light on it
MasterDuke fyi, here's the file gist.github.com/MasterDuke17/aa149...e77a89b03c
1 RMD: Loading settings CORE 1 RMD: going to load Perl6::BOOTSTRAP 1 RMD: Settings CORE loaded 1 RMD: Loading settings CORE.d 1 RMD: Settings CORE.d loaded 21:35
that's the entirety of the output with RAKUDO_MODULE_DEBUG=1 21:36
afk for a bit, should be back later 21:37
jnthn You'd have to `use` it as a module to get any interesting out related to it 21:42
jnthn also afk a bit
21:49 rfold left
Geth App-MoarVM-ConfprogCompiler/master: 7 commits pushed by (Timo Paulssen)++ 22:02

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!