| IRC logs at
Set by AlexDaniel on 12 June 2018.
Geth MoarVM/master: 4 commits pushed by ZhongnianTao++, (Samantha McVey)++ 08:10
13:20 lucasb joined
Geth MoarVM: d24f6952b3 | (Deven T. Corzine)++ | 3rdparty/dyncall
Use updated dyncall library with "execstack" fix.

The following bugs relate to the "execstack" flag being set on MoarVM, even though executable stacks weren't actually needed:
... (18 more lines)
MoarVM: e54cf70d18 | (Jonathan Worthington)++ (committed using GitHub Web editor) | 3rdparty/dyncall
Merge pull request #1157 from deven/execstack

Use updated dyncall library with "execstack" fix.
13:55 Kaiepi left 14:46 brrt joined
brrt \o 15:32
Geth MoarVM: samcv assigned to ZhongnianTao Issue gb2313 clang warnings
samcv self-assigned gb2313 clang warnings

Jit some num ops
15:53 Kaiepi joined 16:18 brrt left
nwc10 o/ 16:20
dogbert17 hi nwc10 16:35
not very much happening here atm it seems
interesting, two, seemingly unrelated, bug reports crashes in the exact same place 16:44
M#3114 and M#3115 16:45
hmm 16:48
16:51 brrt joined 17:00 pamplemousse joined 17:22 bisectable6 left, evalable6 left 17:23 pamplemousse left, bisectable6 joined 17:26 evalable6 joined 17:29 committable6 left 17:30 chloekek joined 17:31 committable6 joined 17:48 domidumont joined 18:23 brrt left
Kaiepi there's this weird bug in the code i'm working on that doesn't seem to be related to what i'm doing 18:26
you can initialize an async udp socket just fine and create an MVMOSHandle repr for it, but when you try passing it to MVM_io_socket_async_write_bytes_to, it ends up with a zeroed out struct instead 18:27
i'm pretty sure i'm not overwriting it myself 18:28
timotimo have you ever tried rr? 18:32
it's possible that you're forgetting to mark the pointer to the MVMOSHandle 18:33
and the GC is clearing out the memory where the struct used to live
and the pointer you're following is out of date
just a theory
18:46 bisectable6 left 18:51 bisectable6 joined
Kaiepi rr timotimo? 18:51
timotimo yeah, it lets you play your program execution backwards 18:52
which is a good combination with watchpoints
but i guess if your situation is easy to reproduce, you don't have to debug backwards
you can just set the watchpoint as soon as the MVMOSHandle is created i guess 18:53
19:44 pamplemousse joined 20:40 domidumont left
dogbert17 .seen timotimo 20:51
tellable6 dogbert17, I saw timotimo 2019-08-25T20:12:01Z in #perl6: <timotimo> *shruuuuug*
20:53 AlexDaniel left 20:54 AlexDaniel joined, AlexDaniel left, AlexDaniel joined
dogbert17 any c-hackers around? 21:11
AlexDaniel dogbert17: ehhh… what's the question? 21:23
dogbert17 a stupid question really. Look at this function 21:46
the function is supposed to return an MVMint32 but the last line is 'return (MVMuint32)result;'. Why? 21:47
AlexDaniel squints 21:49
jnthn dogbert17: Hm, that return type should probably be MVMuint32; there's even code there to ensrue it can't be negative.
dogbert17 aha
I just reported that I belive there's a bug in the commit which introduced that function. Perhaps the return type is the problem. 21:50
AlexDaniel looking at what calls it, it's uint throughout, so yeah, seems like a typo 21:51
dogbert17 AlexDaniel: I suspect that this commit,, is 'partly' responsble for M#3114 21:52
AlexDaniel dogbert17: nah I doubt the return type is the cause 21:53
jnthn dogbert17: Unfortunately, only in so far as it permits more inlining.
dogbert17: Meaning it just exposed something that could have happened differently anyway.
releasable6: status 21:54
releasable6 jnthn, Next release in ≈26 days and ≈21 hours. R6 is down. At least 7 blockers. Unknown changelog format
jnthn, Details:
jnthn Ah, no immediate stress to hunt it down :)
AlexDaniel jnthn: we were aiming at 1st of September, but I don't know if that's happening
dogbert17 ok, still I'm quite certain that the commit introduced a SEGV
AlexDaniel well, it's unlikely to happen, because I don't feel like doing release prep this week… 21:55
jnthn Ah, and this issue is already in the blocker list 21:56
So I can find it when glancing at those
dogbert17 that's AlexDaniels fault :-)
jnthn It's convenient, it's one less thing I have to separately remember :)
AlexDaniel yeah, I marked it, because pure perl6 code causing a segfault is a big no-no in my book
especially at the size of that gist 21:57
but I didn't know at the time it was a regression that we missed :)
at some point I used to write scripts with segfaults in mind, i.e. a loop that simply restarts everything and the program picks up the work from where it was dropped on the floor 21:59
dogbert17 the SEGV happens here:
AlexDaniel I don't do that anymore because rakudo no longer does that as often 22:00
dogbert17 it's definitely getting very hard to find bugs in MoarVM :)
AlexDaniel actually, pretty much never for scripts that I have… but that gist by jmerelo is really bad :D 22:01
dogbert17 there's another SEGV that fails on the exact same line 22:02
AlexDaniel dogbert17: well, previously bugs used to find me :) Stuff like this ticket: 22:03
in 3 days it'll be one year with no updates on that ticket, i.e. no new issues 22:04
bisectable6: uptime
bisectable6 AlexDaniel, 3 hours, 18 minutes, and 6 seconds, 218.886719MiB maxrss. This is Rakudo version 2019.07.1-94-gd1f9d2848 built on MoarVM version 2019.07.1-50-gb614a7b4d implementing Perl 6.d.
AlexDaniel and that's on almost the latest rakudo version :)
dogbert17: btw we should point some static program analyzer on moarvm's code 22:07
dogbert17: I'd expect it to find issues like the one you mentioned
dogbert17 I run cppcheck from time to time 22:08
MasterDuke AlexDaniel: one (some?) are already run, but not everybody has access to the output
AlexDaniel MasterDuke: I see
MasterDuke i don't remember which one it is. coverity maybe? i think timotimo knows
dogbert17 it's coverity 22:09
timotimo yes
MasterDuke what's the one that occasionally does free scans on FOSS software and posts writups? something like viva64 maybe? or does it start with a 'p'? 22:10
AlexDaniel looks like you can also get pvs-studio for one year:
MasterDuke ah, that's what i was thinking of
AlexDaniel MasterDuke: that's pvs-studio
dogbert17 it looks as if the last scan was done in May last year ... 22:12
22:12 chloekek left
MasterDuke while people are around, is pretty simple, anyone mind if i merge it? 22:14
AlexDaniel MasterDuke: this stuff makes me wonder why it wasn't done before 22:17
MasterDuke: also, if you merge now I'll bump right away
MoarVM: a9b44444dd | MasterDuke17++ (committed using GitHub Web editor) | src/jit/graph.c
Merge pull request #1162 from MasterDuke17/jit_some_num_ops

Jit some num ops
MasterDuke yeah, i was a little surprised 22:18
AlexDaniel wow, what a sha
jnthn Just glanced over it; it was indeed an easier review than I expected. :) 22:19
rest time o/
22:34 pamplemousse left 23:09 lucasb left 23:52 Kaiepi left