00:06 AlexDaniel` joined 00:37 ilmari[m] joined 00:53 patrickz_ joined 00:59 evalable6 joined
AlexDaniel huggable: bots 01:43
huggable AlexDaniel, The #perl6 irc channel normally hosts several helpful bots. I am a bot, and everyone else voiced on this channel is a bot. See the full list here: github.com/perl6/doc/issues/711#is...-235414744
AlexDaniel MasterDuke: fwiw, maintainers should be listed for each bot here ↑ 01:44
01:47 domidumont joined 01:50 domidumont1 joined 01:55 domidumont joined 01:58 bisectable6 joined
MasterDuke ah right, forgot about that 01:58
japhb AlexDaniel: Should it be more aware of which IRC channel it is speaking on, rather than referring only to #perl6? 02:56
AlexDaniel huggable: that's just a static factoid, so it's probably ok
huggable AlexDaniel, nothing found
AlexDaniel japhb: ↑ :)
02:58 ilbot3 joined 03:54 domidumont joined 03:56 domidumont1 joined 06:04 releasable6 joined, greppable6 joined 06:05 squashable6 joined
Geth MoarVM: samcv++ created pull request #761:
Improvements to ucd2c.pl
06:33
07:10 geospeck joined 07:35 reportable6 joined 07:58 zakharyas joined 08:13 zakharyas joined 08:18 zakharyas1 joined 08:20 brrt joined
brrt good * #moarvm 08:45
yoleaux 6 Dec 2017 21:00Z <jnthn> brrt: SEGV in linear scan allocator: gist.github.com/MasterDuke17/45323...4409368975
brrt ouch, thank you
08:48 geospeck joined 09:02 domidumont joined
brrt damnit, that doesn't crash for me 09:25
and i can't start valgrind :-(
i'll try and start a virtual machine then 09:26
nwc10 brrt: paste.scsys.co.uk/566116
and only yesterday I was going to "complain that ASAN was getting bored, but then I though "be careful what you wish for" 09:27
does anyone have a colo(u)r pastebot so that I could paste it in all its pavement-pizza glory?
brrt that'd be cool, yes 09:29
oh, that's a pretty good stack trace though 09:31
that gives me virtually all the necessary information
nwc10 I'm probably AFK in about an hour from now, so if you have time to cook up something to (re)test before then, it should be able to 09:38
otherwise, not sure when
09:49 [Coke] joined 09:50 Util joined
brrt i'm a bit puzzled still 09:50
09:56 [Coke] joined
nwc10 oops, be careful what you wish for 10:28
ASAN barfage in JIT code paste.scsys.co.uk/566117 10:29
brrt: is that the same place?
afk for a bit 10:30
10:51 domidumont joined
brrt yes, it is 11:16
same place
not sure why still!
lizmat brrt: I suspect some initialization issue, as it either happens after ~ 200 msecs, *or* it runs to completion without any issues in about 950 msecs 11:20
I have not seen any crashes after 200 msecs
brrt hmm, odd 11:22
anyway, valgrind / ASAN makes it crash quite relaibly
the question is, why the use-after-free
because the valuerefs are allocated from a single buffer that doesn't grow, and in fact that's not the growing buffer 11:23
oh, hang on 11:24
i know why
lizmat whee! 11:30
brrt double-pointer trick wreaks havoc with the realloc 11:47
that would've been a good title for a blog though 12:27
meh, i'll be saddened to see it go 12:29
Geth MoarVM: 5484f048ee | (Bart Wiegmans)++ | src/jit/linear_scan.c
[JIT] Don't take a pointer-pointer to a moving object

This would cause segmentation faults because the alc->values array
  (which backed the spillee object) could move in a realloc, and the
pointer-pointer to the ValueRef would point inside this object. We don't actually need that, we can get by with two separate pointers and a comma.
Note that we'd also assign to the (potentially moved) spillee object, which is what I suspect was the real cause to the SEGV, but ASAN already found the invalid read earlier.
12:42
brrt .tell lizmat jnthn nwc10 fixed
yoleaux brrt: I'll pass your message to lizmat.
nwc10 brrt: paste.scsys.co.uk/566118 -- src/jit/linear_scan.c:720:48: runtime error: member access within null pointer of type 'struct ValueRef 12:49
brrt fuuuuuuuuu 13:07
:-(
oh, i see
yeah
hmmm
13:09 geospeck joined 13:16 domidumont joined
Geth MoarVM: 924a75d556 | (Bart Wiegmans)++ | src/jit/linear_scan.c
[JIT] moreover, don't read a NULL struct

If the next pointer is already NULL it makes no sense to read it.
13:45
brrt .ask lizmat if the —asan option is working for you, because, i think i have a way to fix it on OSX
yoleaux brrt: I'll pass your message to lizmat.
lizmat --asan is not working for me :-) 13:46
yoleaux 13:45Z <brrt> lizmat: if the —asan option is working for you, because, i think i have a way to fix it on OSX
brrt okay, then i may have a fix for that as well 13:47
Geth MoarVM: f2c35e1333 | (Bart Wiegmans)++ | Configure.pl
[ASAN] fix flags for darwin

Apple is ... exceptional, about how to setup flags for ASAN support
13:49
brrt by the way 14:07
i've been thinking of a way to write a generic 'top-down tree matching funciton
14:33 zakharyas joined 14:52 squashable6 joined 14:54 reportable6 joined, releasable6 joined, arnsholt joined, buggable joined, huggable joined, ilmari joined, ZofBot joined, nine joined, mst joined 14:55 bartolin joined, japhb joined, nebuchadnezzar joined, geospeck joined, zakharyas joined, [Coke] joined, Util joined, brrt joined, patrickz_ joined, BooK joined, zostay joined, TimToady joined, harrow` joined, jpf joined, lizmat joined, samcv joined, Voldenet joined, jnthn joined, dogbert17 joined, timotimo joined, krunen joined, eater joined, leedo joined, btyler joined, camelia joined, jsimonet joined, ingy joined 15:05 sivoais joined 15:07 geospeck joined 15:09 geospeck joined 15:16 MasterDuke joined, ChanServ joined
brrt anyway, what i wanted to say was 15:17
the idea was to combine an array with a string that tells me how to interpret the string
almost like data and instructions
this is a pattern, hmmm
15:20 sivoais joined 15:36 AlexDaniel` joined 16:03 geospeck joined 16:32 zakharyas joined 17:24 zakharyas joined 17:55 AlexDaniel` joined 18:03 domidumont joined 18:06 domidumont1 joined 18:38 AlexDaniel` joined 18:43 domidumont joined 18:48 domidumont joined 19:01 AlexDaniel joined 19:12 domidumont joined 19:14 domidumont1 joined 19:16 ilmari[m] joined 19:19 travis-ci joined
travis-ci MoarVM build errored. Bart Wiegmans '[ASAN] fix flags for darwin 19:19
travis-ci.org/MoarVM/MoarVM/builds/312964683 github.com/MoarVM/MoarVM/compare/9...c35e1333e3
19:19 travis-ci left 19:36 unicodable6 joined, committable6 joined 20:22 domidumont joined 20:33 benchable6 joined 21:48 mst joined
timotimo src/strings/unicode.c:72236:83: warning: pointer/integer type mismatch in conditional expression 22:06
? Bidi_Mirroring_Glyph_enums[0] : Bidi_Mirroring_Glyph_enums[result_val]) : bogus;
should we be worried about this at all? i suppose it's not a problem if it's "bogus"?
22:10 bisectable6 joined 22:15 greppable6 joined, squashable6 joined 22:42 mst joined 23:00 evalable6 joined