04:56 vendethiel joined 06:03 pyrimidi_ joined 06:11 colomon joined 07:15 ggoebel joined 08:20 zakharyas joined
nwc10 jnthn: paste.scsys.co.uk/501255 -- ASAN barfs on t/spec/S32-str/encode.rakudo.moar 08:34
09:15 domidumont joined 09:24 kjs_ joined 10:56 Peter_R joined
ilmari nwc10: that's probably my fault 11:27
ilmari has a look
i >= result_alloc - repl_length # fails when repl_length > result_alloc 11:34
all the variables are unsigned
MVMuint32 or size_t
nwc10 I'm not familiar with that code, so can't comment 11:35
and a bit too much $work is happening currently
timotimo well, looks like ilmari is about to fix it 11:36
ilmari how us(able|eful) is ASAN on gcc 4.9.2 (debian)?
timotimo i think it's usable 11:37
you'll need to install that extra package, but it'll be fine
and we have the helpful --asan thing for Configure.pl 11:38
ilmari yeah, I found that
timotimo good good
nwc10 ilmari: the question is ASAN on that debian generally? Or ASAN for MoarVM? 11:45
ilmari nwc10: in general 11:46
nwc10: I was just wondering if I should install a newer gcc or clang, or whether I can just build with --asan on stock debian stable 11:47
nwc10 OK, I don't know.
timotimo actually, i wonder if we should enable ASAN on our travis builds?
especially since we're usually ASAN clean
arnsholt Using the stock compiler should be OK for ASAN and such 11:48
nwc10 ilmari: the ASAN I'm using is gcc (GCC) 4.9.0 11:52
a local build on CentOS $something 11:53
(it's "my" machine, not my machine)
we get some hangs on ASAN builds in the tests for various concurrency features
timotimo oh, damn
nwc10 I assume that these are race conditions we hit due to slower execution times
but they'd really hurt on an automated build 11:54
and however much we go
jnthn++
there still only seems to be one of him
and I think he's got other things to look at first
timotimo that's fair 11:55
jnthn People who can debug concurrency bugs seem to be few and far between 11:57
nwc10 good UGT, jnthn 11:59
jnthn o/
nwc10 there's a lot of contention on the time of the folks who can debug concrrency bugs
timotimo %)
jnthn And, sadly, debugging concurrency bugs seems to need to acquire a mutex on their brane 12:00
12:36 nine joined 12:37 arnsholt joined
konobi how does the build/setup.pm work? (in terms of OS vs Platform vs toolchain) 12:55
13:40 kjs_ joined 14:02 zakharyas joined
ilmari nwc10: I have a fix for the overflow in ascii (and latin1, windows1252 and utf8(_c8)), but utf16 is still borked 14:16
because it does no checking at all, just assumes that any charcter can fit in 4 bytes
16:20 retupmoca joined 17:01 kjs_ joined 17:45 kjs_ joined 17:59 kjs_ joined 18:16 domidumont joined 18:51 kjs_ joined 20:16 kjs_ joined 20:23 zakharyas joined
timotimo ilmari: will we get that as a pull request or patch? :) 20:25
jnthn suspects PR 20:30
I think ilmari is hacking at lunchtimes though, so we'll likely have to wait until lunchtime happens again ;)
nwc10 and being ilmari, we might have to wait several days until he remembers that it's lunchtime 20:31
[Coke] crap, lunch. 20:33
nwc10 oh no, not *another* ilmari 20:34
ilmari nwc10: accept no imitations! 20:39
psch hm, an arg with a non-MVM_CALLSITE_ARG_OBJ arg_type doesn't have a container spec, right?
ilmari jnthn: my home laptop has started having thermal issues, so I prefer anything that involves lots of compiling on my work workstation 20:40
timotimo accept no limitations!
psch or is that just not reachable with .o but instead the corresponding typed... something (register?)
[Coke] I'm not actually hungry, despite it being nearly 4pm here. weird. 20:44
nwc10 I "just" outsource my compiling to shell accounts other people give me 20:45
jnthn psch: Um...I think it's unreachable *at the moment* 20:46
Not sure it should be
psch jnthn: because that's what it looks like is missing 20:48
as in, cached dispatch between Int and Int is rw works 20:49
but bring an int is rw and int in there it only dispatches the the latter
jnthn Oops. :) 20:51
jnthn wonders if the cahce needs to learn about native refs...
psch gist.github.com/peschwa/92605dd3ec50054f9a9d is my test case, just to make sure i'm not messing that up again... :) 20:56
jnthn psch: Need to rest now; will look in the monring :) 21:14
konobi afternoon all 21:26
hoelzro o/ konobi 21:27
21:39 njmurphy joined 22:54 kjs__ joined