01:54 pyrimidine joined 02:06 pyrimidine joined 02:15 pyrimidine joined 02:19 FROGGS_ joined 02:34 pyrimidine joined 02:46 pyrimidine joined 02:48 ilbot3 joined 03:54 pyrimidine joined 04:06 pyrimidine joined 04:13 pyrimidine joined
JimmyZ FROGGS_: I suspect changing readdir_r to readdir may fix your problem :) 04:50
FROGGS_: anyway, I will commit it someday :P
geekosaur um. making it less thread safe does not sound like a good fix 04:52
(granting that readdir/readdir_r thread safety is a bit dubious to begin with)
JimmyZ readdir_r has some problem womble.decadent.org.uk/readdir_r-advisory.html too 04:53
since it's low api, who want to keep it more safe, can write it with a lock or something by perl6 :P 04:56
FROGGS_ JimmyZ: let me do that... then I can hack in some proper detection... 09:04
09:37 brrt joined
brrt \o 09:37
stmuk . o O ( that womble url is scary! )
yoleaux2 18 Oct 2016 17:52Z <FROGGS> stmuk: have a look at github.com/MoarVM/MoarVM/commit/0402247933
brrt i need one minor name thing for the JIT (and then I'm done, sort of) 09:38
stmuk FROGGS_: yeah that moar change fixed one of my probs! cheers 09:43
FROGGS_ JimmyZ: it is not about readdir vs readdir_r directly... it is more about "struct dirent entry;" being not enough 09:45
sizeof(struct dirent)=20
(offsetof(struct dirent, d_name) + NAME_MAX + 1)=274
NAME_MAX=255
nqp and rakudo pass now all tests on solaris 11.3 x86_64 \o/ 09:53
dalek arVM: fdfc1f9 | FROGGS++ | build/setup.pm:
request POSIX.1.c compliance on solaris

Otherwise we get funtions compatible to POSIX.1.c Draft 6 only.
10:05
JimmyZ FROGGS_: yeah, but seems that readdir() doesn't have this issue. 10:34
FROGGS_ true, but I'm going the "safe" way :o) 10:37
10:40 lizmat joined
dalek arVM: c0a2997 | FROGGS++ | src/io/dirops.c:
alloc enough memory for readdir_r on solaris
10:44
arVM/multidimarray_view: deadf05 | timotimo++ | src/6model/reprs (4 files):
WIP on MultiDimArrayView REPR.
12:17
timotimo (transfer from laptop to desktop)
dalek arVM/readdir: 15c8190 | (Jimmy Zhuo)++ | src/io/dirops.c:
change to use readdir insteal of readdir_r
12:41
JimmyZ FROGGS_: Could you try this one? I hope it will fix all the platform which have some issue with readdir_r 12:43
FROGGS_: not just on solaris
12:54 vendethiel joined 13:13 TimToady joined 13:32 lizmat joined
FROGGS_ JimmyZ: and what about thread safety? 13:58
JimmyZ FROGGS_: we don't use dirhandle by multi-thread in rakudo 14:00
FROGGS_: we just have a dir() function
dalek arVM: 15c8190 | (Jimmy Zhuo)++ | src/io/dirops.c:
change to use readdir insteal of readdir_r
14:03
FROGGS_ JimmyZ++ # all tests pass
JimmyZ FROGGS_: I think I can merge it? 14:04
anyone who wants to use dirhandle by multi-thread through nqp ops , should give it a mutex :) 14:06
though readdir is thread safe in gnu libc 14:07
oh, you merged it 14:09
FROGGS_ JimmyZ: I did
dalek arVM: 656143c | (Jimmy Zhuo)++ | src/io/dirops.c:
code clean up
14:13
JimmyZ FROGGS_: thanks 14:14
14:48 domidumont joined
FROGGS_ sad that the qemu sparc64 emulator seems not be reliable... 16:16
dalek arVM: 8047226 | FROGGS++ | build/setup.pm:
use gcc on solaris by default
16:21
16:56 zakharyas joined
timotimo i think i'll just make setstrides an op that explicitly handles *View only, rather than adding that to the associative reprops all things share 17:38
dalek arVM/multidimarray_view: b024ad3 | timotimo++ | / (5 files):
MultiDimArrayView can compile now

need to implement setting and getting stride and calling that via custom ops now ...
18:07
18:24 domidumont joined
dalek arVM: 90d7f1e | FROGGS++ | / (2 files):
give diagnostic output when compiling fails
19:17
arVM: 340cd0d | FROGGS++ | build/setup.pm:
current solaris compiler does not understand -mt
19:31
arVM: 1f41fd5 | FROGGS++ | Configure.pl:
query libffi header directory on solaris
20:39 leego joined 20:42 leego joined 20:54 lizmat joined