00:39 rurban joined 00:48 ventica joined 01:12 FROGGS_ joined
ventica_desktop OK... rakudo cloned, built with MoarVM backend, perl6 works at the command-line... (pause for applause) now, how do I check out the MoarVM test suite? 01:57
02:10 cognome joined 02:29 jimmyz joined
jimmyz ventica_desktop: MoarVM doesn't have test yet, there is a Perl 6 one github.com/perl6/roast/ 02:30
05:10 bcode joined 05:14 lue joined
sergot o/ 06:28
06:44 zakharyas joined 06:53 lizmat joined 07:00 woolfy joined 07:14 Ven joined 07:46 woolfy joined
masak \o 08:56
jnthn o/ 08:57
09:00 lizmat joined
nwc10 \o/ \o/ \o/ 09:00
FROGGS_ ventica_desktop: you can also run nqp's test suite, this is the first thing one can check if MoarVM fundamentals got changed 09:35
masak <lizmat> what would be the best way to fix this? 10:39
can we patch upstream (dyncall)? 10:40
can we send a pull request?
can we apply a patch on our side while we wait for their side to fix it?
10:49 brrt joined
brrt \o #moarvm 10:49
jnthn: thanks for looking into it. i /really/ suspect that this is a call-convention issue (i.e., i borked something when i updated the posix version) 10:50
and that would (or could) explain why bogus stuff is stuffed everywhere
also, i've figured that microsoft lets you borrow VM's (modern.ie), so I've got one now 10:53
whcih works better than my 'real' installation, it seems 11:01
12:19 jnap joined 12:26 jnap joined 13:33 FROGGS[mobile] joined 13:54 btyler joined 14:15 dalek joined 14:22 ggoebel111116 joined 14:46 jnap1 joined 15:17 brrt joined 15:19 btyler__ joined
japhb
.oO( MoarVM running as an ActiveX control )
 15:37
16:03 cognome joined 16:24 cognominal joined, cognome joined
ventica FROGGS: Is it a separate step to clone nqp test suite? I don't see any nqp tests under t/ 16:33
timotimo ventica: these tests are in the nqp repository under t/ 16:34
just think of NQP as a separate language that's meant to be used for creating language compilers and happens to be the one we're using for rakudo 16:35
see also examples/rubyish
ventica ok 16:41
timotimo well, that's not exactly a helpful explanation, i suppose 16:42
ventica jnthn had told me I shud run t/spec/S32-io/IO-Socket-Async.t to exercise the async socket I/O but I don't see any of those folders under t/ 16:44
timotimo ah
well, that's in the rakudo spectests :)
inside the *rakudo* repository, if you "make spectest-checkout" 16:45
ventica hmm, i see
timotimo (the repo itself is github: perl6/roast)
ventica_desktop rakudo$ make spectest-checkout 16:47
make: *** No rule to make target `spectest-checkout'. Stop.
timotimo oh? 16:50
ah
spectest_checkut
damnit :)
spectest_checkout
17:02 zakharyas joined 17:48 lizmat joined 18:00 lizmat joined 18:01 woolfy joined, woolfy left 18:04 btyler joined 18:16 woolfy joined 18:18 jnap joined 18:22 vendethiel joined 18:23 Ven joined
ventica_desktop ah, underscore to the rescue... 18:54
rakudo$ ls t/spec/S32-io/IO-Socket-Async.t 18:55
t/spec/S32-io/IO-Socket-Async.t
nice
FROGGS_ yeah, that tests fails different tests from time to time for some reason 18:57
ventica_desktop rakudo$ ./perl6 t/spec/S32-io/IO-Socket-Async.t 18:58
===SORRY!===
Could not find Test in any of: ., blib, /home/claytonkb/rakudo/rakudo/install/languages/perl6/lib, /home/claytonkb/rakudo/rakudo/install/languages/nqp/lib
8| 18:59
19:01 woolfy left
jnthn -Ilib 19:01
FROGGS_ either that or install rakudo and use the perl6 or perl6-m that got installed (expand your PATH) 19:08
ventica_desktop rakudo$ ./perl6 -Ilib t/spec/S32-io/IO-Socket-Async.t 19:14
1..6
ok 1 - Async listen on bogus hostname
ok 2 - Async connect to unavailable server breaks promise
ok 3 - Async connect to available server keeps promise
ok 4 - Echo server
ok 5 - Discard server
ok 6 - bytes_supply
so looks like test 1..6 are passing on my machine 19:16
FROGGS_ four and six fail sometimes on my box 19:18
maybe only under load, like when running the spectest suite with TEST_JOBS=4 19:19
19:20 Ven joined
jnthn Yeah, there's some race in there, I fear... 19:20
19:24 brrt joined
brrt just figured out the IE VMs are 32 bit... 19:29
anyway, i think i've the solution to moar-jits windows worries 19:32
basically, windows likes 4 arguments in registes, and 4 on the stack, right? 19:33
eh i mean the rest on the stack
posix likes 6 integer arguments in registers, 8 floating point arguments in SSE registers, and the rest on the stack
19:33 Ven joined
brrt most functions i call have fewer than 4 arguments, some have 5 or 6, and i know of only one that has 7 arguments (all integer /pointer) 19:34
well, when i put an argument on the stack, i do so by simply pushing a 64 bit quadword from a register to the stack
but that means that an argument that is expected to be 32 bits wide takes up 64 bits 19:35
jnthn ahh 19:36
I did wonder if it could be ABI related 19:37
brrt and since there is only one function that takes 7 arguments (and there are no more), and since this is a simple boolean, this problem doesn't show on posix 19:38
it had to be ABI related :-) there's no difference anywhere else
fixing will take up a bit of work, though 19:39
brrt afk
20:26 itz_ joined 20:32 itz joined 20:52 ggoebel111116 joined 21:41 brrt joined 22:27 ventica joined 22:28 ventica_desktop joined 23:14 cognome joined, cognominal joined, cognome joined 23:24 ggoebel111116 joined 23:40 itz_ joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!