github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
Zoffix Talk Slides and Recording: "Intro Into Perl 6 Regexes and Grammars Perl 6": rakudo.party/post/Intro-Into-Perl6...-Recording 06:03
nine ~~ 08:35
.ask jnthn The NativeCall JIT issue reported in github.com/rakudo/rakudo/issues/2036 goes away when run with MVM_SPESH_NODELAY=1. Setting MVM_SPESH_BLOCKING=1 only makes the issue appear earlier (after ~260 iterations instead of ~14000). Also in every test run there's only a single iteration giving a wrong result. Do you concur that this sounds very much like an OSR issue? 08:41
yoleaux nine: I'll pass your message to jnthn.
nine .ask jnthn JIT compiled native calls with rw arguments are the only code that uses the getarg_i OP (for reading back the modified value from the arg buffer). Maybe OSR is not treating that correctly? 08:45
yoleaux nine: I'll pass your message to jnthn.
timotimo nine: you think perhaps something's b0rked in spesh/args.c? 12:13
lizmat timotimo: did you happen to follow developments on github.com/rakudo/rakudo/issues/2138 ? 12:19
timotimo not sure about the QAST::WVal vs Var thing, but have you looked at the compile_time_value method and such? 12:22
lizmat afaik, QAST::Var doesn't have a compile_time_value method ? 12:23
timotimo i don't remember clearly how the compile time value stuff works
lizmat tries
timotimo i've had really bad sleep; a combination of "too warm" and "the cat was terrorizing me more than half the night" 12:25
so don't expect too much from me in the near future %) 12:26
lizmat timotimo: :-( feel better soon! 12:27
timotimo i'll try!
Geth MoarVM: 647fcb9f2a | MasterDuke17++ (committed using GitHub Web editor) | src/core/interp.c
Fix typo that prevented MVM_GC_DEBUGging bindlex
 20:43
MasterDuke can a #define work if there's a space between the '#' and the 'define'? e.g., `# define x y` 20:49
timotimo yes
MasterDuke good 20:50
btw, what were you expecting GC_DEBUG to tell us? i never saw any difference in behavior or output
timotimo oh? 20:51
it should give us errors as soon as something bogus is assigned somewhere
instead of when the next GC runs happens to stumble upon it
MasterDuke hm. i had tried with it just set to 1, now trying 2 20:52
just printed to stderr?
timotimo i think it'll panic 20:53
MasterDuke no panic, just segfaults 20:56
timotimo 367 MVM_gc_worklist_add(tc, worklist, &(tc->instance->all_scs[sc_idx]->sc)); 21:01
(gdb) print sc_idx
$2 = 53606176
travis-ci MoarVM build failed. MasterDuke17 'Fix typo that prevented MVM_GC_DEBUGging bindlex' 21:03
travis-ci.org/MoarVM/MoarVM/builds/409616873 github.com/MoarVM/MoarVM/compare/7...7fcb9f2a4d
MasterDuke t/nqp/019-file-ops.t again 21:05
that's a pretty big sc_idx 21:07
travis-ci MoarVM build passed. MasterDuke17 'Fix typo that prevented MVM_GC_DEBUGging bindlex' 21:09
travis-ci.org/MoarVM/MoarVM/builds/409616873 github.com/MoarVM/MoarVM/compare/7...7fcb9f2a4d

Powered by Cro and the Raku® Programming Language.

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

Thank you!