| unmatched} | p6: "hello.o: hello.c" ~~ / .+ ':' \s* $<stuff>=.+ /; say $<stuff> | 02:41 | |
| camelia | rakudo-moar e97fb7: OUTPUT«｢hello.c｣» | ||
| dalek | arVM/even-moar-jit: 6079744 | brrt++ | / (2 files): Work-in-progress pieces for linear scan |
06:35 | |
|
06:52
TheLemonMan joined
07:26
zakharyas joined
|
|||
| TheLemonMan | another day, another bug triaged! and once again it's at the MoarVM level :P | 07:56 | |
| RT#127075, caused by the second branch in the or (github.com/perl6/nqp/blob/master/s....nqp#L890) that causes $_ (a unicode string) to be numified via smrt_numify -> MVM_coerce_smart_numify -> MVM_coerce_s_n which in turn calls MVM_string_ascii_encode and hell breaks loose when it encounters a non-ascii codepoint | 07:58 | ||
| synopsebot6 | Link: rt.perl.org/rt3//Public/Bug/Displa...?id=127075 | ||
| TheLemonMan | I'm also pretty sure there has to be a better way than to recompile mvm, nqp, and rakudo each time :| | 08:03 | |
| nine | TheLemonMan: you don't have to recompile nqp and rakudo after changes to MoarVM | ||
| TheLemonMan | nine, I get some errors I didn't write down if I don't recompile everything :| | 08:04 | |
|
08:35
kjs_ joined
08:50
domidumont joined
|
|||
| jnthn | Curious what errors, I do that all the time and it works out fine... There are changes that need it but you gotta be adding ops, pretty much. | 08:54 | |
| TheLemonMan | then I probably forgot to 'make install' the updated nqp, sorry for the noise | 09:01 | |
| nine | TheLemonMan: big difference between nqp and moar | 09:02 | |
| You do need to recompile rakudo when changing nqp | |||
| TheLemonMan | yeah, learned that the hard way | 09:03 | |
|
10:26
kjs_ joined
10:40
edehont joined
|
|||
| Zoffix | TheLemonMan so you tried Perl 6 because of one of my articles and now you're fixing issues in things I've not even learned yet? | 10:44 | |
| Zoffix feels old and stupid :) | |||
| TheLemonMan | Zoffix, I think the best way to learn a language is to start digging into the internals :) | 11:30 | |
| now I got an 'opcode validation error' that gives me an offset but afaics there's no (immediate) way to dump the bytecode stream | 12:08 | ||
|
12:08
kjs_ joined
|
|||
| jnthn | moar --dump foo.moarvm | 12:09 | |
| TheLemonMan | the next question is about how to get the moarvm dump from p6 | 12:11 | |
| jnthn | Compile with --target=mbc --output=foo.moarvm :) | 12:12 | |
| TheLemonMan | now I feel silly heh, thanks jnthn | 12:13 | |
| hmm, moarvm segfaults when I try to run it now | 13:14 | ||
| and the --dump switch generates a lot of spurious trailing characters | 13:19 | ||
| a patch like ptpb.pw/lWQN stops the segfault, you might come up with a better message though | 13:55 | ||
|
13:56
kjs_ joined
14:03
leedo joined
|
|||
| TheLemonMan | apparently a 'rakudo_scalar' is missing | 16:38 | |
| timotimo, the problem disappears if you disable the jit | 16:59 | ||
| oops, s/jit/spesh/ | 17:02 | ||
|
17:32
zakharyas joined
|
|||
| timotimo | which one are we talking about right now? | 17:38 | |
| the "perfect numbers" thing that segfaults? | |||
| TheLemonMan | yes, that one | 17:40 | |
| timotimo | and the patch you had on ptpb.pw? was that for the same crash? | 17:41 | |
| TheLemonMan | that prevents moarvm to crash when I try to run a .moarvm file dumped from p6 | 17:42 | |
|
17:47
kjs_ joined
|
|||
| timotimo | oh | 17:54 | |
|
18:00
brrt joined
|
|||
| brrt | good * #moarvm | 18:01 | |
| jitbugs? | |||
| unmatched} | \o\\ | ||
| timotimo | good brrt | ||
| brrt | good timotimo | ||
| timotimo | we're potentially kicking out a guard or something | ||
| brrt | oh... | ||
| wasn't that what jnthn was working on yesterday | |||
| timotimo | he had something in that general vicinity | ||
| where a guard was wrongly kicked out in some cases, causing trouble | 18:02 | ||
| brrt | who guards the guards? | ||
| (in unrelated news.. them register allocator 's gonna be /sweet/) | 18:03 | ||
| one of the funnest things i've had the opportunity to write in a while | 18:04 | ||
| timotimo | whoa | ||
| i'm glad to hear it! | |||
| brrt | glad to finally have some time for it | 18:06 | |
| of course i have to fix a whole bunch of stuff before it can be fully implemented | 18:07 | ||
| hmm, and i've decided i need to be able to split according to a range, not just for a certain branch | 18:09 | ||
| eh, not just past a certain step | |||
| that means... that means | |||
| that means we're going to need basic block ranges after all | |||
| which was always the plan, mind you | |||
| but it was never necessary before | |||
| and i've learned in hard ways against doing things that aren't yet necessary | 18:11 | ||
| timotimo | hehehe | ||
| right, when things are not necessary yet, but still implemented, cutting them out is extra painful when it turns out neither that nor a dependency of it ended up being needed | 18:12 | ||
| brrt | uhuh | 18:13 | |
| 'but it was beautiful' | |||
| worse is when you start modifying your solution to the existing systems rather than the problem | |||
| timotimo | %) | 18:14 | |
| brrt | there's still open bits, of course | 18:16 | |
| let me see what are the worst ones | 18:18 | ||
| hmmm | 18:19 | ||
| i still don't know how to cleanly fit ARGLIST in there | 18:20 | ||
| but that's it i guess | |||
| it's more a question of much to do | |||
| TheLemonMan is still trying to figure out what libpaths are needed | 18:29 | ||
|
18:52
kjs_ joined
20:36
Ven_ joined
20:39
zakharyas joined
20:56
kjs_ joined
21:28
kjs_ joined
22:10
Zoffix joined
22:22
lizmat joined
22:43
kjs_ joined
|
|||