Camelia, the Perl 6 bug

IRC log for #moarvm, 2013-07-23

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
02:31 colomon joined #moarvm
03:29 prammer_ joined #moarvm
03:54 lizmat joined #moarvm
04:10 lizmat joined #moarvm
04:35 lizmat joined #moarvm
04:49 birdwindupbird joined #moarvm
06:53 FROGGS joined #moarvm
08:23 donaldh joined #moarvm
09:05 FROGGS diakopter / jnthn: what what generates the instructions for calling an action-method? is that rxtype 'pass', or is it 'subrule'?
09:07 jnthn !cursor_pass is called by rxtype pass
09:08 jnthn uh, code is generated to do taht
09:08 jnthn This in turn calls !reduce, which calls the action method
09:08 jnthn see Cursor
09:08 FROGGS k, thanks!
10:52 colomon joined #moarvm
10:59 chipdude joined #moarvm
11:14 diakopter lololol; as if I'd remember anything at all from that file
11:16 FROGGS :P
11:38 diakopter srsly
11:50 benabik joined #moarvm
12:00 diakopter hm.  IMing your CEO at 5 a.m.... career-limiting-move? :S
12:01 jnthn Depends if you wrote "lol I'm drunk!!" :P
12:13 diakopter first time IMing him
12:13 diakopter no I didn't write that
12:16 jnthn ok, good :P
12:22 diakopter HE DOESN'T NEED TO KNOW I'M DRUNK!!!!!!11!1!!
12:22 diakopter (*giggle*)
12:22 jnthn o.O
12:27 tadzik "I woke you up, lol"
12:36 jlaire joined #moarvm
13:05 FROGGS ha! I fixed it!
13:06 lizmat FROGGS++
13:07 FROGGS was a bit surprising, because I have (almost) no idea what the code does
13:13 diakopter eh, what lines are these
13:13 * diakopter guesses such inscrutable code is mine
13:14 FROGGS diakopter: I'll paste a diff in a sec
13:15 diakopter probably nopaste
13:15 FROGGS diakopter: https://gist.github.com/FR​OGGS/a7aad862205d70c5d7f3
13:16 diakopter well, moving the release() line couldn't have affected anything
13:18 diakopter (and it should go back where it was in case someone adds another fresh_() between them
13:18 diakopter )
13:21 diakopter .. but I can see how completely ignoring what was put into @ins would be a problem...
13:21 diakopter :D
13:22 diakopter (If I wrote that, I probably intended to flatten @ins at the beginning of that trailing []
13:22 diakopter )
13:25 FROGGS I tried flattening @ins rigth after find_meth, but this didnt do the trick
13:26 diakopter well the const_s would also have needed to be fixed :D
13:27 diakopter actually I don't know if | flattens inside [ ]
13:27 FROGGS ahh, yes, that might was the problems
13:32 FROGGS diakopter: that would explain why it exploded like: Cannot find sub &prefix:<|>
13:49 dalek MoarVM: e9fe141 | (Tobias Leich)++ | nqp-cc/src/QASTRegexCompilerMAST.nqp:
13:49 dalek MoarVM: fixed a thinko
13:49 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/e9fe141432
13:49 dalek MoarVM: 5f012d6 | (Tobias Leich)++ | / (6 files):
13:49 dalek MoarVM: added and mappen gethllsym
13:49 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/5f012d6978
13:58 diakopter jnthn: I'm realizing (again) that I can't put moarvm bytecode in the p5 thread at all
13:59 diakopter jnthn: well...
13:59 diakopter actually.
13:59 diakopter I can if P5Val repr throws on every repr op but invoke
13:59 diakopter in fact that needs to happen anyway.
14:00 jnthn invoke isn't really a repr op anyways...
14:01 diakopter argh, I keeping calling those repr ops
14:01 diakopter what do you mean by repr ops
14:01 jnthn Anything you call with REPR(foo)->... :)
14:01 jnthn anything that hangs off st->REPR
14:01 diakopter how is invoke not a repr op?
14:01 jnthn it hangs off st
14:01 jnthn And it deosn't depend on representation
14:02 jnthn It's orthogonal.
14:02 diakopter uhm.
14:02 jnthn (in general, at least)
14:02 jnthn We do have a representation or so that sets a default invoke
14:03 diakopter you mean MVMCode?
14:04 jnthn think that is one, yes...though maybe mvmcfunction does too
14:04 diakopter also MVMCFunction and Lexotic?
14:05 jnthn hm :)
14:05 diakopter ok. I called it a reprop since it's in the repr file
14:05 jnthn yeah. The point is that it is attached to stable.
14:05 jnthn Yeah, agree
14:05 diakopter agree it's in the file?
14:05 jnthn I'm just reluctant to call it that 'cus it doesn't have to be 1:1 with repr.
14:05 jnthn but yes, those ones live int eh repr files
14:05 diakopter no, but neither do any of the other repr ops
14:19 JimmyZ FROGGS: ping
14:19 FROGGS JimmyZ: pong
14:22 FROGGS $ nqp nqp-moar-cc.nqp -e 'use NQPHLLMoar; say(nqp::sprintf("abc %d", [1]))'
14:22 FROGGS abc 1
14:22 FROGGS \o/
14:29 jnthn yay
14:29 FROGGS well, it's not perfect yet:
14:29 FROGGS not ok 3 - two %s directives
14:29 FROGGS #   Actual value: WilliamWilliamBell
14:29 FROGGS # Expected value: William Bell
15:06 FROGGS https://gist.github.com/FR​OGGS/c7bd14c0da153c88f058 # sad FROGGS is sad
15:06 FROGGS I think it does not rest $/ at some point
15:06 FROGGS *rset
15:06 FROGGS *reset
15:24 * JimmyZ push his commits so he can sleep well
15:24 dalek MoarVM: d133f32 | jimmy++ | src/ (2 files):
15:24 dalek MoarVM: small fixes and cleanups
15:24 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/d133f32889
15:24 JimmyZ good night
15:26 FROGGS gnight JimmyZ
15:35 diakopter jnthn: ohh I see; you were saying the repr ops are 1:1 with the repr *at runtime*. I was thinking about at *code time*.
15:40 diakopter FROGGS: does the current .moarvm emitter in nqp bundle nqp runtime too?
15:40 diakopter (as dependencies, even?)
15:41 FROGGS diakopter: I think so
15:42 * diakopter wonders if they're all loaded at startup
15:42 FROGGS hmmm, I'm not sure tbh
15:43 diakopter FROGGS: how long does time say ./moarvm the_program_above.moarvm takes
15:43 diakopter (moar-actions.md)
15:43 diakopter on your computer
15:43 diakopter (once it's compiled)
15:46 diakopter FROGGS: can you time the .moarvm of that gist?
15:46 FROGGS diakopter: where is that gist?
15:47 diakopter sad FROGGS is sad
15:47 FROGGS ahh
15:47 FROGGS hold on
15:49 * diakopter thinks "wow, that is a long time"
15:50 FROGGS real0m0.021s
15:50 FROGGS user0m0.020s
15:50 FROGGS sys0m0.000s
15:51 FROGGS and I had to add 'use QRegexMoar' and had to execute it in the nqp-cc folder, so that it can find ModuleLoader.class
15:52 diakopter o_O
15:52 diakopter probalby should remove that .class extension thing :)
15:53 FROGGS that needs changing nqp
15:53 diakopter oh
15:53 jnthn But now MoarVM ain't a secret, we can do that.
15:53 diakopter you're a secret
15:54 diakopter ;)
15:58 FROGGS diakopter: you dont have an idea where the problem might be? https://gist.github.com/FR​OGGS/c7bd14c0da153c88f058
16:16 FROGGS[mobile] joined #moarvm
16:22 Guest1337 joined #moarvm
16:35 FROGGS joined #moarvm
16:49 diakopter FROGGS: I literally know nothing about the internals of the Actions
16:50 FROGGS diakopter: it is more about $/ IMO
16:50 diakopter yeah, but all that's emitted by Actions, right?
16:50 diakopter or is it in the engine runtime?
16:52 FROGGS $/ is set by the regex engine, the actions are just called after something matched
16:59 FROGGS hmm, this might even happen in nqp-cc/nqp-src/QRegex.nqp
17:08 diakopter I meant in the actions that precede the regex compiler
17:09 FROGGS ahh, k
17:11 FROGGS okay, I guess I can hunt that down... will take a while though
19:21 dalek MoarVM: 9349e0d | (Tobias Leich)++ | nqp-cc/nqp-src/QRegex.nqp:
19:21 dalek MoarVM: use a fresh list/hash for every match
19:21 dalek MoarVM:
19:21 dalek MoarVM: Otherwise we carry matches from previous runs around.
19:21 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/9349e0d73b
19:22 FROGGS jnthn: do you get this too sometimes? "Can't free STables in the GC yet"
19:23 FROGGS I get it for t/moar/threads.t from time to time, and now I get it constantly for the sprintf test when I have more than 93 single tests
19:25 jnthn FROGGS: Yeah, it's a GC bug.
19:25 jnthn FROGGS: Some day, I'll be clever enough to hunt those again.
19:25 dalek MoarVM: b4428c7 | (Tobias Leich)++ | nqp-cc/src/QASTRegexCompilerMAST.nqp:
19:25 dalek MoarVM: little optimization
19:25 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/b4428c7bbb
19:25 FROGGS jnthn: k
19:26 * jnthn hasn't quite been his usual self for the last couple of months...
19:26 jnthn Maybe next week's trip to the Alps will help clear my head...
19:26 FROGGS jnthn: would it help if I add the t/hll/06-sprintf.t? It might be easier to hunt down if it isnt a heisenbug like the thread.t one
19:27 FROGGS jnthn: I wish you all the best :o)
19:27 * TimToady thinks it'll Alp a lot :)
19:27 jnthn FROGGS: Yes, I'm aware it can happen in single threaded contexts too
19:27 jnthn TimToady: :P
19:27 * FROGGS .oO( throws a baguette )
19:54 dalek MoarVM: 0bb8555 | (Tobias Leich)++ | nqp-cc/ (3 files):
19:54 dalek MoarVM: added nqp::sprintf and tests
19:54 dalek MoarVM:
19:54 dalek MoarVM: Note that these tests fail due to a GC bug. The test file is not included in the test targets.
19:54 dalek MoarVM: review: https://github.com/MoarVM/MoarVM/commit/0bb8555f03
21:23 donaldh joined #moarvm
21:59 diakopter jnthn: theoretically we can write a clang extension to teach it about your allocator so Address Sanitizer can track those allocations
21:59 jnthn diakopter: ooh
21:59 jnthn diakopter: Did you get asan to work out on Windows, ooc? Or you been doing it on Linux?
22:00 diakopter just linux; haven't tried it on windows
22:00 jnthn k...what's involved? I think I tried getting it up and running a while ago but failed it
22:00 jnthn Probably 'cus I tried to do it in a VM running an oldish Debian...
22:00 diakopter hrm.
22:00 jnthn Don't need answer now, I will sleep in a ibt :)
22:01 diakopter I will sleep in an ibis.
22:02 jnthn That would be preferable than attempting sleep in this hot room...
22:02 jnthn Pro Tip: don't get a top floor appartment :P
22:18 benabik joined #moarvm
22:30 benabik joined #moarvm
23:50 yoleaux joined #moarvm

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary