01:18 harrow joined 05:27 ssutch joined
[Coke] updated gist.github.com/coke/8250608 05:32
FROGGS: is your stuff committed now?
06:24 ssutch joined
diakopter hi:-) 06:34
FROGGS[mobile] [Coke]: yes 07:35
09:01 FROGGS[mobile] joined 09:42 hoelzro joined
hoelzro morning #moarvm 09:42
anyone awake?
FROGGS I am, but almost afk
hoelzro mkay 09:43
I found a segfault in Moar, but I'm not sure how to proceed
I went through a backtrace and found the routine in question being called by the compiler
and I know it occurs at a particular bytecode offset
FROGGS well, build MoarVM with --debug and run that piece of code in gdb
hoelzro but beyond that, I'm kinda stumped
that's what I did =)
FROGGS next step: post it :o) 09:44
hoelzro heh 09:49
momentje
pastebin.com/iggHGhRM 09:50
FROGGS: ^
FROGGS umm, the key is already NULL 09:53
hoelzro: btw, you pull in MoarVM and nqp? 09:54
hoelzro yes
NQP = 876d500
MoarVM = 4e2fe17
the key's NULL, but it's because sv->value is NULL in compiler.c on line 400 09:55
FROGGS nqp's rev is 2013.12.1-27-gfdc0be3 09:56
you moarvm is up to date 09:57
but please pull nqp before you proceed
because we got a new stage0 yesterday
hoelzro hmm 09:59
hoelzro wonders if the remote is ok
oh, I see
I'm dumb.
hmm 10:02
FROGGS[mobile] hoelzro: did that work out? 11:04
hoelzro yeah, thanks! 11:05
running spec tests now
FROGGS[mobile] good, so we don't have to hunt such a weird Bug :o)
hoelzro =) 11:07
11:13 ingy joined 11:14 wsri joined 11:47 colomon joined 11:51 [Coke] joined, masak joined
timotimo o/ 13:02
FROGGS o/ 13:15
moritz \o 13:16
hoelzro o/ 13:21
13:54 FROGGS[mobile] joined 14:44 FROGGS[mobile] joined
diakopter FROGGS: what's wrong with the "worse than worse" thing 14:55
FROGGS diakopter: that it is hardcoded, and only covers the failing tests that I have spotted 14:59
diakopter o 15:06
FROGGS: why not use the block stuff I had it emit 15:07
15:17 cognominal joined
diakopter FROGGS: I mean, I thought I had it emit the stuff 15:21
(maybe I didn't?)
15:21 FROGGS[mobile] joined
jnthn yay, we broke 95% :) 15:26
15:32 FROGGS[mobile] joined
[Coke] jnthn: \Ƙ/ 15:41
jnthn Ƙ. NƘGNE Ƙ!
ooh, I see FROGGS++ has been doing more on unicode things too :) 15:43
[Coke]: Did yesterday's run not get pushed yet?
[Coke] nuuuuthin'. 15:53
(just pushed)
jnthn ;) 15:54
S05 is by far the failest, but maybe that's better after FROGGS++ recent work, which I guess didn't make it into yesterday's run? 16:10
FROGGS diakopter: I dunno what you mean... can you point me at code and give concrete advices? 16:12
diakopter ponders 16:14
FROGGS jnthn: we should be at 96.5% next time
jnthn \o/
FROGGS (at least)
jnthn FROGGS: If you fancy another thing to explore, fixing wordcaes probably gets us a handful more passes 16:16
FROGGS jnthn: I will look into that
(besides doing v5 stuff)
jnthn When I get a moment, I can take a look at why we abort 300 typed exceptions tests too 16:17
moritz compiles MoarVM on his raspberry pi 17:11
\o/ success! 17:14
jnthn woo 17:15
Can you get NQP to build on it? :)
17:19 _ilbot joined
[Coke] r: say 27507 - 27135 # moar tests today. 17:21
camelia rakudo-parrot 58307f, rakudo-jvm 58307f: OUTPUTĀ«372ā¤Ā»
jnthn 372 moar passing? :) 17:22
nice
[Coke] that's in today's not-yet-pushed-waiting-for-jvm run, aye.
17:27 _ilbot joined
[Coke] only 637 failures (plus a few fudges that were done.) 17:27
FROGGS I need way more time to kill the 261 failures, because I need to understand better how it is supposed to work
[Coke] 13 TODO Passed in moar. 17:28
r: say 261 / 637
camelia rakudo-parrot 58307f, rakudo-jvm 58307f: OUTPUTĀ«0.409733ā¤Ā»
17:30 _ilbot joined
FROGGS ohh 17:33
jnthn [Coke]: Which segfaulted?
I think it may be related to exception handlers and gather/take...
FROGGS S26-documentation/why.t............................ 18 0 0 1 19
1 skipped: segfault
this?
moritz doing things on a CPU-starved device really gives you a different perspective on speed issues :-)
jnthn I'd imagine... 17:34
FROGGS: No, I think segv is elsewhere...in S14 or so
If it's smae one as ysterday
FROGGS S02-types/subset.rakudo.moar.......................Segmentation fault 17:35
S12-subset/subtypes.rakudo.moar....................Segmentation fault
that is sub...
optimal 17:36
moritz /home/pi/p6/nqp/install/bin/moar --libpath=src/vm/moar/stage0 src/vm/moar/stage0/nqp.moarvm --bootstrap --setting=NULL --no-regex-lib --target=mbc \ 17:48
--output=gen/moar/stage1/nqpmo.moarvm gen/moar/stage1/nqpmo.nqp
MAST::Local index out of range
at gen\moar\stage2\QAST.nqp:4757 (src/vm/moar/stage0/QAST.moarvm:assemble_to_file:21
that's on the raspberry pi 17:49
[Coke] hurls gist.github.com/coke/8250608 - segfault at S02-types/subset.rakudo.moar
[Coke] wonders if there's any more he can do to avoid hacking on moar actual. 17:57
18:00 FROGGS[mobile] joined
jnthn moritz: I think that may be an out of date NQP? 18:01
moritz jnthn: I cloned like an hour ago 18:02
jnthn oh, and your MoarVM is also up to date? 18:07
moritz aye, freshly cloned too 18:11
jnthn hm, odd 18:12
'cus I updated stage0 yesterday...
18:12 FROGGS[mobile] joined
jnthn I saw somebody had this in the backlog 18:12
And I think then got it working
bbl
18:33 FROGGS[mobile] joined 19:49 ilbot3 joined 19:57 ilbot3 joined
jnthn yay, think I fixed the S32-exceptions/misc.t crash and got us a bunch more tests 21:36
250 or so maybe
timotimo wowza 21:38
and i thought you were running low on tuits :)
can you give guidance on where to implement the floordiv-operations on parrot? i'm thinking of putting them into nqp's parrot ops
jnthn floordiv? 21:39
timotimo integer division that always rounds towards -Inf
i.e. what we want for perl6
jnthn For bigint?
We have an file of bigint ops in the NQP repo 21:40
*a
timotimo for bigint and normal ints
also, it seems like there's some work that needs to be done for negative mod as well
FROGGS wow
timotimo probably exactly the same thing, haven't looked too closely yet
FROGGS jnthn++
dalek arVM: 116aa55 | jonathan++ | src/core/ (2 files):
Correctly handle active handlers and continuations

Need to slice off handlers since the reset and restore them on invoke.		 21:41
timotimo who's tackling openpipe, btw? FROGGS? :) 21:42
FROGGS >.o 21:43
I dunno
maybe
jnthn Yeha, FROGGS is doing it 21:44
:P
:D
FROGGS :P
gah
I already suffered unicode pain :o)
jnthn Yes, we seem to fail much less of the char props tests now 21:45
timotimo in the .ops files, what does :base_core mean again? 21:46
FROGGS but that mostly satisfies the tests, at least what I did about properties-derived.t
jnthn timotimo: I forget...I think we just scatter it everywhere...
timotimo yeah, that's what i thought %) 21:47
jnthn I think originally it may have been part of some op security model 21:48
timotimo vestiges ... vestiges everywhere 21:51
jnthn ah, I think my fix may only have got us another 150 tests rather than 200...we got up to 490 before, not 400ish 21:52
Gonna look in to the parameterized type stuff 21:53
timotimo parrot checks the result of mp_div for MP_VAL and returns division by zero if that's the case; should moarvm have a similar check in place, too?
jnthn Probably 21:55
timotimo where do i have to look to see how to report such an error in moarvm? 21:56
diakopter timotimo: search "by zero" :) for the other times
timotimo good idea, thanks 21:57
jnthn timotimo: If it's juts throwing an exceptin, look for MVM_exception_throw_adhoc 21:58
r: role R2[$] { }; class C3 does R2[R2[Int]] { }; say nqp::istype(C3, R2[R2[Int]]); 22:00
camelia rakudo-parrot 58307f, rakudo-jvm 58307f: OUTPUTĀ«1ā¤Ā»
timotimo oh, you're tackling that one 22:01
jnthn r: role R2[$] { }; class C3 does R2[Int] { }; say nqp::istype(C3, R2[Int]);
camelia rakudo-parrot 58307f, rakudo-jvm 58307f: OUTPUTĀ«1ā¤Ā»
jnthn r: role R2[$] { }; class C3 does R2[R2[Int]] { }; say nqp::istype(C3, R2[R2[Int]]);
camelia rakudo-parrot 58307f, rakudo-jvm 58307f: OUTPUTĀ«1ā¤Ā»
jnthn Yeah. That line gives 0 on Moar... 22:02
timotimo i saw that line in the test run and thought "oh, maybe i could look at NOPE NOPE NOPE" :P
jnthn :P 22:03
diakopter you just noped on out of there
jnthn Well, I think this bug underlies a few other failures
So, worth hunting
timotimo i NOPed out of there :P
timotimo gets on his NOPsled and rides off
jnthn Interesting that it needs two levels of wrapping to show up 22:04
timotimo www.pauldotcom.com//images/nopsled.jpg ā† someone made this and i am happy
jnthn r: role R2[$] { }; class C3 does R2[R2[Int]] { }; say R2[R2[Int]].HOW.accepts_type(R2[R2[Int]], C3) 22:07
camelia rakudo-parrot 58307f, rakudo-jvm 58307f: OUTPUTĀ«1ā¤Ā»
jnthn r: role R2[$] { }; say R2[Int].ACCEPTS(R2[Int]); 22:13
camelia rakudo-parrot 58307f, rakudo-jvm 58307f: OUTPUTĀ«Trueā¤Ā»
timotimo hm, there shouldn't be many test failures of the parrot perl6 on moar-support, right?
jnthn r: role R2[$] { }; say nqp::istype(R2[Int], R2[Int]); 22:14
camelia rakudo-parrot 58307f, rakudo-jvm 58307f: OUTPUTĀ«1ā¤Ā»
FROGGS timotimo: no 22:15
roast did not move much these days
timotimo weird. 22:16
FROGGS hmmm, wordcase fails because it is unable to look up grapheme cluster breaks... :/ 22:25
I mean, it resolvs the property code and property value code, but the things in the string are not marked as cluster breaks
how do we gain that knowledge? compare properties of a grapheme to the thing right before it? 22:26
FROGGS reads useless-factor.blogspot.de/2007/08/...art-4.html 22:27
jnthn r: role R2[$] { }; say R2[Int].HOW.type_check(R2[Int], R2[Int]); 22:28
camelia rakudo-parrot 58307f, rakudo-jvm 58307f: OUTPUTĀ«0ā¤Ā»
timotimo No such method 'defined' for invocant of type 'Hash' - this seems to kill at least some of my tests 22:29
i'll have a look at nom instead of moar-support 22:31
yeah, no such errors there 22:42
FROGGS unicode.org/reports/tr29/ <--- ohh noes /o\ 22:43
jnthn 96.66% \o/ 22:45
FROGGS very very nice :o)
timotimo FROGGS: what is oh noes about that? 22:46
FROGGS timotimo: I have no idea how to implement that
diakopter it's all the things
timotimo oh
jnthn I've figured out what's going on with the parameterized type stuff
diakopter jnthn: needs one more layer of meta-parameterization? 22:48
jnthn No, one more level of CPS :P
diakopter throws Compiling With Continuations at jnthn... literally 22:49
jnthn ducks
TimToady feels like an accessory before the fact
tadzik related: www.youtube.com/watch?v=7DP7bfAfK2o 22:50
diakopter tadzik: 22:51
dalek arVM: f62b406 | (Timo Paulssen)++ | src/math/bigintops.c:
throw div by 0 in bigint division
 22:56
diakopter timotimo: same for mod? 22:58
timotimo probably should, yeah
diakopter DIOT
i mean
DOIT
timotimo IDIOT 22:59
er i mean
I DO IT
FROGGS *g*
TimToady I dot i.
timotimo should it say "Division by Zero" or "Division by Zero in mod" or something? 23:00
timotimo tends towards the former
23:03 colomon joined
jnthn Not sure Zero needs uppercasing 23:04
timotimo it would if it were "Division by Zorro"
diakopter Division by g0v 23:05
TimToady it's not really division by zero, because that's impossible
timotimo Diakopter by 0
diakopter o_O 23:06
timotimo is it sensible to check for the second operand to be 0 to throw an error or should i rather catch SIGFPE or something? 23:07
dalek arVM: 7cd98af | jonathan++ | src/6model/6model.c:
Fix check/accepts non-cache case of istype.
 23:08
jnthn That helps some tests
timotimo i made a big mistake in the zero-division test anyway
dalek arVM: c73a5ed | (Timo Paulssen)++ | src/math/bigintops.c:
throw div by 0 in mod, too
 23:09
arVM: 6550f75 | (Timo Paulssen)++ | src/math/bigintops.c:
fix copypasto, lowercase "zero"
arVM: 2bcb384 | (Timo Paulssen)++ | src/core/interp.c:
div_i should throw Division by zero, too.
 23:12
jnthn afk & 23:15

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!