ugexe fyi: i'm getting TravisCI build failures with `language: perl` on OSX (if the next rakudo commit fails, this is likely why) 02:14
03:29 vendethiel joined 05:03 vendethiel joined 05:59 sno joined 06:52 RabidGravy joined
dalek kudo/nom: a63deba | (David Warring)++ | lib/NativeCall (2 files):
define native type ssize_t. size_t should be unsigned

Fixes RT #127823
07:08
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=127823
dalek kudo/nom: 6cb2319 | niner++ | lib/NativeCall (2 files):
Merge pull request #760 from dwarring/ssize_t

define native type ssize_t. size_t should be unsigned
07:37 travis-ci joined
travis-ci Rakudo build errored. niner 'Merge pull request #760 from dwarring/ssize_t 07:37
travis-ci.org/rakudo/rakudo/builds/126580095 github.com/rakudo/rakudo/compare/3...b231924d80
07:37 travis-ci left
[Tux] This is Rakudo version 2016.04-48-g6cb2319 built on MoarVM version 2016.04 08:01
test 22.245
test-t 13.100
csv-parser 23.469
RabidGravy Hmm it has been wibbling upward slightly for the last few days 08:17
psch grr jvm *shakes fist* 09:08
TAP uses &signal 09:09
which means we have a non-building r-j again
in the interest of portability i'd try and expose only those that translate to "shut down everything" and expose a ShutdownHook for that 09:14
i think that's the best we can do without relying on sun.misc, which is unofficial and all that 09:15
jnthn Interesting. S12-methods/qualified.t fails its second test under GC torture 09:18
m: class A { }; say ~A.new
camelia rakudo-moar 6cb231: OUTPUT«A<140212425128552>␤»
jnthn Reason being that if you do
m: class A { }; say ~A.new eq ~A.new
camelia rakudo-moar 6cb231: OUTPUT«False␤»
jnthn It uses the memory address
Which relies on the object not moving. But you're doing string concat to form the string A<...> 09:19
Which can trigger GC
uh, I shoulda done the same A instnace there of course
m: class A { }; loop { my \a = A.new; die "oops" unless a eq a; }
camelia rakudo-moar 6cb231: OUTPUT«oops␤ in block at /tmp/jufArJEyz4 line 1␤␤»
jnthn m: class A { }; loop { ++state $i; my \a = A.new; die "oops $i" unless a eq a; } 09:20
camelia rakudo-moar 6cb231: OUTPUT«oops 12371␤ in block at /tmp/QXxR9HCHk3 line 1␤␤»
jnthn m: class A { }; loop { ++state $i; my \a = A.new; die "oops $i" unless a eq a; }
camelia rakudo-moar 6cb231: OUTPUT«oops 12371␤ in block at /tmp/5lx8DtDidF line 1␤␤»
jnthn This has quite possibly caused various heisenbugs over time.
nine_ So it should use nqp::objectid instead? 09:27
jnthn Yeah, got a local patch for that 09:28
And building a MoarVM without a 13KB nursery so I can build Rakudo fast enough :) 09:29
Also a fix for a spurious wakeup vulnerability in Promise.result.
Yeah, patch fixes the busted test and the above loop seems to run "forever" without dying 09:39
nine_ \o/ 09:41
jnthn isn't have any luck making S17-procasync/no-runaway-file-limit.t erupt with all the various fixes 09:59
Wonder if that means I fixed it or the bug's just playing hide and seek really well...
Ah, it deadlocks 10:03
RabidGravy boom 10:11
nine_ I guess a boom would be easier to debug than a silent deadlock ;) 10:18
jnthn Well, the deadlock is seemingly that GC orchestration one that bites now and then 10:20
Which I plan to take on soon, but know is independent of the call frames changes.
In fact, I suspect everything I'm looking at is, by now.
Though really want to get to the bottom on this S15-nfg/many-threads.t one 10:21
*of
dalek kudo/js: f16c03d | (Pawel Murias)++ | src/vm/js/Perl6/Ops.nqp:
[js] Add missing file.
11:38
11:39 pmurias joined
dalek p: 428de5f | (Pawel Murias)++ | src/vm/js/ (2 files):
[js] Stub nqp::neverreposses. It's won't do anything till we implement repossession.
11:39
p: 9f8a952 | (Pawel Murias)++ | src/vm/js/Operations.nqp:
[js] Add more support to defining ops from compiled code.
RabidGravy just in case people aren't reading the other channel, I'm going to start a "Perl 6 Noise Gang" organisation on GH to promote and/or bring focus to audio/music stuff, any objections from anyone? 12:17
jnthn So...Perl 6 can result in line noise too? :D 12:18
Sounds cool to me :) 12:19
RabidGravy yeah, been meaning to do something like it for a while
psch oh bugger 12:34
this whole signal stuff probably can't work with a shutdown hook i think
'cause that might run too late for the TC and all that to still be around
actually, i'm just doing things wrong, probably... 12:43
fwiw, it does seem to want to run the tap'd code on the signal 13:15
but i don't really know how i can pass arguments there
and well, it hangs and has to be kill -9'd, too
hm, is there already a *handle that could let me pass the argument through to the asynctask..? 13:22
actually, i think i misunderstood this 13:30
every signal from @signals gets tapped with a closure that takes the signum
so i can just create AsyncTasks that invokeDirect and pass the signum for the signal they belong to, right 13:31
13:59 skids joined
psch oh ffs java 14:22
jnthn A common reaction to writing Java :)
14:23 vendethiel joined 14:25 dalek joined 14:26 geekosaur joined, psch_ joined
psch 14:22 < psch> "public Foo(int tpyo) { this.typo = typo }" 14:27
14:22 < psch> of course that compiles
14:23 < psch> well, plus the semicolon
from before my early-aborted timeout... :) 14:28
timotimo ugh
psch i suppose that's an argument against "name the constructor arguments the same as the attributes" 14:29
and it really makes me appreciate our BUILD syntax magic 14:30
and, well, the inherited .new as well
timotimo good point! 14:31
14:37 JimmyZ joined, masak joined, [Coke] joined
psch $ ./perl6-j -e'say "ok..."; signal(SIGINT).tap({ say "k, bye" }); sleep 1000' 14:43
ok...
^Ck, bye
that's not that bad i suppose :)
unfortunately it NPEs during install-core-dist.pl /o\
psch summons bartolin++ 14:44
there's a precedence for bartolin and fixing NPEs... :P
hm, unless that's the NPE that PR 748 fixes..? 14:54
...apparently so 15:05
dalek kudo/nom: 6ea0927 | usev6++ | src/core/Parameter.pm:
Avoid NPE when parameter has no coerce_type

This lead to NPE in some files under S32-trig/ One golfed example for dying code:
   class Foo is Real { }
15:06
kudo/nom: 6964266 | usev6++ | src/core/Parameter.pm:
rakudo-j no longer needs nqp::getattr_i here

was fixed with nqp commit 5981bf0439
rakudo/nom: 120f56d | peschwa++ | src/core/Parameter.pm:
15:11 sno joined
dalek p: d54d719 | peschwa++ | src/vm/jvm/ (2 files):
Add rudimentary signal workaround.

This is the best we can do considering the mechanism of supporting signal() on the JVM. The supported signals could probably be expanded to "everything that ends the receiving process", but I don't know enough about POSIX signals to decide which signals belong into that group.
15:11
kudo/nom: eaa5834 | peschwa++ | tools/build/NQP_REVISION:
Bump NQP_REVISION

This gets us basic almost-signal-handling on nqp-j, see the corresponding nqp commit message for details.
kudo/nom: 1acf805 | peschwa++ | / (3 files):
Handle the few supported signals on R-J.
psch there, nom head builds r-j again
now for a spectest too see how that looks \o/ 15:12
16:02 travis-ci joined
travis-ci Rakudo build passed. peschwa 'Merge pull request #748 from usev6/NPE_S32_trig 16:02
travis-ci.org/rakudo/rakudo/builds/126679039 github.com/rakudo/rakudo/compare/6...0f56d200c9
16:02 travis-ci left 16:04 vendethiel joined
psch m: say "{6069 / 60} minutes for make j-spectest" 17:03
camelia rakudo-moar 1acf80: OUTPUT«101.15 minutes for make j-spectest␤»
psch total of 23 tests failed
gist.github.com/peschwa/fb61ff5f57...53d42f7a45 17:04
well, unless i miscounted... :)
bartolin psch: nice! (the fails in S05-mass/properties-general.t don't happen with jdk 1.8 (cause of newer unicode version)) 17:19
dalek ast: feda328 | usev6++ | S06-other/main-usage.t:
Fudge (todo) test for RT #127977
17:22
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=127977
dalek ast: 0e0084f | usev6++ | S02-types/version-stress.t:
Skip 3 dying tests on JVM
19:16
psch oh, yeah, the whole NFG stuff is gonna be fun too 19:19
well, and the unicode support before that even
like unival etc
well, maybe that'll finally teach me about unicode ;) 19:20
timotimo it'll be a challenge, yeah :\
i mean, you can just steal our code for moarvm
that part generates C code, which is almost compatible with java code anyway 19:21
psch well, i don't really want to side-step the jvm-inherent unicode support complete if i don't have to... 19:24
otoh, i probably do have to, because we don't want to rely on a given jvm version or higher...
anyway, maybe *i* specifically don't have to because someone comes and along and says "oh boy implementing unicode in java!" and does it instead :P 19:25
[Coke] psch: I've been dreading it for a year now. :) 19:26
timotimo :D
jnthn It shouldn't be too hard to crib the NFG impl from Moar 19:27
Updating everything to use it will be trickier.
Though I guess you can introduce overloads of all the ops that take String to make them take NFGString or something 19:28
And then update the code-gen, rebootstrap, and whee
psch never have ever i heard the debug-cycle in java describes as "whee" :P 19:30
timotimo :D 19:32
java has hot code swapping!
geekosaur [29 15:30] <psch> never have ever i heard the debug-cycle in java describes as "whee" :P ...do you count sarcasm? :p 19:33
vendethiel timotimo: yeah, when it works :v 19:37
timotimo i've never tried it
but i'm jealous 19:38
that's now for android, too
22:16 skids joined 22:21 tomboy64 joined 22:40 tomboy64 joined