timotimo it turns out i didn't end up bringing ndarrayviews to a functioning POC today 00:35
MasterDuke timotimo: how good are you with Java? 02:00
timotimo not good 02:01
MasterDuke ha, neither am i (that's why i'm asking!) 02:02
well, to anybody really, any suggestions for my ArrayIndexOutOfBounds exception when trying to create a ClassReader with an InputStream? 02:03
timotimo i don't know anything about classloaders
well, when i worked on the jvm port of rakudo, AIOOB was always about having push operations not followed by the right amount of poppiness 02:04
maybe the other way around, too
MasterDuke i was definitely surprised to get that error, i'm not doing anything with an array myself, this is the problem line: github.com/MasterDuke17/nqp/blob/s....java#L331 02:06
timotimo well, if you're using getResourceAsStream, you're probably giving ASM an array of bytes or something
ASM is *really* bad at turning exceptions that relate to ASM internals to sensible exceptions about its functionality
like, it's the worst
MasterDuke yeah. i've never used it before, it hasn't been the easiest stuff to understand 02:07
timotimo but yeah ... got no clue :(
it's actively resisting being understood, clearly.
anyway, i haven't touched the *Reader or *Visitor stuff of ASM at all 02:08
i'll be off to bed now :|
but i wish you the best of luck
MasterDuke .tell psch still having the same problem, any suggestions? this is the problem line: github.com/MasterDuke17/nqp/blob/s....java#L331
yoleaux2 MasterDuke: I'll pass your message to psch.
MasterDuke timotimo: thanks, later... 02:09
samcv perl6.wtf uses roast right? 04:35
timotimo that's correct 04:36
samcv ok cool :)
timotimo be aware: it's not unfallible 04:37
like, it seems to say there's not a single test for subbuf-rw
... i didn't actually check
Generated on 2016-10-07 - so it's one and a half months out of date
it takes exceedingly long to generate, i suppose 04:40
viki timotimo: actually, last time I ran the script, it failed because it couldn't merge master moarvm on top of the branch with coverage support 05:29
psch . 05:36
yoleaux2 02:08Z <MasterDuke> psch: still having the same problem, any suggestions? this is the problem line: github.com/MasterDuke17/nqp/blob/s....java#L331
MasterDuke psch: i tried upgrading asm to 4.2 and 5.1, but got the exact same behavior 05:38
psch MasterDuke: i'm trying the other accept candidate on ClassVisitor right now
"Any attribute whose type is not equal to the type of one the prototypes will not be parsed"
MasterDuke: not sure that will help, but yeah 05:39
MasterDuke "cr.accept(cv, new Attribute[] { new CommentAttribute("") }, 0);" from the asm docs?
psch yeah
well, with the right class name :) 05:40
timotimo viki: probably because of ops not really being mergable by git itself 05:41
MasterDuke yep. i think i tried that before and it didn't seem to work, but that was a while back in my experimenting
psch well, the asm user guide has quite a bit of "attributes are hard and clunky to use" 05:43
so maybe using normal annotations -- or just stuffing a String where just the line number used to be -- is the cleaner solution
MasterDuke well, it did work when it loaded the class file from the local directory 05:44
yeah, annotations might be better overall, but i think they'd still have the same problem 05:45
that is, you'd still need to create a ClassReader+Visitor, and that's where it's failing 05:46
psch no, it's not
it's failing during reading, not during creation
at least i'm fairly sure about that, because that's what AIOOB usually means 05:48
MasterDuke but then how come it works when using the "String classname" candidate when the class is extracted out of the jar and in the CWD? 05:49
psch well, stepping through it with jdb it does find the resource 05:51
main[1] dump info.compUnit.getClass().getClassLoader().getResourceAsStream("3CE0973F925D7507D9B062C5E98B7D9871761704") 05:52
info.compUnit.getClass().getClassLoader().getResourceAsStream("3CE0973F925D7507D9B062C5E98B7D9871761704") = {
buf: instance of byte[12827177] (id=1887)
so the .accept doesn't do the right thing
MasterDuke i also tried creating a byte[] manually and using that candidate, exact same thing 05:53
psch yeah, sure would be nice if asm had reasonable error messages :/ 05:58
aha (?) 06:01
MasterDuke intellij-support.jetbrains.com/hc/...-Exception 06:02
psch so, apparently readClass calls readUTF8 inside the ClassReader
MasterDuke ^^^ is that any help?
psch there we read an unsigned short
and *that* is out of bounds of the constant pool
...probably not
but i'll look
"I had an entry "/home" which also got scanned for classes. This is where this error occured." from the link 06:03
that does look completely unrelated
MasterDuke thought so, but grasping at straws 06:04
psch hmm
so, yeah 06:05
readUTF8 looks into the CONSTANT_Utf8 items
and apparently there isn't 40k of those
well, the 'SETTING::*' strings are in the class file... 06:06
maybe the ClassReader messes something else up and doesn't read the Attributes correctly first..? 06:07
like, i'm guessing that custom attributes don't really go into the normal constant pool
'cause we'd get "class file to large", just from adding all the filenames
like, we've been scraping on the upper limit of the constant pool for quite a while 06:08
soo i'd guess it's really just looking at the wrong thing, apparently
well, i'm pretty much out of ideas 06:11
MasterDuke hmm, i saw some SO Q&As that mentioned manually walking the classpath and opening jars and looking inside them (or something like that). think that's required?
psch it shouldn't be
...unless the ClassLoader we're using throws custom attributes away
which is possible, i suppose?
MasterDuke well, i used the same one before when it worked 06:12
psch right, except in one instance you're loading from CLASSPATH and in the other you're loading what has been loaded to instantiate the class 06:13
...i think
MasterDuke ohhhhh
psch i'm a bit fuzzy on what exactly getResourceAsStream does there
i'm honestly not really sure that makes sense 06:14
MasterDuke i mean it sounds plausible, but my Java-fu is weak 06:15
looks like arnsholt wrote ./src/vm/jvm/runtime/org/perl6/nqp/jast2bc/JASTCompiler.java, maybe he'll have a suggestion 06:17
psch maybe, yeah 06:20
MasterDuke i just built the version that works when the class is in CWD, maybe i'll try using jdb to see it's doing 06:21
because i do like the look of "in throw (SETTING::src/core/Exception.pm:64)" 06:23
psch wiki.jvmlangsummit.com/pdf/23_Kuleshov_asm.pdf looks like you're doing exactly the right thing, anyway 06:24
pages 13-16 is what i'm looking at fwiw
MasterDuke that's reassuring 06:27
btw, how did you run jdb? 06:30
psch perl6-jdb-server 06:31
it starts suspended, so you need to run 'jdb -attach 8000 -sourcepath $paths-seperated-by-colons' and "run" then
"src/vm/jvm/runtime/:nqp/src/vm/jvm/runtime/:src/:nqp/src/:.:nqp/" is my $paths-seperated-by-colons, when running -attach from the rakudo repo root 06:32
MasterDuke perl6-jdb-server with my -e '...'? 06:35
psch whatever you like, actually 06:36
perl6-jdb-server is a drop-in replacement for perl6-j
it just starts a jdb session as well and suspends it before execution starts
MasterDuke it's past sleep time for me, i'll have to pick this up later. thanks for the help 06:40
.ask arnsholt if you don't mind backlogging, psch and i are running into some ASM troubles and git suggests you might be able to help out 06:42
yoleaux2 MasterDuke: I'll pass your message to arnsholt.
moritz lizmat++ # typo fix 07:03
yoleaux2 20 Nov 2016 22:51Z <lizmat> moritz: in blogpost: "and nothing as fixed" s/as/was/ ?
[TuxCM] This is Rakudo version 2016.11-15-g014d4cf built on MoarVM version 2016.11-8-gb582aeb 07:29
csv-ip5xs 3.120
test 13.836
test-t 6.430
csv-parser 14.789
lizmat Files=1154, Tests=53764, 210 wallclock secs (12.89 usr 3.76 sys + 1280.05 cusr 116.85 csys = 1413.55 CPU) 11:04
viki .tell TimToady never mind. japhb said the reason is mostly due to val() being done prior to addition of fancy pants minus ( irclog.perlgeek.de/perl6-dev/2016-...i_13587383 ) 12:25
yoleaux2 viki: I'll pass your message to TimToady.
lizmat re. val() and MAIN: perhaps we *do* need a EnumStr type and val processing of strings looking like enums? 12:27
that would make it at least consistent across the board ?
nine lizmat: I've had similar thoughts, too
viki The idea occured to me, but the problem I see is it makes code sensitive to allomorphs flaky 12:28
m: my $set = set <meows>; say $set{'meows'}
camelia rakudo-moar 014d4c: OUTPUT«True␤»
viki ^ now that will break the second I happen to have an enum "meows" in the scope, because it'll now be an EnumStr
masak that is a really good point 12:29
lizmat but wouldn't $set<meows> then do work ? 12:30
viki Yes, but what if 'meows' is a variable, coming from elsewhere? 12:31
lizmat yeah, treu
true
.tell TimToady any thoughts on val() also processing enums in scope ? 12:51
yoleaux2 lizmat: I'll pass your message to TimToady.
lizmat .tell TimToady aka: having val("True") returning Bool::True 12:54
yoleaux2 lizmat: I'll pass your message to TimToady.
arnsholt That sounds kind of icky. For that to work, val() would have to go groping in its caller's lexical environment 13:07
yoleaux2 06:42Z <MasterDuke> arnsholt: if you don't mind backlogging, psch and i are running into some ASM troubles and git suggests you might be able to help out
lizmat arnsholt: it would be as simple as ::("True") 13:09
arnsholt Sure, that'd work for enums defined in the setting 13:11
But if you have an enum defined in user code, that's not in val's lexical environment
Ohnoes, ASM! That's a long time ago... 13:12
arnsholt goes scrollbacking
lizmat m: enum E <A B C>; dd ::("A") # arnsholt
camelia rakudo-moar 014d4c: OUTPUT«E::A␤»
lizmat ah, in val() lexical env
hmmm.... 13:13
m: sub a($a) { ::($a) }; sub b { enum E <A B C>; dd a "A"}; b # still seems to work outside of lexical env, arnsholt ? 13:16
camelia rakudo-moar 014d4c: OUTPUT«E::A␤»
lizmat ah, that's because enums are our by default
jnthn iirc enum is our-scoped by default
yoleaux2 20 Nov 2016 23:21Z <lizmat> jnthn: something needs fixing in the bowels: github.com/rakudo/rakudo/commit/014d4cff2f
lizmat :-)
jnthn I don't think that one is too far up the bowels 13:17
arnsholt psch, MasterDuke: It's not quite clear to me what exactly is going wrong. Do you have a gist or something? (Also, the JAST compiler is mostly an integration of code jnthn originally wrote which was external to the compiler)
jnthn I think in BOOTSTRAP.nqp you'll find it registering the int/num ref types and it's probably copy/paste-tweak a line or two
arnsholt Oh, I thought enums were lexical. In that case I guess it'll be less icky 13:18
lizmat jnthn: will look
jnthn arnsholt: Well, they are if you stick a "my" on 'em
psch arnsholt: afaiu, a ClassVisitor over an InputStream/byte[] cannot find the CONSTANT_Utf8 pool correctly, in comparison to a ClassVisitor over a .class file from the filesystem
arnsholt: mind though that the behavior stays consistent over asm 4.1, 4.2, 5.0, so it's unlikely to be an actual bug in their code
arnsholt jnthn: True, true. Then I guess the question is what would val() do in that case 13:19
psch: Weird. So where does the InputStream come from?
psch arnsholt: github.com/MasterDuke17/nqp/blob/s....java#L331 throws ArrayIndexOutOfBoundsException
arnsholt: i dug into the asm source, and that's an access onto a String[] that gets populated with CONSTANT_Utf8 13:20
arnsholt: github.com/MasterDuke17/nqp/blob/s....java#L291 is where we write the Attribute, fwiw 13:21
arnsholt Never encountered that stuff, I'm afraid
psch i don't think we can annotate java methods with different source files, can we? 13:24
as in, a SourceFileAnnotation is actually classfile-scoped
arnsholt I haven't a clue, sorry
psch 'cause that's what i'm thinking is more sensible in the end. the asm user guide (paraphrased) says Attributes are weird, and Annotations are easier, so maybe we can coopt an existing Annotation instead... 13:25
MasterDuke the builtin SourceFile attribute is once per class, so won't work since we have a single giant class 13:27
but maybe there's an Annotation that would work 13:28
irclog.perlgeek.de/perl6-dev/2016-...i_13595901
docs.oracle.com/javase/specs/jvms/...vms-4.7.10 13:29
jnthn: looks like arnsholt passed the blame on to you. any thoughts re the AIOOBE with a ClassVisitor over an InputStream? 13:36
jnthn AIOOBE could be anything 13:41
It generally means "you called/mis-matched methods"
I've had it blow up that way when leaving too many things on the stack, iirc
Just crappy error reporting on ASM's part :(
dalek kudo/nom: 6d726f8 | lizmat++ | src/Perl6/Metamodel/BOOTSTRAP.nqp:
Hook up native shaped str arrays

  jnthn++ for the pointer		 13:43
lizmat afk& 14:28
FROGGS o/ 14:41
timotimo yo 14:42
dalek kudo/nom: cb9df2b | (Zoffix Znet)++ | src/core/allomorphs.pm:
Teach val() to handle U+2212 minus

Includes handling in forms as −42, 42e−72, and 42−72i		 15:22
ast: d904753 | (Zoffix Znet)++ | S32-str/numeric.t:
Test stringification of numerals with U+2212 minus in them
 15:24
viki eeeehhhh.... s/stringification/parsing from string/; 15:25
timotimo unstringification, then? 15:34
viki Precicely :) 15:36
japhb wishes it was Sunday, not Monday ... still in the mood for personal hacking
viki It's not Sunday!? 15:37
OH CRAP I FORGOT TO GO WORK!!
viki giggles
timotimo the bad thing about sundays is that stores don't sell you stuff
i much prefer saturdays for that reason
jnthn Many of them here actually do.
viki timotimo: what country is that at?
timotimo germany
psch stores here have open from 9 am till 12 am on sundays i think 15:38
timotimo you can go to gas stations and get stuff there, but the prices are pretty steep
psch which is kinda sorta like not open
viki You can get stuff here any day of the week. But I dislike national holidays for that same reason.
timotimo mhm 15:39
jnthn I used to live walking distance to Austria. First time I took a stroll there on a Sunday, I quickly understood I would not be getting a snack/drink there to help the walk back. :)
japhb (day of week discrimination)-- 15:40
viki Thou shall not buy snaks on a Sabbath 15:41
ilmari how about snakes?
perlpilot That's just one of the older forms of social engineering. Take a day of rest, don't do anything ... no one else will be doing anything either. 15:42
viki speaking of days off... 15:45
viki goes back to playing video games and eating cookies
DrForr And don't buy them on planes, definitely. 15:46
dalek p: b45f0e3 | (Pawel Murias)++ | src/vm/js/RegexCompiler.nqp:
[js] Get $!cstack from a cursor correctly and prepare for a VMArray being stored there.
 17:24
p: 62f0624 | (Pawel Murias)++ | src/vm/js/nqp-runtime/reprs.js:
[js] Clone VMArrays correctly.

  [js] Handle more things returned from .Num when numifying.
p: bc6f14a | (Pawel Murias)++ | src/vm/js/Compiler.nqp:
[js] Fix bug.
kudo/nom: bfe89a5 | lizmat++ | / (2 files):
shaped(int|num|str)array.AT-POS at least 15% faster

By rewriting in nqp ops: please note there is no optimization for 1/2/3dimmed native shaped arrays yet		 19:51
kudo/nom: e0a415f | (Zoffix Znet)++ | src/core/Str.pm:
Make prefix:<~> actually call .Str on an item

This fixes the issue that using ~ does not coerce an allomorph into its Str component. This is in contrast to prefix<+>, which does.
Remove return type constraint, since we don't know what .Str would return on custom Str subclasses.		 20:00
rakudo/nom: bd42363 | (Zoffix Znet)++ | src/core/IO/ArgFiles.pm:
rakudo/nom: Fix incorrect $*ARGFILES.lines() limit counter
rakudo/nom:
rakudo/nom: Patch by lucasb++
rakudo/nom:
viki github.com/rakudo/rakudo/commit/bd...29b6466566 20:01
dalek ast: c84496e | (Zoffix Znet)++ | S16-filehandles/argfiles.t:
Test $*ARGFILES.lines($limit) works across files

Test for fix in commit
  github.com/rakudo/rakudo/commit/bd...29b6466566		 20:14
ast: e7adb53 | (Zoffix Znet)++ | S02-literals/allomorphic.t:
Test prefix:<~> coerces allomorphs to Str

  irclog.perlgeek.de/perl6/2016-11-21#i_13606506
Fixed in github.com/rakudo/rakudo/commit/e0a415f453		 20:18
lizmat and another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2016/11/21/...perl-6-入门/ 21:50
cog_ lizmat++ 21:51
vendethiel lizmat++ :) 21:54
jnthn lizmat++ 22:05
A nice week :)
lizmat indeed :-) 22:10
[Coke] www.learningperl6.com/2016/11/16/q...sequences/ - "this site can't be reached" -- my restrictive work proxy, probably. 22:19
perlpilot aye, it works here. 22:20
lizmat [Coke]: works for me too :-) 22:35
dalek kudo/nom: ecc202e | lizmat++ | / (2 files):
shaped(int|num|str)array.ASSIGN-POS at least 10% faster

By rewriting in nqp ops: please note there is no optimization for 1/2/3dimmed native shaped arrays yet		 22:57
|Tux| .tell FROGGS hpux.connect.org.uk only ships 32bit objects :( :( - I will install libffi by hand 22:58
yoleaux2 |Tux|: I'll pass your message to FROGGS.

Powered by Cro and the Raku® Programming Language.

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

Thank you!