RabidGravy eugh 08:19
lizmat good *, #perl6-dev!
Files=1154, Tests=53760, 213 wallclock secs (12.92 usr 3.70 sys + 1297.88 cusr 121.07 csys = 1435.57 CPU)
RabidGravy has something changed in the implementation of return in the last few months that anyone can think of? 08:20
something really really weird here
[TuxCM] This is Rakudo version 2016.10-307-gbca0ae5 built on MoarVM version 2016.10-172-g1ba1dd2 08:23
csv-ip5xs 3.106
test 13.877
test-t 6.873
csv-parser 14.471
RabidGravy This github.com/sergot/http-useragent/issues/161 seems to be entirely contingent on the return in github.com/retupmoca/P6-Compress-Z...w.pm6#L160 08:26
which seems somewhat odd 08:27
take the return out and it works, put it back in it fails 08:29
I'm not sure how it can be reproduced for yet another RT in the category of "Cannot invoke this object"
lizmat does it also happen if you disable spesh / optimization / jit ? 08:41
[TuxCM]: last bench seems to have a rise all across the board ? 08:48
RabidGravy MVM_SPESH_DISABLE doesn't seem to make any difference, nor does the optimisation llevel 09:07
but yeah it does feel like an optimization sort of thing 09:08
lizmat does it still have the problem if the sub isn't "our" 09:12
would be my last guess
stmuk_ any RPM packager want to fix pkgs.org/fedora-24/fedora-updates-...4.rpm.html ? :) 09:26
(see comments on perl6-users) 09:27
I'm not a RPM user myself
dalek ast: fa97aa9 | usev6++ | S12-methods/attribute-params.t:
Unfudge tests for native attribute binding on JVM
nine stmuk_: fix in what way? 09:58
stmuk_ nine: well by adding the rakudo star modules to it 10:00
nine RabidGravy: return became a proper control exception this summer
stmuk_ "No that's an incomplete version lacking the modules which makes rakudo into rakudo star. Most importantly panda (the module installer) is missing. Neither does it even have full docs since perl6intro.pdf is missing.
nine I really question the utility of distro packages for star. Why not simply package rakudo and the modules individually and maybe create a rakudo star meta package that just depends on those? 10:02
stmuk_ yes I say exactly that on perl6-users .. I was hoping to link to www.nntp.perl.org/ but my email isn't indexed yet 10:04
anyway I predict more perl6 newbies asking how to install panda on Rakudo Star 10:06
nine Sounds like a good little project for my next Freestyle Friday: give packaging modules a real try 10:08
dalek kudo/nom: a3be654 | lizmat++ | src/core/Main.pm:
Show possible enum values in usage message

If we know it's an enum, we might as well show which values would be legal. Make sure though it fits on one line.
10:21
kudo/nom: 3dcc52b | lizmat++ | docs/ChangeLog:
Tell about a3be654
10:23
RabidGravy nine, I was thinking the same thing the other day, we could do with something like Distzilla with plugins for packaging and stuff 10:34
stmuk_ maybe perl 5's cpan2rpm might help 10:39
timotimo it sounds problematic that googling for perl6 instructions didn't get them anywhere near our docs 10:50
lizmat perhaps google bubble: for me "perl6 instructions" rakudo.org/how-to-get-rakudo as first, and perl6.org/getting-started/ as second 10:52
stmuk_ timotimo: I don't understand
timotimo "how to run perl6 script" - "We have included how to run perl script. 10:53
Limit results to how to run perl6 script?
"
wow, duckduckgo is *super* unhelpful for this
stmuk_ I'm getting better perl 6 google results than last time i looked
moritz ddg has some feedback mechanism 10:54
please use it :-)
stmuk_ I would hope "private mode" would break out of the google bubble
lizmat afk& 11:07
MasterDuke does anyone else have problems rebuilding rakudo-j after making changes to nqp-j? sometimes i can make a change in nqp, rebuild, and then rebuild rakudo just fine. other times i get "java.lang.RuntimeException: Missing or wrong version of dependency 'gen/jvm/stage2/QRegex.nqp'" when rebuilding rakudo 12:52
psch MasterDuke: yeah, it's either QRegex or Pod for me, usually 12:53
MasterDuke: i tend to always 'gif clean -xdf' first nowadays
MasterDuke and the only solution seems to delete everything in my --prefix and then rebuild nqp
yep, QRegex this time
annoying, my system doesn't build nqp-j and rakudo-j all that quickly 12:54
psch well, it shouldn't be too hard to figure out what causes the dependency mismatch
or rather, whether it's a real mismatch or if we're just checking the wrong thing
bartolin uses 'git clean -dfx; perl Configure.pl --backends=jvm --gen-nqp && make' 12:55
psch almost the same, except i use --make-install for Configure.pl :)
Zoffix Actually, my last night's concern seems to have been valid. It was just the falsy bug that prevented from the issue from showing up 12:56
m: my @*ARGS = BEGIN <Meow>; sub MAIN (Str $blah) { say "$blah $blah.^name()" }
camelia rakudo-moar 3dcc52: OUTPUTĀ«Meow Strā¤Ā»
MasterDuke perl Configure.pl --prefix=/home/dan/Source/perl6/install/ --backends=jvm && make -j8 install
Zoffix m: my @*ARGS = BEGIN <Truew>; sub MAIN (Str $blah) { say "$blah $blah.^name()" }
camelia rakudo-moar 3dcc52: OUTPUTĀ«Truew Strā¤Ā»
Zoffix m: my @*ARGS = BEGIN <True>; sub MAIN (Str $blah) { say "$blah $blah.^name()" }
camelia rakudo-moar 3dcc52: OUTPUTĀ«Usage:ā¤ <tmp> <blah> ā¤Ā»
Zoffix m: my @*ARGS = BEGIN <True>; sub MAIN ($blah) { say "$blah $blah.^name()" }
camelia rakudo-moar 3dcc52: OUTPUTĀ«True Boolā¤Ā»
psch m: say val("True")
camelia rakudo-moar 3dcc52: OUTPUTĀ«Trueā¤Ā»
psch m: say val("True").WHAT
camelia rakudo-moar 3dcc52: OUTPUTĀ«(Str)ā¤Ā»
Zoffix It doesn't get to val() 12:57
psch hm, and i thought Bool means "this switch doesn't take an arg and is either negated or not"
MAIN parsing is hard :S
Zoffix huh?
My MAIN doesn't expect a Bool
It wasn't a string. I'm passing a string that just happen to be an Enum
psch ohh, right, Enum 12:58
Zoffix s/wasn't/wants/;
psch ...geez, that's been in for like a year iirc
that Bool is an Enum i mean
Zoffix No, since yesterday.
Oh
So basically now our MAIN will convert anything that matches an Enum to its value, regardless of whether we're actually have an Enum type-contraint 12:59
timotimo wait, it goes for every single enum that's in scope?
Zoffix Yes 13:00
uses ::()
github.com/rakudo/rakudo/blob/nom/...ain.pm#L19
timotimo that's magical
Zoffix I'm thinking of reverting this feature for the release, so we could shake out issues like that 13:03
psch shouldn't it be sufficient to first check if the constraint is an Enum?
assuming we have the actual Parameter in scope at that moment
timotimo yeah, the commit is so fresh it might be a good idea to revert it just for the release 13:04
yeah, $m should have all the candidates and such
we should be able to go through the signatures and look for enum-HOWed type constraints
psch oh hm, that kinda replicates MMD though, doesn't it
Zoffix psch, yeah, that should work, but currently the conversion happens before we get to the sub and even then we don't check individual args, but rather check if there are any matching candidates and then just give all the args to Main
psch i mean, if we walk through candidate and whathaveyou...
psch dimly recalls some bits of the structure of MAIN handling 13:05
dalek kudo/nom: e8f938b | bartolin++ | src/core/Rakudo/Internals.pm:
Remove bandaid for r-j, native attribute works no

  psch++ for implementing attributive parameter binding on the JVM
13:07
kudo/nom: e23566e | (Zoffix Znet)++ | src/core/Rakudo/Internals.pm:
Merge pull request #920 from usev6/patch-3

Remove bandaid for r-j, native attribute works now
ast: 47a9ec1 | (Zoffix Znet)++ | S06-other/main.t:
Temporarily fudge enum MAIN tests for release

  irclog.perlgeek.de/perl6-dev/2016-...i_13595383
13:17
kudo/nom: a264f0a | (Zoffix Znet)++ | t/04-nativecall/20-concurrent.t:
Revert "Test no longer flops."

This reverts commit 84eaff47f274b203b8600b78d4149d3de1bc52a1.
As AppVeyor shows the test still flops on Windows:
  ci.appveyor.com/project/moritz/rak...5at2axp4i7
Ticket re-opened:
  rt.perl.org/Ticket/Display.html?id=130042
13:24
jnthn So why not skip it just for Windows? 13:31
Zoffix Fair point. Will do shortly
jnthn Thanks.
afaik it's not failed again on Linux/OSX... 13:32
Zoffix Yeah, seems to be good there.
jnthn Which is...odd, but if that remains the case, it points at a Windows-specific issue.
(And it's nice to keep it running on Linux/OSX Travis to have more data.) 13:33
dalek kudo/nom: 1549161 | (Zoffix Znet)++ | / (2 files):
Temporarily revert sub MAIN accepting Enum

  irclog.perlgeek.de/perl6-dev/2016-...i_13595383
13:35
kudo/nom: 25b0a55 | (Zoffix Znet)++ | t/04-nativecall/20-concurrent.t:
Unfudge test for non-Windows platforms

Floppiness is not observed there.
13:36
Zoffix NeuralAnomaly, status
NeuralAnomaly Zoffix, [āœ”] Next release is today. Since last release, there are 38 new still-open tickets (0 unreviewed and 0 blockers) and 0 unreviewed commits. See perl6.fail/release/stats for details
Zoffix fires up a test release run on test repos 13:37
viki Hm, this test fails for me on Windows 10: github.com/perl6/roast/blob/master...ing.t#L401 14:18
never mind, works after running `chcp 65001` 14:19
.oO( wonder if the test should do that... )
MasterDuke idle curiosity while waiting for rakudo-j to build. is there anything in particular that causes the JVM build to be so much slower? sure, the JVM doesn't start up all that quickly, but stage parse takes 170s for me, which i assume isn't starting up the JVM multiple times 14:34
any low-hanging fruit i could look into? 14:35
psch MasterDuke: well, consider that parsing a Perl 6 or NQP class means generating a Java class during runtime 14:36
Zoffix hm... lots of stresstest failures on Windows 10.... Seems like at least some of them are likely due to my using the default command prompt :/ 14:37
psch MasterDuke: add to that the "normal" enhancing-the-grammar overhead on the Perl 6 level -- which is basically "create a new grammar class that inherits the old one and adds the new rule"...
Zoffix Ummmm
"Windows Firewell has blocked some features of this app: "moar.exe"" :( 14:38
psch MasterDuke: plus we rely on a lot on invokedynamic, which basically means "decide this dispatch at runtime", which kills the JVM JIT
Zoffix Wonder where it was connecting to (didn't even look before clsoing the windows :())_
psch well, not quite "kills", more like "makes you do the same work manually", which i don't think we do right now 14:39
that might be kind of LHF-y, actually
lemme confirm my suspicion first... :)
jnthn invokedynamic was introduced in order to *inform* the JIT about things
psch jnthn: right, but the JIT can only work on indy if it has the exact same set of arguments 14:40
MasterDuke psch: hmm, sounds like not much LHF
psch jnthn: otherwise the FALLBACK needs to cache different, oft-occuring paths itself
Zoffix Failures: gist.github.com/zoffixznet/1b703b9...583b77309e
psch jnthn: at least that's what my digging around indy and jvminterop around christmas 2014 turned up 14:41
jnthn psch: You can update the callsite with further conditions in the fallback, though, PIC-style. Think we do that somewhere or other.
It's been a while since I looked at it, mind.
psch right, same here :)
i'm not aware of the meaning of "PIC-style" though 14:42
from the looks of it we're hanging invocation targets onto the CallSite, which seems sensible enough 14:43
MasterDuke while we're discussing the JVM, how come its backtrace and --ll-exception output is different from moar in the first place? 14:45
psch do you have an example of how it's different?
MasterDuke perl6-m --ll-exception -e 'say 1/0' Attempt to divide 1 by zero using div at gen/moar/m-CORE.setting:24539 (/usr/share/perl6/runtime/CORE.setting.moarvm:throw) 14:46
perl6-j -e 'say 1/0' Attempt to divide 1 by zero using div in method floor at gen/jvm/CORE.setting line 53 14:47
using a moar from before my line directive support 14:48
psch so it's a line break difference? might be we print instead of println in ExceptionHandling.java
MasterDuke not that, moar mentions m-CORE.setting and its line number (which is what i changed) and the method in the runtime 14:49
psch oh
$ ./perl6-j -e'say 1/0' 14:50
Attempt to divide 1 by zero using div in method floor at gen/jvm/CORE.setting line 25499
is what i got locally?
MasterDuke jvm just mentions CORE.setting and its line number (which is picking up the line directive)
perl6-j --ll-exception -e 'say 1/0' Attempt to divide 1 by zero using div in throw (gen/jvm/CORE.setting:26881)
psch oh, yes. we don't add the backend prefix for r-j
Zoffix jnthn, you develop on Windows, right? And don't get spectest failures? I'm just wondering what do I do with all of my failures as far as release goes.
MasterDuke that's from a r-j before my line directive patch
psch that's somewhere in Makefile-JVM.in 14:51
i mean, we used to have m-CORE.setting directly in gen/
and there it makes sense to prefix it with the backend
...or maybe it was even in the repo root, i don't remember
i mean, gen/jvm/ vs gen/moar/ is probably pretty sensible 14:52
and, for conformitys sake, i'd want to say that moar should stop adding that prefix
although maybe r-js will add the prefix too and r-j is in the wrong... :) 14:53
jnthn Zoffix: Mostly on Linux these days
Zoffix: I used Rakudo on Windows (which is why I ran into and patched a pre-comp issue or two earlier on in the month)
There's long been a small number of spectest failures on Windows, though. 14:54
psch MasterDuke: i'm afraid i still don't recognize the actual problem, honestly
r: say 1/0
camelia rakudo-jvm 76b061: OUTPUTĀ«Attempt to divide 1 by zero using divā¤ in block <unit> at <tmp> line 1ā¤ā¤Actually thrown at:ā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
..rakudo-moar 25b0a5: OUTPUTĀ«Attempt to divide 1 by zero using divā¤ in block <unit> at <tmp> line 1ā¤ā¤Actually thrown at:ā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
Zoffix OK
psch those do look identical, though i suppose camelia has a reason for not merging them 14:55
maybe the differing gitref
Zoffix A lot of failures appear to be bogus test that expect too much precision out of floating point math
jnthn Zoffix: The power and num ones are long-standing for sure; iirc some oddity around NaN or Inf handling
MasterDuke moar mentions the source+line number *and* the runtime, jvm just mentions the source+line number
when you add --ll-exception
psch ahh, this bit "(/usr/share/perl6/runtime/CORE.setting.moarvm:throw)"? 14:56
MasterDuke yeah
psch hm, i'd have to look into what moar does there, actually
the place it should be added would still be ExceptionHandling.java i suppose
in nqp/src/vm/jvm/runtime/org/perl6/nqp/runtime iirc 14:57
jnthn iirc, it's just choosing to report both the high level language location from resolving annotations, and also the bytecode location
I'd not except --ll-exception to look the same across backends.
psch same here, actually
--ll-exception is the more-user-friendly version of NQP_VERBOSE_EXCEPTIONS 14:58
...in my mind
yes, the impl details are different
but it's more a "hand this to someone who knows more" kinda deal
and NQP_VERBOSE_EXCEPTION will always be different across backends, except if we bend backwards to a perilous degree 14:59
MasterDuke ok, that's not a big deal. however, i'm still trying to get jvm to report the filename in the line directive (it *is* getting the line number) instead of gen/jvm/CORE.setting
psch MasterDuke: filename of the file in src/core?
MasterDuke yep 15:00
psch MasterDuke: that's probably a really fun romp through nqp/src/vm/jvm/QAST/Compiler.nqp
Zoffix m: say 108556850480095346844596918561061146623193876257401611187638859870020063756287.base: 2
camelia rakudo-moar 25b0a5: OUTPUTĀ«1111000000000000111111111111111100000000000011111111111111110000000000001111111111111111000000000000111111111111111100000000000011111111111111110000000000001111111111111111000000000000111111111111111100000000000011111111111111110000000000000111111111111111ā¤Ā»
psch MasterDuke: that's where we generate the Java annotations, which, currently utilize the file we're actually compiling
MasterDuke yeah, that's where/what i'm playing around with now
psch MasterDuke: i imagine whatever your moar solution there was is probably also along the lines of "figure out which file this bit of code originally came from and find the right line" or something? 15:01
MasterDuke i.e., got it to pick up the line numbers in the directive so far
Zoffix Well, I'm gonna go ahead with the release. I see pick test is likely bogus. Some of the failing procasync tests say "The syntax of the command is incorrect."
And both master and 6.c-errata run clean on Linux
psch i guess you always parse the '## from file $some-core-file ' or whatever exactly gen-cat puts there and count the lines yourself..? 15:02
MasterDuke psch: pretty much. find the offset from the most recent line directive and report that
Zoffix Hm, gonna build a 2016.10 and see if the precomp test fails there too
MasterDuke psch: github.com/perl6/nqp/commit/0538ca...bedbba80ec 15:03
Zoffix (it's the first time I'm testing the release on Windows too, so I dunno if any of this stuff is brand new)
psch MasterDuke: the QASTCompilerMAST.nqp diff shouldn't be to hard to replicate around QAST/Compiler.nqp:4195, from the looks of it 15:05
MasterDuke: it might depend on changes to JAST::Annotation though 15:06
MasterDuke psch: this is what i've done so far, which get the line number (and does parse the filename, but doesn't put it in the backtrace): gist.github.com/MasterDuke17/6a8be...b52ef52ab5
psch ah, right. there was a 53 there in your paste earlier
which seems awefully early for CORE.setting for div-by-zero 15:07
MasterDuke: org/perl6/nqp/jast2bc/JASTCompiler.java:290 or there abouts is where we put that into the bytecode
MasterDuke: mind, you still have to extend JAST::Annotation to actually know the file, which might mean diving a bit deeper into what exactly where doing at that spot in JASTCompiler 15:08
MasterDuke: the objectweb.asm docs are pretty good though, at asm.ow2.org/asm40/javadoc/user/inde...mmary.html 15:09
MasterDuke psch: ahh, this looks promising, many thanks!
psch ++MasterDuke 15:10
:)
Zoffix oh 2016.10 doesn't even build on Windows. 15:12
Oh well. Lesson to learn for Zoffix: test on Windows more frequently.
NeuralAnomaly, cut the release
NeuralAnomaly Zoffix, Will do! If you're feeling particularly naughty, you can watch me at perl6.fail/release/progress or go look at some cats icanhas.cheezburger.com/lolcats
Zoffix, ā™„ā™„ā™„ā™„ā™„ā™„ Prep done
Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  ABNORMAL EXIT!
Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  NQP: Bump MoarVM version
Zoffix :o
MasterDuke Zoffix: i think that's why stmuk_ didn't do a .msi for r* 15:13
Zoffix Oh, nqp is already bumped
NeuralAnomaly, status
NeuralAnomaly Zoffix, [āœ”] Next release is today. Since last release, there are 38 new still-open tickets (0 unreviewed and 0 blockers) and 0 unreviewed commits. See perl6.fail/release/stats for details
Zoffix NeuralAnomaly, steps 15:14
NeuralAnomaly Zoffix, all pre nqp r post pre-r6 pre-blank-slate nqp-clone nqp-bump-vers nqp-build nqp-tar nqp-tar-build nqp-tag nqp-tar-sign nqp-tar-copy r-clone r-prep-ann r-bump-vers r-build r-p5 r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp
Zoffix NeuralAnomaly, run pre nqp r post pre-r6 pre-blank-slate nqp-clone nqp-build nqp-tar nqp-tar-build nqp-tag nqp-tar-sign nqp-tar-copy r-clone r-prep-ann r-bump-vers r-build r-p5 r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp
:/
NeuralAnomaly Zoffix, ā™„ā™„ā™„ā™„ā™„ā™„ Prep done
Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  NQP: Bump MoarVM version
Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  ABNORMAL EXIT!
Zoffix NeuralAnomaly, I told you not to bump it! 15:15
jnthn heh, is that 'cus I bumped it yesterday and so it tries to make an empty commit? :)
Zoffix Yeah :)
jnthn d'oh :)
Being helpful fail :)
Zoffix The bot just needs to check if it's already the right version
NeuralAnomaly, cut the release 15:19
NeuralAnomaly Zoffix, Will do! If you're feeling particularly naughty, you can watch me at perl6.fail/release/progress or go look at some cats icanhas.cheezburger.com/lolcats
Zoffix, ā™„ā™„ā™„ā™„ā™„ā™„ Prep done
dalek p: 7d9c105 | (Zoffix Znet)++ | VERSION:
bump VERSION to 2016.11
psch MasterDuke: the other end is probably something like teaching every CompilationUnit that the Annotations don't only contain a file number, but also a file name 15:21
MasterDuke: that probably involves changes to CodeRefAnnotations.java, as well as CompilationUnit.java to read those properly, although i've only skimmed this a bit right now 15:22
MasterDuke psch: cool, thanks for the pointers 15:25
Zoffix spots a failed JVM test :/ 15:27
t/nqp/059-nqpop.t ......................
NeuralAnomaly Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  ABNORMAL EXIT! 15:28
Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  NQP: build and test
Zoffix NeuralAnomaly, run nqp-build nqp-tar nqp-tar-build nqp-tag nqp-tar-sign nqp-tar-copy r-clone r-prep-ann r-bump-vers r-build r-p5 r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp
MasterDuke man, there's visitLineNumber, but no visitFileName, couldn't be that easy...
NeuralAnomaly Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  NQP: build and test
Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  ABNORMAL EXIT!
Zoffix /o\ 15:29
My awesome bot is showing all its cracks today :P
moritz time to awesomeify it some more!
psch MasterDuke: maybe the easier approach is to convert that saved value to a String (a la "$file:$line") and rewrite visitLineNumber into visitFileNameAndLineNumber?
Zoffix NeuralAnomaly, run nqp-clone nqp-build nqp-tar nqp-tar-build nqp-tag nqp-tar-sign nqp-tar-copy r-clone r-prep-ann r-bump-vers r-build r-p5 r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp 15:30
psch bbl & 15:31
MasterDuke well, visitLineNumber is part of the objectweb.asm API 15:33
jnthn docs.oracle.com/javase/specs/jvms/...vms-4.7.10 may be bad news 15:34
Zoffix man... that test failed again :/ 15:37
.... despite passing during my test run... on the same box :S 15:38
NeuralAnomaly Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  ABNORMAL EXIT! 15:39
Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  NQP: build and test
MasterDuke jnthn: hmm, is a new ClassFile created for each Perl 6 class?
jnthn No 15:40
One Java class file per compilation unit
MasterDuke and gen/jvm/CORE.setting is a single compilation unit? 15:41
jnthn Right
MasterDuke that does not seem promising 15:42
jnthn Seems they didn't consider/chose not to handle the "class file produced from multiple source files" case
Zoffix What do I do now? prove -r -v --exec ./nqp-j t/nqp/059-nqpop.t consistently fails test 278 with java.lang.NullPointerException: gist.github.com/zoffixznet/cfc354c...5b0d4e1d23 15:43
MasterDuke does anything here look helpful? stackoverflow.com/questions/530945/...ava-source
Zoffix psch, you around? 15:44
bartolin, or you...
j: use nqp; nqp::isnull(nqp::null_s()) 15:45
camelia ( no output )
Zoffix j: use nqp; nqp::isnull_s(nqp::null()) 15:46
camelia rakudo-jvm 76b061: OUTPUTĀ«java.lang.NullPointerExceptionā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
Zoffix there, this seems to be the offending test
timotimo wow 15:47
trying to be careful checking for nulls, but blows up anyway
bartolin Zoffix: hmm, that test was added about a month ago by pmurias++ 15:48
Zoffix Oh...
jnthn m: use nqp: nqp::isnull_s(nqp::null())
camelia rakudo-moar 25b0a5: OUTPUTĀ«===SORRY!=== Error while compiling <tmp>ā¤Confusedā¤at <tmp>:1ā¤------> use nqpā: nqp::isnull_s(nqp::null())ā¤Ā»
jnthn m: use nqp; nqp::isnull_s(nqp::null())
camelia rakudo-moar 25b0a5: OUTPUTĀ«Cannot unbox a type object (VMNull) to a str.ā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
Zoffix Didn't we fudge some test on nqp during last release
Zoffix reads the log
jnthn tbh I'd expect such a test to blow up in some way
On any backend 15:49
Zoffix Hm, no, those were different tests: github.com/perl6/nqp/commit/298e2281ec
The full test is ok(!nqp::isnull(nqp::null_s())); 15:50
nqp: ok(!nqp::isnull(nqp::null_s()));
camelia nqp-moarvm: OUTPUTĀ«ok 1ā¤Ā»
Zoffix nqp-j: ok(!nqp::isnull(nqp::null_s()));
Well, I'll fudge it then
bartolin Zoffix++ 15:51
bartolin didn't look at the nqp tests for a while
dalek p: f707728 | (Zoffix Znet)++ | t/nqp/059-nqpop.t:
Temporarily fudge tests for release

Tests fail on JVM: irclog.perlgeek.de/perl6-dev/2016-...i_13595918
15:53
Zoffix NeuralAnomaly, run nqp-clone nqp-build nqp-tar nqp-tar-build nqp-tag nqp-tar-sign nqp-tar-copy r-clone r-prep-ann r-bump-vers r-build r-p5 r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp 15:54
NeuralAnomaly Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  NQP: Clone repo 15:55
Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  ABNORMAL EXIT!
MasterDuke jnthn: "Compilers are permitted to define and emit class files containing new attributes in the attributes tables of class file structures." So i can't use the builtin SourceFile attribute, but could i create my own?
Zoffix NeuralAnomaly, run nqp-clone nqp-build nqp-tar nqp-tar-build nqp-tag nqp-tar-sign nqp-tar-copy r-clone r-prep-ann r-bump-vers r-build r-p5 r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp
NeuralAnomaly Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  NQP: Clone repo 15:56
Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  ABNORMAL EXIT!
Zoffix /o\
jnthn MasterDuke: Seems so; then just need to find a way to get hold of them. 15:57
Zoffix NeuralAnomaly, run nqp-clone nqp-build nqp-tar nqp-tar-build nqp-tag nqp-tar-sign nqp-tar-copy r-clone r-prep-ann r-bump-vers r-build r-p5 r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp 15:59
If today proves anything is we're pretty far from humanity-anihialating robot uprising 16:00
These things only look smart.
geekosaur that's usually the case 16:06
NeuralAnomaly Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  ABNORMAL EXIT! 16:08
Zoffix, ā˜ ā˜ ā˜ ā˜ ā˜ ā˜  NQP: build and test
Zoffix NeuralAnomaly, c'mon, bruh. 16:10
t/qast/01-qast.t is now failing 16:13
psch j: use nqp; use Test; ok(!nqp::isnull(nqp::null_s())); 16:16
camelia rakudo-jvm 76b061: OUTPUTĀ«ok 1 - ā¤Ā»
Zoffix java.lang.RuntimeException: Unimplemented case of read_ref 16:17
t/serialization/01-basic.t : gist.github.com/zoffixznet/bc547c4...4378d6b141
bartolin Zoffix: some of the failing tests in t/qast/01-qast.t were already a problem during the last release: github.com/perl6/nqp/commit/298e2281ec 16:18
Zoffix Yeah 16:19
m: say 1494 -975 16:20
camelia rakudo-moar 25b0a5: OUTPUTĀ«519ā¤Ā»
bartolin again, I'd prefer to skip all those for the release. (and I hope we can find a clean solution before the next release ...)
bartolin should really start to run nqp tests before his nightly spectest runs 16:22
dalek p: e8433d7 | (Zoffix Znet)++ | t/ (2 files):
Temporarily fudge tests for release

due to failures on JVM
Half of these are the same tests fudged on previous release 3e142894cf3f651abb77a7f9dd5fc738092e6e80
16:25
Zoffix NeuralAnomaly, run nqp-clone nqp-build nqp-tar nqp-tar-build nqp-tag nqp-tar-sign nqp-tar-copy r-clone r-prep-ann r-bump-vers r-build r-p5 r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp
NeuralAnomaly Zoffix, ā™„ā™„ā™„ā™„ā™„ā™„ nqp tests OK 16:35
stmuk_ is nom close to being tagged?
Zoffix stmuk_, yeah, why? 16:37
Why? 16:38
I can still abort if needed
stmuk_ no no
I was just going to try a windows build
Zoffix Oh, actually the tagging is at the end of the process... right now it's building nqp tar, then it'll run tests, and then it'll build rakudo, run its stress test on master; run stress test on 6.c-errata, make the tar, then repeat the build + stresstest on the tar and only then tag 16:40
and each stresstest takes about 154seconds
And you can watch progress here: perl6.fail/release/progress 16:42
psch ohh progress 16:43
NeuralAnomaly Zoffix, ā™„ā™„ā™„ā™„ā™„ā™„ nqp release tarball tests OK
stmuk_ will it echo your passphrase? :) 16:44
NeuralAnomaly Zoffix, ā™„ā™„ā™„ā™„ā™„ā™„ nqp release DONE
dalek kudo/nom: 64d66e5 | (Zoffix Znet)++ | docs/announce/2016.11.md:
Generate release announcement for 2016.11
kudo/nom: 4d0dd8c | (Zoffix Znet)++ | tools/build/NQP_REVISION:
[release] bump NQP revision
kudo/nom: f66f8be | (Zoffix Znet)++ | VERSION:
[release] bump VERSION to 2016.11
psch + echo -e '*****\n' # probably?
Zoffix stmuk_, I thought of that ;) 16:47
(and no, it's not that short :P)
lizmat++ kicking me off first spot in number of commits :) 16:48
psch Zoffix: what, your passphrase *isn't* five asterisksā€½ 16:50
NeuralAnomaly Zoffix, ā™„ā™„ā™„ā™„ā™„ā™„ Rakudo stresstest (master) OK 16:51
Zoffix psch, nope: github.com/zoffixznet/na/blob/mast...m6#L67-L70
NeuralAnomaly Zoffix, ā™„ā™„ā™„ā™„ā™„ā™„ Rakudo stresstest (6.c-errata) OK 16:53
moritz NeuralAnomaly++ 16:54
Zoffix++ 16:55
NeuralAnomaly Zoffix, ā™„ā™„ā™„ā™„ā™„ā™„ Rakudo release DONE 17:00
Zoffix, ā™„ā™„ā™„ā™„ā™„ā™„ Post: upload tarballs to rakudo.org
Zoffix, šŸŽŗšŸŽŗšŸŽŗšŸ“ÆšŸ“ÆšŸ“ÆšŸ“ÆšŸ“ÆšŸ“ÆšŸŒˆšŸŒˆšŸŒˆšŸ“¦šŸ“¦šŸ“¦
Zoffix, The release of **Rakudo #105 2016.11** has now been completed
Zoffix, šŸŽŗšŸŽŗšŸŽŗšŸ“ÆšŸ“ÆšŸ“ÆšŸ“ÆšŸ“ÆšŸ“ÆšŸŒˆšŸŒˆšŸŒˆšŸ“¦šŸ“¦šŸ“¦
NeuralAnomaly celebrates with an appropriate amount of fun
psch oh geez, *nothing* on my computer can display all of those graphemes
well, unless the second and third triad are boxes 17:01
Zoffix psch, i.imgur.com/WOUEW0a.png 17:02
psch ah, thanks Zoffix++
dalek ar: 3a52702 | (Steve Mynott)++ | modules/ (26 files):
bump star versions and modules
moritz my browser displays it correctly 17:03
dalek ar: 940e010 | (Steve Mynott)++ | tools/star/Makefile:
bump rakudo, nqp and moar
MasterDuke "java.lang.RuntimeException: org.perl6.nqp.runtime.UnwindException", wtf? 17:05
timotimo oh, what fun
psch ah, those
i think it's somewhere between "missed exception handler" and "missing continuation handler" 17:06
but, i'll readily admit, i never really understood what UnwindException actually do
MasterDuke this is the problem line: String file = (String) getattr_s(insn, jastAnnotation, "$!file", 0, tc);
psch well, consider you're generating bytecode there
so the UnwindException is thrown when that bytecode gets executed 17:07
MasterDuke oh, don't need the cast
well, still runtime exception
ahh, that 0 should have been a 1 17:09
timotimo that's the hint or something? 17:10
MasterDuke index i think
psch i'd honestly just not call hinted geattr
*getattr
timotimo well, you can just supply -1
psch should be the hint i think
timotimo rather, we should have a constant somewhere for NO_HINT
psch yeah well, but if you pass STable.NO_HINT you can also just call the not-hinted variant
timotimo fair enough
might as well go for it if you have a language with overloading 17:11
psch i mean, of course it doesn't really matter in the end
timotimo JVM be like "i'll inline this shit!"
psch my aesthetic is just more along the lines of "less arguments, especially if the ones you leave out mean 'this does not mean anything'" 17:12
timotimo right 17:13
no way to accidentally pass the wrong kind of sentinel "doesn't mean anything" value
MasterDuke what does the hint do?
psch MasterDuke: the jvm class behind a Perl 6 class has a number of fields called 'field_\d+' 17:14
timotimo it tells the getattr what slot to get the attribute from
psch MasterDuke: the hint is basically that \d
++
timotimo we pass that hint when we generate code that already knows what index the field we're getting things from has
dalek kudo/nom: d1ef82a | (Zoffix Znet)++ | src/core/Main.pm:
Bring back sub MAIN (Enum) support
MasterDuke "No such attribute '$!file' for this object", even though i added it to class JAST::Annotation
dalek ast: c00301e | (Zoffix Znet)++ | S06-other/main.t:
Revert "Temporarily fudge enum MAIN tests for release"

This reverts commit 47a9ec10c95776c99ca0f47474419a5278575121.
timotimo like when we have private attributes, for example
psch j: say <1/0i>.re
camelia rakudo-jvm 76b061: OUTPUTĀ«Method 're' not found for invocant of class 'Str'ā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
psch j: say val("1/0i").re
camelia rakudo-jvm 76b061: OUTPUTĀ«java.lang.ArithmeticException: BigInteger divide by zeroā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
psch j: say val("1+0i").re
camelia rakudo-jvm 76b061: OUTPUTĀ«1ā¤Ā»
psch j: say val("1+0i").im
camelia rakudo-jvm 76b061: OUTPUTĀ«0ā¤Ā»
psch huh, apparently that got fixed
that was a nice way to show how hints can mess things up 17:16
not sure who fixed it either :P
dalek p: dd51b3c | (Zoffix Znet)++ | t/ (2 files):
Revert "Temporarily fudge tests for release"

This reverts commit e8433d7bac541582fe6aae1aa3fc5fb3bce9959f.
p: 8e22038 | (Zoffix Znet)++ | t/nqp/059-nqpop.t:
Revert "Temporarily fudge tests for release"

This reverts commit f707728e61789905037c4cc0b9c468afeebdeb9c.
psch aha, apparently it was me
with a terrible bandaid fix too 17:17
...i wonder how that keeps happening
dalek kudo/nom: 9071142 | (Zoffix Znet)++ | docs/release_guide.pod:
2016.11 is now in the past
psch well, FSDO "terrible"
as long as P6Opaque is the only REPR that does MI we're good
timotimo "nichts hƤlt lƤnger als ein provisorium"
Zoffix Aaaannndd... I'm done for today 17:18
Zoffix leaves to do drugs and play video games
psch timotimo: well, the fix isn't *that* bad. i mean, the issue was that we generated wrong hints for some multiple-inheritance classes' attributes
timotimo: so obviously the solution is to not listen to the hint when we have a MI-class 17:19
it is somewhat bandaid-y, but i don't think we'll ever heavily rely on non-P6Opaque with MI
and i also don't think the performance impact is immediately concerning, because honestly, hinting doesn't really buy us that much in the first place 17:20
sure it's nice to have, but still
timotimo well, yeah. moar does it the exact same way, though :)
psch the general invocation and codegen overhead is so much higher...
timotimo also, the implementation of getattr is per-REPR
oh, wait
that's what you meant by that "only repr" comment 17:21
psch yeah, i think jnthn++ pointed me at the corresponding moar bit
dalek ar: f7b277a | (Steve Mynott)++ | docs/announce/2016.11.md:
draft announce
17:26
ar: 83e974d | (Steve Mynott)++ | / (2 files):
more version bumps
MasterDuke any idea why i'm getting that "no such attribute"? 17:27
psch MasterDuke: you might have to rebootstrap i think?
MasterDuke ah, how do i do that? 17:28
timotimo make j-bootstrap-files
i think
psch MasterDuke: don't use the new thing you added, run 'make j-bootstrap-files' to get a stage0 that knows them
timotimo right
psch MasterDuke: *then* use the things you added
timotimo the "don't use" part is important
psch MasterDuke: also commit the new bootstrap files first
MasterDuke cool, thanks 17:29
dalek ar: 86105ef | (Steve Mynott)++ | modules/perl6-pod-to-bigpage:
revert the perl6-pod-to-bigpage bump to prevent star being dependent on OpenSSL
17:51
ar: d16ef6f | (Steve Mynott)++ | docs/announce/2016.11.md:
markdown fixes
17:55
viki "json_fast: Incredible speed gains" I like that changelog entry :) 18:01
stmuk_ the main compiler changelog might look more familiar to you :) 18:04
I'm seeing a *lot* more missing test files now 18:12
timotimo incredible? not so sure about that ... 18:14
stmuk_ ah the second line of that commit says "seriously. i don't understand it. at all. 18:15
github.com/timo/json_fast/commit/d...5a0e21ca2e 18:16
timotimo oh
stmuk_ you wrote it!
timotimo yes, it's incredible because i can't believe it works
it's not an incredible amount
dalek ar: 56e67bd | (Steve Mynott)++ | docs/announce/2016.11.md:
different sort of incredible so delete that word, also remove pod-to-bigpage
18:19
viki stmuk_: well, they aren't really "missing". It's just we added a lot more new files on top of 6.c
stmuk_ viki: I was wondering whether star should patch spectest.data to remove 6.d tests 18:23
people are more likely to notice now
or wait for a proper upstream solution 18:25
viki I can look into fixing it for next Rakudo release. 18:33
stmuk_ viki++ 18:55
jnthn Why would we need to patch spectest.data for star? 18:58
viki jnthn: not just star, but regular releases too. The ton of "missing test file" messages is confusing to the uninitiated, making them think something is wrong. 19:01
jnthn: and to clarify: when testing 6.c-errata, it shows a ton of "missing test file" messages due to spectest.data referencing test files added since 6.c-errata, which 6.c-errata does not have
jnthn OK, so my next question is "why do we get those"? :)
Ah
It's probably less hassle to make the target to test 6.c-errata give the harness a flag that tells it to be quiet about missing things 19:02
dalek ar: c48de65 | (Steve Mynott)++ | modules/perl6-lwp-simple:
bump perl6-lwp-simple for test fix
19:08
kudo/nom: 1e10ce3 | bartolin++ | src/core/Failure.pm:
Remove no longer required special case for r-j

I was unable to find out why this special case was needed back in 2015. A spectest run looked good on rakudo-j (no new failures).
19:11
kudo/nom: 84ef8f3 | (Zoffix Znet)++ | src/core/Failure.pm:
Merge pull request #921 from usev6/patch-3

Remove no longer required special case for r-j
MasterDuke has anyone used MethodVisitor.visitAttribute() before? 19:14
"m.visitAttribute(sfa);" is giving me an NPE, even though both m and sfa are not null
.seen arnsholt 19:15
yoleaux2 I saw arnsholt 18 Nov 2016 14:51Z in #perl6-dev: <arnsholt> Definitely looks like you're getting Nums in there somewhere
psch MasterDuke: are you observing the order? 19:21
asm.ow2.org/asm40/javadoc/user/org/...sitor.html
at the very start of the docs there
MasterDuke yeah, i'm visitAttribute()ing first
psch are you visitAnnotationDefault before that is what i mean :S 19:22
+()ing # probably
i suppose we're using some of that functionality already, so i kinda wanna guess "yes" anyway :/ 19:23
MasterDuke i didn't add visitAnnotationDefault
psch hm, we don't have that either, actually 19:24
so maybe that helps, i'd guess it sets up the annotation in the first place, and you get the NPE because there's not annotation default that could contain your attribute..?
MasterDuke adding 'AnnotationVisitor av = m.visitAnnotationDefault(); av.visit("name", method.crName);' didn't help 19:28
psch tbf, it was the first thing that leaped at me from the docs page 19:29
i haven't worked with the annotation part of the MethodVisitor at all, honestly
maybe the book they wrote has some insight
download.forge.objectweb.org/asm/asm4-guide.pdf
MasterDuke ooh, looks like that might have some useful info 19:32
ah ha! needed to override read() and write() for my extended Attribute 19:45
psch++
(now, where should those attributes get read and how?) 19:48
dalek ar: 095b9f3 | (Steve Mynott)++ | / (3 files):
Patch a url in one module test
20:02
MasterDuke and i just confirmed the attribute(s?) shows up in r-j's CORE.setting.class with the names from the line directives 20:05
RabidGravy At line 794 in World.nqp there is a "# XXX Can give deprecation warning in the future, remove ... before 6.0.0." can that happen 20:23
psch isn't sure of the *current* EXPORT supersede syntax, never mind the legacy one... 20:25
apparently it's a stash that has some set of keys
and if one of the keys is lowercase, it's assume to mean supersede, but unchecked 20:26
m: gist.github.com/anonymous/63c3c581...e0fe36f0c2 20:29
camelia ( no output )
psch okay, that apparently already doesn't work
from perl6advent.wordpress.com/2011/12/...y-and-how/
sooo, yeah, someone who actually knows that EXPORT stuff should probably look at it :)
stmuk_ odd p6-Template-Mustache no longer passes tests on windows 20:44
at least using mingw
MasterDuke the API docs for StackTraceElement.getFileName() say this "... Generally, this corresponds to the SourceFile attribute of the relevant class file (as per The Java Virtual Machine Specification, Section 4.7.7). ..." 20:48
so looks like i need to getClass() or getMethod() instead and find my custom Attribute
stmuk_ neither does Linenoise install on windows/mingw 21:16
ah it does if I revert back before Native::Resources removal 21:20
dalek ar: 4b2ad3c | (Steve Mynott)++ | modules/Linenoise:
revert Linenoise submodule before Native::Resources removal to fix windows
21:46
MasterDuke i added some debugging lines to backtrace() in .../ExceptionHandling.java where i print out the filename, class name, and method name of each StackTraceElement 22:38
for a filename of "gen/jvm/CORE.setting" i get class names like "E9A5E7E19C54FEEABE867F852164187D6DF7E6E5" and method names like "qb_2389" 22:39
i'm trying to create a ClassReader to get to my custom Attribute, but it doesn't like those class names 22:40
any suggestions for how to get a usable class name? 22:46
oh, that's the name of the giant CORE.setting class 22:52
dalek kudo/nom: e9487d6 | lizmat++ | src/core/ (4 files):
Allow *-N on .pick/pickpairs/grab/grabpairs

Indicating you want all but N from the set. This makes sense, as you can specify * to indicate all from the set.
22:57
lizmat .tell jnthn was there a reason why native shaped str arrays are not implemented? or was it an oversight that they aren't? 23:58
yoleaux2 lizmat: I'll pass your message to jnthn.
lizmat good night, #perl6-dev!