lizmat . 01:08
yoleaux 27 Nov 2017 20:41Z <DrForr> lizmat: theperlfisher.blogspot.cz/2017/11/t...hings.html I think the link got missed in the P6W.
lizmat .tell DrForr the link is in the Meanwhile on FaceBook section ? 01:10
yoleaux lizmat: I'll pass your message to DrForr.
Geth nqp: f22b0b2d0e | usev6++ | src/vm/jvm/QAST/Compiler.nqp
[jvm] Remove capture from cstack upon backtracking

This is a translation of similiar fixes for js (626e60c81f) and moarvm (d181c8b8d5). It fixes three currently failing test in S05-match/capturing-contexts.t (see also RT #116895 and RT #125285).		 05:57
synopsebot RT#116895 [open]: rt.perl.org/Ticket/Display.html?id=116895 [BUG] [JVM] LTM alternation captures the wrong stuff when backtracking in Rakudo
RT#125285 [resolved]: rt.perl.org/Ticket/Display.html?id=125285 Capturing separator (like /something % (.+?)/ ) spews out backtracking
TimToady m: enum Pows (<A B C> Z=> 1,2,4); say B # wonder if we have this in our bugs yet... 07:24
camelia 5===SORRY!5=== Error while compiling <tmp>
Undeclared name:
B used at line 1
[Tux] Rakudo version 2017.11-38-g1668b4f0b - MoarVM version 2017.11-2-g26ad7cc55
csv-ip5xs1.117 - 1.128
csv-ip5xs-2013.315 - 13.810
csv-parser12.697 - 12.923
csv-test-xs-200.445 - 0.447
test12.137 - 12.238
test-t3.260 - 3.271
test-t --race1.329 - 1.392
test-t-2059.335 - 59.452
test-t-20 --race20.508 - 20.638
 07:25
TimToady yeah, turns out to be a variant of #130041 07:40
Geth rakudo: d9021cf16e | TimToady++ | src/Perl6/Actions.nqp
enums from generated list should handle pairs

The enum actions code could handle literal pairs correctly but didn't handle generated pairs such as might be generated from X=>.
Fixes #130041		 07:44
TimToady m: enum Foo ( 'A' => 42, 'B', 'C', 'D' ); say +B 07:54
camelia 43
TimToady m: enum Foo ( 'A' => 42, 'B', 'C', 'D' ); say +A
camelia 42
TimToady m: enum Pows (<A B C> Z=> 1,2,4); say +B
camelia 2
masak TimToady++ 09:49
TimToady: you might have answered already, but... 10:09
quasi { my $x; {{{ #`[ $x not visible here ] }}} }
would you agree?
both these combinations are visible: macro->quasi, macro->unquote 10:10
but not quasi->unquote
(my argumentation is that quasi code is not "real"/"incorporated" enough for it to have real declarational side effects. this is in line with recent design decisions around quasis and BEGIN blocks.) 10:11
or, phrasing it from another angle, it kinda makes sense for the code in the unquote to be "unaware" of its surrounding quasi context. in a way it's *not* surrounded by it, because an unquote is... undoing the quasiquoting. 10:16
("it's only surrounded by it textually!")
jnthn masak: Yeah, it's kinda like how in `my $x = 101; Q:c/my $x = 42; { say $x }/` we expect 100, not 42 10:19
masak I'm deeply insulted at having quasiquotes being compared to text... but yes :) 10:20
jnthn Just with the oddity that the language being parsed inside of the quote construct happens to start in the MAIN grammar rather than the Q grammar
masak the thing that makes it tricky, of course is that something like `quasi { my $x = 42; $x }` still needs to work 10:21
jnthn An explanation that's good *and* insulting. Go me! :P
masak that is, the quasi *does* declare things
just not for the unquotes
in terms of nesting depth, the unquotes do a rare -1 instead of a +1, kinda 10:22
at least handwavingly-conceptually
jnthn It declares things, but inside of its own little world, from which the outer program looks kinda like a setting 10:23
masak a recent theme for me has been realizing (and re-realizing) that the quasiquote is a *template* 10:24
so while declarations can happen in there, they're not the "real thing" that will eventually happen in the injected code
and in a sense they can't be, because templates aren't fully built yet 10:25
the fine balance Perl 6 tries to walk here is that you're doing all these *clearly insane* things with scopes and program fragments... but you still wan't full lexical scoping to apply 10:26
want*
jnthn wan't :D
masak I blame macros for that typo, somehow
jnthn tries to guess the missing letter :)
masak a short progress report from the field: 007 is about to turn 3 years old. I've had some encouraging progress with quasis in the past few days/weeks. 10:28
jnthn But yeah, I think "it's just a quote construct that can interpolate, parsing a very interesting language inside" is a fairly nice way to think about it.
Wow, time flies
masak 007 is by far the toughest thing I've ever implemented. some things I have to work with on "good days" because they're just impossible otherwise
I'm still not 100% convinced I have the "right" model to bring back to Perl 6 -- I keep finding cases that don't work well 10:35
happy to discuss it with somebody, with lots of examples
though brain meltdown hazard applies, of course 10:36
tbrowder .tell AlexDaniel I added answers to your questions. 12:05
yoleaux tbrowder: I'll pass your message to AlexDaniel.
tbrowder .tell Zoffix the pod spec tests now require roast with PR #353 merge. 12:16
yoleaux tbrowder: I'll pass your message to Zoffix.
dogbert2 ZOFFLOP t/spec/MISC/bug-coverage-stress.t (test #5) 12:20
Geth rakudo: e7c0a64475 | (Elizabeth Mattijsen)++ | src/core/Rakudo/Iterator.pm
Fix the .toggle() case

  - if no checkers specified, $!done would not be properly initialized		 13:51
timotimo TimToady: will you spend a little time on the Z+= thing brian d foy reported (once a few months ago, again more recently)? 13:55
AlexDaniel . 15:22
yoleaux 12:05Z <tbrowder> AlexDaniel: I added answers to your questions.
Geth roast/master: 7 commits pushed by (Tom Browder)++, (Aleks-Daniel Jakimenko-Aleksejev)++ 16:31
TimToady timotimo: well, maybe, but no promises; in masakian terms, I haven't had any "good days" for the last couple weeks, due to a tooth extraction that got infected, resulting in a two-week long headache 16:51
masak: I'm on the mend, but not sure I'm up to brain-melting just yet 16:52
[Coke] TimToady: ugh, hope you feel better soon. 16:56
I'm doing my part with some acute tonsillitis, but I think I got the easy job. :) 16:57
AlexDaniel yay! SEGV 17:39
squashable6: next 17:41
squashable6 AlexDaniel, āš šŸ• Next SQUASHathon in 2 days and ā‰ˆ16 hours (2017-12-02 UTC-12āŒUTC+14). See github.com/rakudo/rakudo/wiki/Mont...Squash-Day
AlexDaniel hm, has anybody noticed any memory leaks lately? 17:43
just noticed that some inactive bots are eating a bit too much 17:44
timotimo TimToady: damn, that sucks. i wish you a swift and lasting recovery :) 17:47
jnthn TimToady: Hope you'll be feeling better soon; take care! 17:51
AlexDaniel jnthn: hello. I think I see a reproducible gc-related segfault here. I'm thinking, maybe you want to take a quick look? 18:27
I did catch it under valgrind but it took more than half an hour and mosh ate the output :( 18:28
but I happens under gdb also
it*
dunno if this helps: gist.github.com/AlexDaniel/5e0f9e0...adc4c2c4ff 18:30
o! maybe this is golfable 18:56
nvm then 18:57
timotimo AlexDaniel: does it happen with jit turned off? 18:58
AlexDaniel timotimo: yes 19:06
timotimo that ought to make the stack trace a bit more digestable
AlexDaniel timotimo: gist.github.com/AlexDaniel/5e0f9e0...adc4c2c4ff 19:08
(updated) 19:09
timotimo it's probably possible to get this to blow up earlier with GC_DEBUG set?
AlexDaniel what's GC_DEBUG? github.com/rakudo/rakudo/wiki/dev-env-vars 19:10
timotimo oh, it's actually a #define inside moar's c source
AlexDaniel wellā€¦ that'd be a bit harder 19:11
timotimo if you set it to 2 rather than just 1, it'll also make things a lot slower 19:12
AlexDaniel ok created this to track the problem: github.com/rakudo/rakudo/issues/1278 19:18
timotimo "GC-related" is a bit of a red herring; it's usually just that moar stumbles upon some prior corruption the moment GC runs, because GC basically touches everything 19:19
AlexDaniel timotimo: changed. 19:21
fwiw, feel free to edit any of my bug reports/comments on github
timotimo oh
moritz can somebody please try to create the category 2017 on perl6advent.wordpress.com/wp-admin/ ? 19:50
it fails for me
creating 2018 worked :/
oops, should've been in #perl6
jnthn AlexDaniel: Was away doing dinner when you wrote 19:54
Hm, I'm sure I've seen that stack trace before
And it was 'cus sc_idx was a bonkers value
AlexDaniel well let me know if you want to reproduce it on whateverable server
in theory you should also be able to do it on any machine, I'd just have to tar some input files for it 19:55
but I don't have any easy-to-follow steps yet :)
jnthn I'm probably a bit too tired for debugging such things today, alas :( 19:56
AlexDaniel sure. I'll update github.com/rakudo/rakudo/issues/1278 accordingly once I have more info
.tell lizmat I've done the bot. It tends to segfault a little bit (GH#1278), but if that happens just rerun the command again. github.com/perl6/whateverable/wiki/Reportable 21:02
yoleaux AlexDaniel: I'll pass your message to lizmat.
gfldex .tell lizmat basic html support is in. I had to cheat because Telemetry::report still resists overloading (and there are a few things not exported). 21:30
yoleaux gfldex: I'll pass your message to lizmat.
Geth roast: c39830a610 | usev6++ | S15-unicode-information/uniname.t
[jvm] Fudge two failing unicode tests
 23:12

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!