Zoffix What's dispatch:<var> used for? 00:02
Ah $foo."meth-name" 00:04
Oh wait no, misread the regex.
jnthn $foo.$var() 00:05
Zoffix Ah
jnthn Time for me to go rest. And off for vacation tomorrow. Good luck with the releases; back on Wednesday. :-)
o/
Zoffix Right, that's what handles $foo.&meow
\o
AlexDaniel` samcv: that got me thinking. Whateverable also builds moarvm for every commit, and every build (compressed) is ≈3.5 MB 03:46
say 3.5 × 600
evalable6 2100
AlexDaniel` that's assuming that all intermediate commits are buildable :)
and all that what for… 03:47
bloatable6: releases
bloatable6 AlexDaniel`, No! It wasn't me! It was the one-armed man! Backtrace: gist.github.com/ccceeb85ab5de83fa8...2d44d01a0a
AlexDaniel` :S
oh I see 03:53
bloatable6: releases 04:07
bloatable6 AlexDaniel`, gist.github.com/68c51dabe931605592...5f07819a20
samcv AlexDaniel`, thinking about what? is bloatable6 what does the moar releases? 04:11
AlexDaniel` thinking that this merge of 600 commits had a drastic impact on the tiny ssd my server has :) 04:12
AlexDaniel` looks. “6.7G free”
samcv heh 04:13
i'm sure!
AlexDaniel` never thought that this kind of stuff can work: 04:19
c: 2017.09..HEAD say 42
boom 04:20
c: 2017.09..HEAD say 42
committable6 AlexDaniel`, gist.github.com/ce1b313d45b37e5744...f4903ff0db 04:23
samcv nice 04:29
AlexDaniel` samcv: fwiw I'm still looking at some modules. I'll let you know when it starts looking safe from my side 04:51
samcv awesome, thanks!
AlexDaniel` ok so Test::Scheduler no longer passes its tests 05:34
created this issue for now: github.com/jnthn/p6-test-scheduler/issues/3 05:39
samcv thanks AlexDaniel`
AlexDaniel` OO::Actors is failing too github.com/jnthn/oo-actors/issues/6 05:50
Geth rakudo/nom: 2262cc477c | usev6++ | src/core/Match.pm
Fix JVM build (re-add removed sigil to 'match')

The sigils have been removed (unintentionally, for sure) with commit dd880cad42. Because this code was part of a '#? if !moar' branch, it didn't explode while building Moar.
06:39
AlexDaniel` good catch 06:59
Geth rakudo/nom: 7277aa5470 | (Aleks-Daniel Jakimenko-Aleksejev)++ | tools/build/NQP_REVISION
Pre-release NQP_REVISION bump
07:10
rakudo/nom: version bump brought these changes: github.com/perl6/nqp/compare/2017....g4494e7049
d10d697748 | (Aleks-Daniel Jakimenko-Aleksejev)++ | docs/ChangeLog

Deliberately not logged: 7cf5ce7e fece49cf 312cac78 9a137598 ba718f45
nqp: 4494e70496 | (Aleks-Daniel Jakimenko-Aleksejev)++ | tools/build/MOAR_REVISION
Pre-release MOAR_REVISION bump
¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...-g19523568
AlexDaniel` hm, how does Inline::Perl decide what libperl.so version it wants? 07:48
samcv you can install multiple perl's like that?
and not in separate directories? 07:49
AlexDaniel` well, problem is that releasable fails on Inline::Perl5 tests 07:50
it wants libperl.so.5.20 and I don't have it
now I'm wondering why it was working in the past
samcv because you had 5.20 before?
AlexDaniel` I guess I really had v5.20 before, yeah
samcv idk i have /usr/lib64/libperl.so.5.24 /usr/lib64/libperl.so.5.24.1 and /usr/lib64/libperl.so 07:51
AlexDaniel` so how do I use it with, say, v5.24?
samcv well it loads the module for me. idk i have a Gentoo based os
AlexDaniel` ooh 07:54
I see something
samcv my constant $p5helper = %?RESOURCES<libraries/p5helper>.Str;
AlexDaniel` alright, I filed this ticket: RT #132343 08:38
synopsebot RT#132343 [new]: rt.perl.org/Ticket/Display.html?id=132343 [REGRESSION] better-sched and other async improvement ecosystem fallout
AlexDaniel` if somebody has more information on these issues please let me know
AlexDaniel` lunch &
samcv night & 08:42
AlexDaniel` good night o/ 08:49
[Tux] This is Rakudo version 2017.09-487-gd10d69774 built on MoarVM version 2017.09.1-614-g19523568 09:23
csv-ip5xs 1.214 - 1.241
test 12.255 - 12.594
test-t 3.283 - 3.388
csv-parser 12.696 - 13.172
lizmat Files=1229, Tests=75766, 316 wallclock secs (14.77 usr 5.27 sys + 2181.17 cusr 213.29 csys = 2414.50 CPU) 09:33
lizmat doesn't see any commits that would explain the drop in test-t timing 09:34
AlexDaniel` github.com/rakudo/rakudo/compare/2...gd10d69774 09:44
that's for the 3.101→3.283 change
Geth: ver github.com/rakudo/rakudo/commit/67...95b30fe9bb 09:46
Geth AlexDaniel`, version bump brought in these changes: github.com/perl6/nqp/compare/2017....6-g252fd89
AlexDaniel` Geth: ver github.com/perl6/nqp/commit/252fd8...405d029e89
Geth AlexDaniel`, version bump brought in these changes: github.com/MoarVM/MoarVM/compare/2...9-g1a9be0a
lizmat AlexDaniel`: my personal benchmarks show no detoriation 10:01
gfldex m: my @a = (^10).list; @a = @a.hyper(:batch(5)).map({ LAST say "last"; „∅“ }); 10:03
camelia ( no output )
gfldex are LAST meant to be executed at the end of a batch?
I would need that or something similar to provide progress indication when working with large lists. 10:04
jnthn: please take note of ^^^ 10:06
AlexDaniel` gfldex: file a ticket if you want that to be noted
gfldex RT doesn't take my e-mails anymore
gfldex tests if RT likes him again 10:08
AlexDaniel` gfldex: ok, open a ticket here then: github.com/rakudo/rakudo/issues
gfldex AlexDaniel`: I will if my RT test fails. 10:09
AlexDaniel` buggable: tags 10:12
buggable AlexDaniel`, Total: 1628; 6.D: 2; 9999: 9; @LARRY: 28; ANNOYING: 8; BOGUSTEST: 1; BOOTSTRAP: 4; BUG: 587; BUILD: 12; CONC: 39; DOCS: 1; EXOTICTEST: 5; FLAP: 1; GLR: 3; IO: 19; JVM: 48; LHF: 7; LTA: 178; MATH: 6; META: 2; MOAR: 2; MOLD: 233; NATIVECALL: 21; NYI: 58; OO: 13; OPTIMIZER: 8; OSX: 2; PARSER: 5; PERF: 27; POD: 19; PRECOMP: 15; REGEX: 46;
AlexDaniel`, REGRESSION: 38; REPL: 6; RFC: 61; RT: 2; SECURITY: 2; SEGV: 30; SINK: 1; SITE: 1; SPESH: 1; STAR: 7; TESTCOMMITTED: 12; TESTNEEDED: 38; TODO: 13; UNI: 27; UNTAGGED: 283; WEIRD: 2; WINDOWS: 4; See fail.rakudo.party/ for details
AlexDaniel` gfldex: sorry for the issues with RT. I feel your pain. 10:16
and yes, I did open rakudo/issues. My justification is here: github.com/rakudo/rakudo/issues/11...-338466171 10:24
Geth roast/6.c-errata: a63f653374 | (Jonathan Worthington)++ (committed by Aleks-Daniel Jakimenko-Aleksejev) | S02-types/WHICH.t
HyperWorkBuffer is going away

This was the only mention of it in the whole test suite. It was an implementation details of the previous hyper/race mechanism, which is being replaced.
10:30
AlexDaniel` lizmat: hello. It seems that this commit should be cherry-picked to 6.c-errata github.com/perl6/roast/commit/be53...b01b797L23 10:34
lizmat: but I don't really understand what's going on there, any chance for a more verbose explanation? :) 10:35
also, “valuek” 10:36
lizmat AlexDaniel`: first of all, I don't think these tests have a place in roast as they're really rakudo specific
having said that, the change I made in the BUILDALL autogen, is that if the default value of an attribute is a compile-time value
(such as "foo" or 42) 10:37
it will store that value in the Attribute.build_closure slot
which before only contained something Callable (even if the value was compile-time only)
the $*FOO // 42 forced the initializer to be a code closure (like before) 10:38
so that the test didn't need change 10:39
in the other case, I did *not* change the default value, so now we need to check for that value, rather than being a callable thing
does that make sense to you?
the alternative would have been to add an attribute to the Attribute class 10:40
(and yes, it's turtles all the way down :-)
AlexDaniel` yeah, it makes sense, although I don't see the resolution for the issue yet :) 10:41
I mean, do I just cherry-pick it, or do we move it to rakudo and say that these test shouldn't've been in roast at all? 10:42
greppable6: \.build
greppable6 AlexDaniel`, Sorry, can't do that
AlexDaniel` attempts to fix greppable6 10:43
LOL 10:45
that's a blocker I think 10:46
≠ op is broken
I should really pay more attention to whateverables.
lizmat I would say, cherry pick for now 10:47
AlexDaniel` m: say (0 != 0 | 1 or 0 != 0) 10:48
camelia False
AlexDaniel` m: say (0 ≠ 0 | 1 or 0 ≠ 0)
camelia True
AlexDaniel` my problem with using whateverable as a reliable canary is that we won't release anything for months if we wait for rakudo issues to be fixed so that the bots work again :) 10:49
for example, right now some test files get stuck in the middle. No idea what's going on, I can't bisect bisectable unfortunately. 10:50
bisect: say (0 ≠ 0 | 1 or 0 ≠ 0) 10:51
bisectable6 AlexDaniel`, Bisecting by exit code (old=2015.12 new=d10d697). Old exit code: 1
AlexDaniel`, bisect log: gist.github.com/a54ba13b52b3353237...7b9787ee89
AlexDaniel`, (2017-03-05) github.com/rakudo/rakudo/commit/d2...45de7a6703
AlexDaniel` oops that's a wrong range
bisect: old=2017.09 say (0 ≠ 0 | 1 or 0 ≠ 0)
bisectable6 AlexDaniel`, Bisecting by output (old=2017.09 new=d10d697) because on both starting points the exit code is 0
AlexDaniel`, bisect log: gist.github.com/dd059148d35e3fd4d5...9c6498b5df
AlexDaniel`, (2017-09-25) github.com/rakudo/rakudo/commit/1a...26389ecf05
lizmat AlexDaniel`: I think it's known that some commits caused hangs in the test-suite, so why not in a particular piece of code?
so, you should give a commit only X resources, and kill it if it doesn't finish in time 10:52
no>
right ?
AlexDaniel` right, but bisectable hangs in its own code
IIRC when trying to fuzzy-match tag names and shas
bisectable6: old=DEAD say 42 10:53
bisectable6 AlexDaniel`, Cannot find revision “DEAD” (did you mean “HEAD”?)
AlexDaniel` like here ↑
my wild guess is that it is running out of handles, or something like that, and it could very well be my mistake 10:54
at the same time no handles are needed for fuzzy-matching things? I really don't know actually.
commit: 1af2a745fcd^,1af2a745fcd say (0 ≠ 0 | 1 or 0 ≠ 0) 10:55
committable6 AlexDaniel`, ¦1af2a745fcd^: «False» ¦1af2a74: «True»
AlexDaniel` commit: 2017.09,1af2a745fcd^,1af2a745fcd,HEAD say (0 ≠ 0 | 1 or 0 ≠ 0)
committable6 AlexDaniel`, ¦2017.09,1af2a745fcd^: «False» ¦1af2a74,HEAD(d10d697): «True»
AlexDaniel` okay: RT #132346 10:59
synopsebot RT#132346 [new]: rt.perl.org/Ticket/Display.html?id=132346 [REGRESSION] ≠ is not identical to != in some cases (0 ≠ 0 | 1 or 0 ≠ 0)
Zoffix AlexDaniel`: when are you releasing? 11:01
lizmat hmmm... I'm not sure we are in agreement on the behaviour of phasers in hypered blocks 11:02
AlexDaniel` Zoffix: not any time soon. samcv++ is resting and the moarvm changelog is not ready yet I think
lizmat I was thinking of outlawing them in this release while we figure out the semantics in the coming releases
AlexDaniel` Zoffix: it should also be decided if what we see in jnthn's modules is very wrong or just a little bit wrong
or maybe not wrong at all 11:03
Zoffix Which is what?
AlexDaniel` releasable6: status
releasable6 AlexDaniel`, Next release will happen when it's ready. 2 blockers. 485 out of 487 commits logged
AlexDaniel`, Details: gist.github.com/bab260db219b8aab65...84769207c1
AlexDaniel` RT #132343
synopsebot RT#132343 [new]: rt.perl.org/Ticket/Display.html?id=132343 [REGRESSION] better-sched and other async improvement ecosystem fallout
AlexDaniel` I did not have any time yet to golf these things 11:04
and just the fact that the tests are failing does not tell much
Geth roast/6.c-errata: b7252b0c7b | (Elizabeth Mattijsen)++ (committed by Aleks-Daniel Jakimenko-Aleksejev) | S12-introspection/attributes.t
Adapt tests to new possible Attribute.build values

Detailed explanation by lizmat++:
   First of all, I don't think these tests have a place in roast as
   they're really rakudo specific.
... (10 more lines)
11:09
AlexDaniel` lizmat: I hope you don't mind ↑ :)
lizmat AlexDaniel`: no pb :-)
AlexDaniel` greppable6: \.build
greppable6 AlexDaniel`, gist.github.com/5b1779b3c7d6c77327...4dd95e9890
Zoffix AlexDaniel`: "not for a while" is what? Like a week before jnthn++ can get back and fix the scheduler problesm if there are any? 11:10
Trying to figure out what I should push into nom and what into post-release branch
AlexDaniel` Zoffix: sorry for the inconvenience. It really depends, it's possible that some things may simply be reverted and jnthn++ will not be needed. We will see how it goes. Or maybe one of the mere mortals will find a way to fix it :) I think it's ok to push to post-release branch by default, I'll merge it after the release (and will take care of merge conflicts if there are any) 11:14
you can use committable with any branch
although it's a bit inconvenient that you'll have to specify the sha
for example 11:15
commit: 738908be4 say 42
see, that works great.
commit: 738908be4 say 42
committable6 AlexDaniel`, ¦738908b: «Cannot find this revision (did you mean “456c439”?)»
AlexDaniel` :(
Zoffix don't worry about it
AlexDaniel` Zoffix: the fix for ≠, if there is one, I'm hoping to see in master :) 11:18
Zoffix Working on it ATM 11:19
AlexDaniel` Zoffix++
commit: 738908be4d79f1472821cd363ec82a1ef0bc3f32 say 42 11:24
committable6 AlexDaniel`, ¦738908b: «Cannot find this revision (did you mean “50c24a8”?)»
AlexDaniel` bah. The build is right there, but committable6 is refusing to use it 11:25
AlexDaniel` will be away for ≈2 hours 11:41
Geth rakudo/nom: a042fd927c | (Elizabeth Mattijsen)++ | src/core/Rakudo/Internals/HyperRaceSharedImpl.pm
Make phasers in hyper/race throw an NYI

Which will allow us to shake out the semantics of phasers like LAST until the next release, without getting code out there that may depend on the current semantics (which will probably change)
12:03
lizmat afk for a few hours& 12:11
Zoffix The (0 ≠ 0 | 1 or 42) thing isn't a regression. It was implemented this way originally, until it was rewritten in a13bad9 and now 1af2a745fcd551 reverted that and it's broken again 12:28
Something with optimization with junctions
"Negated operators are special-cased when it comes to autothreading. $a !op $b is rewritten internally as !($a op $b)" 12:29
I'm guessing this isn't being done for ≠ which is what the bug's about
AlexDaniel I thought it was working in 2017.09 12:31
Zoffix No, there was a bug 12:38
m: $ = (1 ≠ 1|2 or 42) for ^100_000; say now - INIT now 12:53
camelia 1.509733
Zoffix m: $ = (1 != 1|2 or 42) for ^100_000; say now - INIT now
camelia 0.01753618
Zoffix m: say 1.509733/.0175
camelia 86.270457
Zoffix *sigh* 12:55
Gonna fix it with messy-but-easy way
the capture slipping doesn't get folded, which causes both the 86x slowage and apparently the semantic differences 12:56
or maybe semantic diff is still there for != case too when its args can't be folded 13:00
(nope, doesn't look like it) 13:01
timotimo oh, the problem with ≠ was in the junction optimizer? 14:13
Zoffix m: ["{.file}:{.line}", .signature].say for &infix:<!=>.candidates 14:14
camelia No such method 'file' for invocant of type 'Any'
in block <unit> at <tmp> line 1
Zoffix anotehr bug
timotimo: there's more than one problem. The fix for 86x perf drop doesn't handle all the cases either. 14:15
Gonna just duplicate the ≠, ≤, and ≥ candidates in source, for now. Don't fully know how to swap these to ASCII equivalents in the Grammar/Actions, since we also have to take care to swap it only when they'd resolve to core ops. And we could make a rule that ALL Texas operators are just in-source aliases for ASCII versions and swap all of them that way 14:17
AlexDaniel` back 14:41
Geth roast/6.c-errata: 57eada3912 | TimToady++ (committed by Aleks-Daniel Jakimenko-Aleksejev) | S32-exceptions/misc.t
overspecific test disallows eject in correct spot
14:51
roast/6.c-errata: f4c2ad4a50 | (Zoffix Znet)++ (committed by Aleks-Daniel Jakimenko-Aleksejev) | S32-io/pipe.t
Fix broken Proc piping test

  .close on a pipe returns the Proc and sunk Proc awaits the
process to finish. This causes proc1 in the test to finish off before the second proc had a chance to grab its pipe.
The only reason this test was passing in the past is due to a bug where only the .close on the last open handle actually returned a Proc
AlexDaniel` Zoffix: interestingly, your original commit message claimed “The test is not part of 6.c-errata.” :)
Zoffix AlexDaniel`: what test?
AlexDaniel` github.com/perl6/roast/commit/220e...e0726ade82
Zoffix I went by RT number 14:57
But in the past it was passing only 'cause of a bug where sunk proc didn't wait for anything
Man, the one thing I immediatelly notice with rakudo's Issues is I get notifications about them all the time. With RT, they only go to email and I read them a few times a week 15:00
AlexDaniel` Zoffix: what about github.com/settings/notifications ? 15:01
you can turn off web notifications there 15:02
Zoffix they're global settings
And also, you can't unmark notifications as read. On email I could. 15:04
I don't even get github email
Geth rakudo: usev6++ created pull request #1201:
[jvm] Fix p6bindattrinvres for native attributes
15:05
Zoffix ZofBot: it's different and I don't like things that are different! 15:12
ZofBot Zoffix, But above all they contain a bottomless reserve of cheap labour
Zoffix m: say 'x'␤␤ 15:15
camelia x
Zoffix m: say 'x'␤␤#x 15:16
camelia x
Zoffix m: use Test; is-deeply (0 ≠ 0 | 1 or 42), 42, 'Junctions with ≠ works like for negated ops'␤␤#x
camelia not ok 1 - Junctions with ≠ works like for negated ops
# Failed test 'Junctions with ≠ works like for negated ops'
# at <tmp> line 1
# expected: 42
# got: Bool::True
Zoffix Weird. In a file the missing semicolon is complained about
Geth rakudo/nom: 6ad06fad9f | (Zoffix Znet)++ | 11 files
Make ≤, ≥, and ≠ work like ASCII versions

Fixes RT#132346: rt.perl.org/Ticket/Display.html?id=132346 Alt Fix for RT#131626: rt.perl.org/Ticket/Display.html?id=131626
The previous fix did not catch all the cases even with core ops and did not (on purpose) resolve issues when user defined their own ... (6 more lines)
15:21
synopsebot RT#132346 [new]: rt.perl.org/Ticket/Display.html?id=132346 [REGRESSION] ≠ is not identical to != in some cases (0 ≠ 0 | 1 or 0 ≠ 0)
synopsebot RT#131626 [open]: rt.perl.org/Ticket/Display.html?id=131626 [PERFTEST][PERF] ≥ and ≤ are 36x slower than Texas version; ≠ is 15x slower
Geth roast: e039bab7c5 | (Zoffix Znet)++ | S03-operators/misc.t
Test junctions with ≠

RT#132346: rt.perl.org/Ticket/Display.html?id=132346 Rakudo fix: github.com/rakudo/rakudo/commit/6ad06fad9f
AlexDaniel` Zoffix++ 15:23
timotimo ah, it's because we special-case negated ops 15:31
oh nice 16:42
bartolin Zoffix: the other day I stumbled above this todo'ed test for RT #121940: github.com/perl6/roast/blob/e039ba...ay19.t#L21
synopsebot RT#121940 [resolved]: rt.perl.org/Ticket/Display.html?id=121940 [@LARRY] say True but False # should be 'True' (advent2010-day19)
timotimo m: loop (my $i = 0.0; $i < 100_000.0; $i += 0.1) { }; say now - INIT now
camelia 2.049006192
bartolin Zoffix: you closed the ticket, explaining that the current behaviour is correct. should we adjust the test? 16:43
timotimo the #3 most time consuming part of that was DON'T_DIVIDE_NUMBERS, by rewriting it with ?? !! and nqp::p6bindattrinvres i got it down to the #4 spot with a comfortable margin
that one alone had 552ms before and 242ms afterwards
bartolin timotimo++ :-) 16:44
timotimo maybe it's still something for after the release?
i'll stresstest it 16:45
Zoffix bartolin: yeah, should be changed, since Bool.Str relies on Bool.Bool's value, which is changed by the role
bartolin I'll do that, thanks 16:46
timotimo i'm now investigating whether giving Rat and FatRat their own numerator and denominator accessor methods is worth doing 17:01
nope. not worth anything 17:04
m: say 2813 / 3352 17:07
camelia 0.839200
timotimo that's from making DON'T_DIVIDE_NUMBERS faster
well, this is with profiler enabled
which might skew the results
Geth roast: eb3b95e9c4 | usev6++ | integration/advent2010-day19.t
Fix test for 'True but False'

cmp. RT #121940
17:23
synopsebot RT#121940 [resolved]: rt.perl.org/Ticket/Display.html?id=121940 [@LARRY] say True but False # should be 'True' (advent2010-day19)
bartolin Zoffix: ^^ I hope, that's matching your explanation 17:24
MasterDuke timotimo: think github.com/perl6/nqp/pull/376 is worthwhile? 17:29
timotimo hm, is it safe to remove the declaration? 17:31
also: weird! we have a hash in there, but it has an @ sigil
as default value i mean
nqp-m: say(nqp::elems(@*nosuchvar)) 17:32
camelia 0
timotimo nqp-m: nqp::push(@*nosuchvar, 1)
camelia ( no output )
MasterDuke i think the declaration is from before it got turned into a dyn var
timotimo OK, so there's still something that holds a declaration for it? 17:33
MasterDuke hm, not sure
timotimo in that case it's probably nice to have. we'll be calling add_comp_line_directive like 50 times?
MasterDuke but line numbers still work after thechange
timotimo that's what gives us proper line numbers for inside the core setting, right?
might want to have it after the release 17:34
MasterDuke yeah, saves 621 Ns in the dyn var log
sure, no reason it has to go in before
timotimo BBL 17:36
i wonder how to stress DIVIDE_NUMBERS a little bit to see if rewriting the final part just like DON'T_DIVIDE_NUMBERS makes a good difference here, too 18:12
ah, just add 1/30 over and over instead of 0.1 18:13
oh, haha 18:15
that gets the denominator to be "the same" right away
cool, this benchmark has DIVIDE_NUMBERS at the top 18:18
it does degenerate into num at some point, though, which is unfortunate for benchmarking reasons
using FatRat is probably more representative 18:19
yikes
this benchmark now has gcd as its most expensive operation m( 18:23
found one that spends 24% of its time in divide_numbers and only 3% in gcd 18:28
m: say 2.8 / 3 18:29
camelia 0.933333
Geth rakudo/microoptimize_dont_divide_numbers: 332e34d3b9 | (Timo Paulssen)++ | src/core/Rat.pm
microoptimize (DON'T_)DIVIDE_NUMBERS
18:44
timotimo passes stresstest and such just fine, but i'd still rather have it after the release
dogbert17 timotimo: does this tell you anything? gist.github.com/dogbert17/909a5661...93af9c9d4f 18:46
timotimo this is about joining threads? 18:47
dogbert17: could you get the perl6 ops compiled with debug symbols, too? 18:50
it's part of the rakudo build process
dogbert17 github.com/perl6/roast/blob/master...ead.t#L106
is it an option to rakudo's make ? 18:51
timotimo hm, i'm not sure
line 437 in my local makefile has it
$(M_PERL6_OPS_DLL): $(M_PERL6_OPS_SRC) $(M_PERL6_CONT_SRC)
this is the directive 18:52
dogbert17 hmm, moar was not compiled with --no-optimize
timotimo you'll want to have -g in the M_CC as well as the M_LD calls
actually, there's just a single capturelex call in that function
so i actually know which line it corresponds to 18:53
dogbert17 could it be a missing MVM_ROOT ?
timotimo actually ... MVM_frame_capturelex probably allocates, right? 18:55
if it does that, p6_code_obj might move?
it doesn't look like the local variable p6_code_obj is ever written to? 18:56
so could "just" assign reg 2's .o to reg 0's .o 18:57
can you try that?
dogbert17 which line
timotimo the last line 18:59
i wonder why it'd leave the original value of the target register untouched if the first check causes an early return?
dogbert17 I'm a moron. I've already reported this as github.com/MoarVM/MoarVM/issues/728. On the other hand the asan output is slightly different 19:00
timotimo hmpf. but what is 20 bytes into the tc? 19:16
on my machine locally 20 bytes into the tc is between two fields :\ 19:17
dogbert17 and if you pretend to be using a 32 bit system what then ? 19:28
timotimo oh!
hm, is that where the nursery_fromspace pointer lives or something? 19:30
could you print offsetof(MVMThreadContext, nursery_fromspace) in gdb or something?
dogbert17 let me try 19:36
(gdb) p offsetof(MVMThreadContext, nursery_fromspace) 19:39
$1 = (void **) 0x14
Geth roast: 2ac82855b6 | usev6++ | 2 files
[jvm] Fudge some newly added tests
19:40
timotimo well, 0x14 is 20, so that's what
so the thread is falling apart from under the code running on it 19:41
dogbert17 that sounds quite bad :) 19:42
is there some place where the tc itself needs to be rooted? 19:44
timotimo i'm not sure how exactly the thread gets ended 19:46
first i'll have dinner though :)
dogbert17 have a nice meal
Geth rakudo/nom: 3f595acfbd | (Elizabeth Mattijsen)++ | docs/ChangeLog
Add phaser in hyper/race NYI commit to ChangeLog
22:12
samcv is still working on mvm changelog :P 23:11
timotimo does that mean my latest change to make MVM_SPESH_LOG not segfault can make it into the release! :3 23:13
samcv yes 23:18
timotimo cool 23:19
that would have been quite the blunder
samcv maybe i should get brrt to write the changelog for his stuff... 23:22
hah
i mean since it's a new thing that's introduced. all the small commits are not really changes from last month
timotimo you think we need much more than one line about the new jit in there?
samcv since it's wholey new
timotimo exactly