melezhik is it possible to invoke method by symbol link? 00:45
sorry. meant to ask that in #raku 00:46
Geth nqp: c10e950ec5 | (Samantha McVey)++ | tools/templates/MOAR_REVISION
Bump MoarVM for Unicode 13.1

Changes: 2020.10..2020.10-20-g70e3608cb 2cac07c93 Regenerate unicode_db to fix minor comment difference 1eda12a0a Use `foo(void)` instead of just `foo()` for functions that take no arguments 004e4bc77 The probe for `pthread_setname_np` needs prototypes from <string.h> 608b90eb1 No need to MVMROOT `root` twice in `get_attribute`. ... (12 more lines)
rakudo: d3acd861a3 | (Samantha McVey)++ | t/09-moar/General_Category__extracted-DerivedGeneralCategory.t
Fudge unicode test in preparation for Unicode 13.1

Disabling this, just so all tests still pass after I bump NQP/MoarVM, but before the tests are updated.
rakudo: 23f67fee0e | (Samantha McVey)++ | tools/templates/NQP_REVISION
Bump NQP/MoarVM for Unicode 13.1

Changes: 2020.10..2020.10-63-gc10e950ec c10e950ec Bump MoarVM for Unicode 13.1 49bd5b2c1 fix some MD formatting 9f14b2b6c Add note about HLL to opsdocs 27ad522ff Documentation cleanup ... (59 more lines)
rakudo: bd803d5eda | thundergnat++ | 15 files
Update tests for Unicode 13.0, 13.1
rakudo: 42c4c1ed3e | (Samantha McVey)++ (committed using GitHub Web editor) | 15 files
Merge pull request #3942 from thundergnat/unicode13

Update tests for Unicode 13.0, 13.1
rakudo: c54193e817 | (Samantha McVey)++ | t/09-moar/General_Category__extracted-DerivedGeneralCategory.t
Revert "Fudge unicode test in preparation for Unicode 13.1"

This reverts commit d3acd861a3d371013fcd0786e260d002db699e8e.
roast/master: 4 commits pushed by thundergnat++, (Samantha McVey)++ 01:58
samcv ok. merging unicode 13 is complete. please ping me if anything bad happens :) but that should not happen 01:59
02:21 melezhik left 03:10 camelCaser left 03:15 camelCaser joined 04:49 travis-ci joined
travis-ci Rakudo build failed. Samantha McVey 'Revert "Fudge unicode test in preparation for Unicode 13.1" 04:49
04:49 travis-ci left 04:54 frost-lab joined 06:26 frost-lab left 07:25 frost-lab joined 08:01 sena_kun joined 08:08 Altai-man joined 08:11 sena_kun left
samcv ^ not sure why `t/02-rakudo/reproducible-builds.t` failed on two apple CI jobs. Does that happen sometimes? 08:18
08:25 domidumont joined
nine Unfortunately, yes. I'm pretty sure that there's at least one bug left but so far haven't been able to reproduce 08:32
09:13 MasterDuke joined
samcv is it a flapper? 09:29
lizmat samcv: t/09-moar/Line_Break__LineBreak.t inconsistently either reports test 1 as TODO and passing, or test 2 09:56
Geth rakudo: 78f164517c | (Elizabeth Mattijsen)++ | src/core.c/Distro.pm6
Make MacOS nick lookup hash a constant

So we don't have to do this at each startup. And possibly fixes non-reproducible build issues on MacOS.
rakudo: 2ad244b17f | (Elizabeth Mattijsen)++ | src/core.c/Distro.pm6
Be consistent in how we use my constant $foo =
nine samcv: yes 11:16
11:21 travis-ci joined
travis-ci Rakudo build passed. Elizabeth Mattijsen 'Make MacOS nick lookup hash a constant 11:21
11:21 travis-ci left
Geth rakudo: 0447b15e8a | (Elizabeth Mattijsen)++ | src/core.c/Match.pm6
Constantize empty list / hash in Match
rakudo: 8af0c463b6 | (Elizabeth Mattijsen)++ | src/core.c/Rakudo/Internals.pm6
Constantize encoding and clean parts lookup
[Tux] WHOT? \o/ 11:29
Rakudo v2020.10-37-g2ad244b17 (v6.d) on MoarVM 2020.10-20-g70e3608cb
csv-ip5xs0.792 - 0.796
csv-ip5xs-207.791 - 7.876
csv-parser24.621 - 25.320
csv-test-xs-200.384 - 0.390
test7.547 - 7.830
test-t1.773 - 1.878
test-t --race0.812 - 0.855
test-t-2030.545 - 31.404
test-t-20 --race9.383 - 9.937
Geth roast: 2277f397df | (Elizabeth Mattijsen)++ | 4 files
Unmark now passing Unicode tests as TODO
timotimo [Tux]: is that a new all-time low? 11:59
lizmat no, it's not afaik
MasterDuke i think 1.6 is the lowest 12:00
but if so, it was over 180 days ago 12:02
timotimo we're clawing back the wins we got from a performance regression that was necessary to get things right? 12:05
MasterDuke don't think anybody ever tried to figure out what happened 12:06
12:09 sena_kun joined 12:11 Altai-man left
samcv lizmat, will check that later 12:21
[Tux] timotimo, all time top5: 20190418 1.624❙20190425 1.623❙20190718 1.621❙20190429 1.605❙20190420 1.594❙ 12:23
12:24 travis-ci joined
travis-ci Rakudo build passed. Elizabeth Mattijsen 'Be consistent in how we use my constant $foo =' 12:24
12:24 travis-ci left
[Tux] hopes for "next" optimization (being goto instead of exception when possible) 12:26
MasterDuke has some trend lines 12:33
12:35 Kaiepi left 12:43 lucasb joined 12:51 patrickb joined
patrickb [Tux]: How hard would it be to validate the top5 by rerunning/retiming the exact version again? 12:52
[Tux] time-consuming, and I don't have time. 13:04
I however can tell someone else exactly what to do
and note that now in corona-times, I am actually working on this box, so the balance of processes is a bit different 13:05
Text::CSV has not changed a lot, so that part should be easy
does perl6 have a public accessible smoke box that I could use to setup something like that? 13:07
raku, sorry
lizmat tsk :-)
[Tux] ideal would be a tool that could reproduce the complete test environment for a certain situation, like nqp commit x, or MoarVM commit y 13:13
than find the closest commit in any of the related projects on or before that point in time and check those out for those 13:14
rebuild the whole kaboodle and run the timing
I think nothing really change in Text::CSV since 2018-03-30 that has any influence on these timings 13:20
[Coke] I wonder if we could get azure credits or something via TPF. 13:25
I will ask around.
timotimo getting the right rakudo versions should be possible with sharable6 13:32
though it will take commit hashes rather than date/time
which git can help you with by giving you all commits from one day, or something like that
AlexDaniel` well, there isn't really a technical reason not to run a benchmark agains every rakudo version 13:34
so the datetime logic is unnecessary 13:35
timotimo i guess that's fair
AlexDaniel` and when I say that I mean prebuilt deep rakudo-nqp-moar versions too, although this is not implemented in whateverable just yet :( 13:36
timotimo right; moarvm versions can at least be switched out without much trouble as long as there's compatibility 13:38
i.e. if you can build the moarvm with the nqp and rakudo version you have, you can drop in that version without running the build again
but if you change nqp, rakudo will have to be recompiled completely 13:39
so the amount of possible duplication would be immense if you wanted to pre-build everything
AlexDaniel` not talking about switching anything underneath existing installs 13:45
it's possible to just take 3 versions, build it, compress it… ??? Profit! 13:46
timotimo ah, that'll work too
AlexDaniel` nqp and moar bumps are surprisingly linear, it's not even n²
anyway, there's some help here: 13:47
lizmat notable6: weekly 13:48
notable6 lizmat, 2 notes:
sena_kun lizmat, not sure if Comma 2020.10 release was mentioned. 13:50
lizmat aha! tweet ? link ??
sena_kun lizmat, jnthn has asked to do one, I'll ping again. 13:51
s/asked/promised/ 13:52
lizmat notable6: weekly reset
notable6 lizmat, Moved existing notes to “weekly_2020-11-02T13:52:40Z”
14:07 evalable6 left, linkable6 left
Geth rakudo: ccc50f7ae6 | (Elizabeth Mattijsen)++ | src/core.c/Distro.pm6
Add support for MacOS 11.0 nick
14:09 evalable6 joined, linkable6 joined
[Coke] updates, lizmat+= 14:15
lizmat, no support for "Kodiak" in Rakudo? :)
lizmat nah... :-) 14:17
14:21 frost-lab left 14:28 patrickb left 14:39 Kaiepi joined
lizmat bisectable6: dd "\x1F62E\x200D\x1F4A8".chars 14:42
bisectable6 lizmat, Will bisect the whole range automagically because no endpoints were provided, hang tight
lizmat, Output on all releases:
lizmat, Bisecting by output (old=2017.06 new=2017.07) because on both starting points the exit code is 0
lizmat, bisect log: 14:43
lizmat, (2017-07-09)
lizmat, Output on all releases and bisected commits:
lizmat ok, so that's not a good test to test support for Emoji 13.1 14:44
[Coke] - some passing TODO tests on rakudo-latest. 14:53
(I don't normally run rakudo make test, no clue how long they've been passing) 14:54
looks like they rely on UnipropCheck.pm6 for TODO setting. 14:55
Geth rakudo: 7176cbc357 | Coke++ | 4 files
track new file name
[Coke] UnipropCheck.rakumod, that is.
.seen skaji 15:07
tellable6 [Coke], I saw skaji 2019-04-25T13:24:52Z in #perl6: <skaji> ugexe:
16:08 Altai-man joined 16:11 sena_kun left
nine Funny what bugs one comes across when investigating in-process-precompilation issues: 16:38
m: my $split = IO::Spec::Unix.splitdir("foo/bar"); IO::Spec::Unix.splitdir("wrong/value"); say $split;
camelia (wrong value)
nine splitdir always modifies and returns the same List object
MasterDuke doh 16:40
nine Introduced in 16:41
[Coke] good find.
nine It's not just that it's always the same List. Another question is, if it ought to be a List at all. Before that it returned a Seq, like we usually do 16:44
Geth roast: 49b5330059 | Coke++ | docs/unicode-generated-tests.asciidoc
Track change to filename in rakudo/rakudo
[Coke] nine: doesn't look like roast cares about the return type. 16:47
easy path: revert zoffix's commit, add a test for splitdir called on the class object, and verify that calling it twice works.
nine Funny pattern I've seen in much documentation (across multiple languages): it's usually quite well documented what you have to give to a routine, but much less so what a routine actually returns to you 16:48
[Coke] I'd keep it vague where possible (mention the role, not the specific class we're returning today) 16:49
(e.g the return type here should be Iterable)
lizmat and another Rakudo Weekly News hits the Net: 16:52
Geth rakudo: 613da571e8 | (Elizabeth Mattijsen)++ | 2 files
Fix an issue with $*TZ and long-running processes

This was spotted in production recently: before this commit, $*TZ was being set *once* on the first access to $*TZ. If a long running process would cross a DST change, it would report the wrong value after that change.
... (5 more lines)
nine I guess its a good sign that raku processes can run long enough for TZ changes to become an issue :)
lizmat yup :-) 17:00
well, actually DST changes...
TZ changes were pretty much covered already :-)
Geth rakudo: 78935ebdc0 | (Elizabeth Mattijsen)++ | src/core.c/DateTime.pm6
Some fixes and tweaks to $*TZ

  - should set last-minute-seen, duh! spotted by jnthn++
  - fix spello in variable name, spotted by MasterDuke++
rakudo: 7b2d9af7be | (Elizabeth Mattijsen)++ | src/core.c/PseudoStash.pm6
Pseudostash code doesn't need a sub

A Block should do. This doesn't break any spectests, but maybe these aren't tested enough. Please revert if this causes breakage.
18:04 domidumont left
lizmat afk for a few hours& 18:06
18:07 [Coke] left 18:28 HarmtH_ left 19:35 [Coke] joined, [Coke] left, [Coke] joined 19:45 domidumont joined
Geth rakudo: 151fd31083 | (Samantha McVey)++ | t/09-moar/UnipropCheck.rakumod
Make unicode property tests deterministic

In these tests, we test thousands of properties and then collect the results into two synthetic tests (otherwise the test would be much too slow and noisy).
Add a sort when we iterate the keys so that the tests don't end up running in different orders.
samcv lizmat, fixed the test order consistency
19:50 domidumont left 20:09 sena_kun joined 20:11 Altai-man left
lizmat samcv++ 20:34
[Coke] wonders why is showing 3 unicode letters as special when those are all the same font. 20:35
(some of them were converted to <img> tags for some reason. weird. 20:37
lizmat uninamically they all look like NEGATIVE SQUARED LATIN CAPITAL LETTERs
samcv [Coke], twitter is converting them to their own icon set. But some letters are not part of the icon set, so display with your font 20:38
[Coke] weird that their icon set has PAO but not LET 20:40
samcv i guess you could check the actual text doesn't contain any special presentation codepoints. otherwise it's probbaly the icon set 20:42
Geth rakudo: 89211e2eef | (Elizabeth Mattijsen)++ | src/core.c/DateTime.pm6
Use alternate test for DST change

As AlexDaniel pointed out, what if there is more than an hour between calls? Fixed by taking the absolute hour value since epoch and compare against that.
gfldex m: sub s($c where Int|Str) {}; s(½); 21:14
camelia Constraint type check failed in binding to parameter '$c'; expected anonymous constraint to be met but got Rat (0.5)
in sub s at <tmp> line 1
in block <unit> at <tmp> line 1
gfldex m: multi sub s($c where Int|Str) {}; s(½);
camelia Cannot resolve caller s(0.5); none of these signatures match:
($c where { ... })
in block <unit> at <tmp> line 1
gfldex Would it be possible to show the Junction instead of { ... } in such cases? 21:15
I'm asking because that would be an awesome error message. :)
MasterDuke looks like it might be happening here 21:22
and i guess it's because the constraint could be anything, it pretty much just punts 21:27
lizmat samcv: so now I get 2 TODO's passed in "make test", how to get rid of the TODO's passed ? 21:30
Geth rakudo: 5b3691817a | (Elizabeth Mattijsen)++ | src/core.c/Str.pm6
Make sure match processing can handle $/ being a Junction

As reported in R#4002. The problem was caused by some private methods not having a Mu constraint on the $/ container being passed, causing them to be called for each of the eigenstates in the Junction.
linkable6 R#4002 [open]: "Cannot modify an immutable Match" after matching junction against m//
AlexDaniel` timotimo: you aren't getting github notifications, are you? There's a problem-solving PR waiting for your review 21:36
I wonder if tony-o is nowadays active. Would've been nice to get it merged without waiting for the timeout 21:38
timotimo oh! yes, i really need to figure out the frickin github notifications 21:39
done 21:40
not figuring out the notifications, just that one thing 21:41
Geth nqp/NotQuiteParadise: d75f48d087 | (Elizabeth Mattijsen)++ | README.pod
Change NQP to Not Quite Paradise
nqp: lizmat++ created pull request #681:
Change NQP to Not Quite Paradise
rakudo: ede52fb02e | (Elizabeth Mattijsen)++ | src/core.c/DateTime.pm6
Check for actual changes in DST

nqp::decodelocaltime actually returns whether we're in DST or not, so we only need to check that.
lizmat and that concludes my hacking for today& 22:20
22:36 sena_kun left
Geth rakudo: vrurg++ created pull request #4003:
Don't inherit mixin class language revision
rakudo/rakudo_3998: 2991016a99 | (Vadim Belman)++ | src/Perl6/Metamodel/Mixins.nqp
Don't inherit mixin class language revision

Use the compiler-default. This fixes #3998 and makes the following code work:
   use v6.e.PREVIEW;
   role R { }; my $v = 0 but R;
Yet, `$v.^language-revision` will now return `e`, contrary to
  `0.^language-revision` returning `c`.
rakudo/rakudo_3998: eb454e99ae | (Vadim Belman)++ (committed using GitHub Web editor) | src/Perl6/Metamodel/Mixins.nqp
Merge pull request #4003 from vrurg/rakudo_3998

Don't inherit mixin class language revision
AlexDaniel` .seen tony-o 23:29
tellable6 AlexDaniel`, I saw tony-o 2020-09-02T15:45:05Z in #raku: <tony-o> [Coke]++