[Coke] spec tests only test for the former (where there is only a single trailing one in the units param) 00:00
s/units/scale/ 00:25
00:31 Xliff left
Geth roast: vrurg++ created pull request #695:
Add testing of mixins under v6.e
00:49
00:53 kawaii left 00:55 zostay left 00:56 kawaii joined, zostay joined
Geth rakudo: 6b87d3e52c | (Vadim Belman)++ | src/core.c/operators.pm6
Make auto-generated mixin role be of language revision 'c'

Fix `$a but $b` under `use v6.e` pragma.
01:12
rakudo: 9be9a5cfe4 | (Vadim Belman)++ (committed using GitHub Web editor) | src/core.c/operators.pm6
Merge pull request #3997 from vrurg/rakudo_3990

Make auto-generated mixin role be of language revision 'c'
roast: c0055b3383 | (Vadim Belman)++ | 4 files
Add testing of mixins under v6.e

Primarly to make sure that cases like `0 but True` are not broken. So,
  `mixin-6e.t` is basically a copy of `mixin-6c.t` with few cases removed
where 6.e-declared roles are mixed into core classes. These tests would require the roles to be declared externally, in a 6.c versioned module which would eventually wind down to the same testing we already do in mixin-6c.t.
  - Added a test for in-role declared constructor to be invoked upon mixin
   under 6.e
roast: 2b6c8e6c92 | (Vadim Belman)++ (committed using GitHub Web editor) | 4 files
Merge pull request #695 from vrurg/rakudo_3990

Add testing of mixins under v6.e
05:26 frost-lab joined 06:14 Kaiepi left 06:20 Kaiepi joined 08:21 jmerelo joined 08:23 Xliff joined
Xliff lizmat AlexDaniel: gist.github.com/Xliff/f4af67ed380d...975e4ed759 08:23
Also, let me know style changes, too! 08:24
08:37 domidumont joined 08:49 frost-lab left 08:51 frost-lab joined
[Tux] Rakudo v2020.10-25-g9be9a5cfe (v6.d) on MoarVM 2020.10-13-g1eda12a0a
csv-ip5xs0.836 - 0.958
csv-ip5xs-208.139 - 8.499
csv-parser27.202 - 27.595
csv-test-xs-200.385 - 0.388
test8.051 - 8.163
test-t1.923 - 1.938
test-t --race0.846 - 0.928
test-t-2033.471 - 34.084
test-t-20 --race10.331 - 10.490
08:54
nine got unstuck :) 09:33
So what happens is that during precompilation of Foo::Bar, that package gets added to Foo's Stash. Thus it remains available globally even when precomp is done. When loading the module, we run merge_globals which is supposed to hook up the stashes from the module to the globally available ones. 09:35
gfldex m: use v6.e.PREVIEW; role R {}; my $a = Int but R; 09:36
camelia Don't know setting RESTRICTED
at gen/moar/ModuleLoader.nqp:249 (/home/camelia/rakudo-m-inst-2/share/perl6/lib/Perl6/ModuleLoader.moarvm:previous_setting_name)
from gen/moar/ModuleLoader.nqp:254 (/home/camelia/rakudo-m-inst-2/share/perl6/lib/…
nine merge_globals differentiates the cases of the symbol not being known at all, the exact same object already being there, the source package being a stub and the target package being a stub and both being stubs.
gfldex m: use v6.e; role R {}; my $a = Int but R;
camelia Don't know setting RESTRICTED
at gen/moar/ModuleLoader.nqp:249 (/home/camelia/rakudo-m-inst-2/share/perl6/lib/Perl6/ModuleLoader.moarvm:previous_setting_name)
from gen/moar/ModuleLoader.nqp:254 (/home/camelia/rakudo-m-inst-2/share/perl6/lib/…
gfldex That is not the error message I wanted to show. :( 09:37
nine In my case though the source and the target are actual full blown packages. In that case we usually see the dreaded "P6M Merging GLOBAL symbols failed: duplicate definition of symbol" message. However I already disabled that message earlier as a workaround for similar situations.
A possible solution is to merge the stashes anyway
09:39 sena_kun joined
gfldex There error message I wanted to show can be found in #3998 . 09:45
nine A cleaner solution will probably be to remove the packages generated during precompilation so the usual merge rules work when loading the same module right after compilation 09:49
So, now down to 2 tests and 1 spec test failing. 09:59
nwc10 and how much coffee remaining? 10:00
nine Enough for the lock down :D
nwc10 now, I know the details here - the relevant shops are still open, so arguably there never wasn't a problem with coffee. 10:01
and also, the length of all these lock downs is "patches welcome"
so it's a bit hard to quantify
unless you have a private coffee plantation.
or you are like my boss and don't drink coffee. Which means that he can be confident that he has enough. (His wife, who does, I guess less so) 10:02
nine Oh I enjoy my morning coffee far too much to give up on it 10:05
nwc10 I think that he never did drink coffee.
I also think that I enjoy all my current "sins" far too much to give up on them.
jmerelo nwc10: in which country/state are you located? 10:24
tellable6 2020-10-30T23:53:41Z #raku <melezhik> jmerelo: another day you talked about testing some (potentially problematic) Raku modules during Hackaton. I suggest Rakudist as one of the solutions. So I just wonder if you've thought about that ...
2020-10-31T19:35:21Z #raku <[Coke]> jjmerelo - to avoid github ticket volleyball, do you agree that we should reject tickets regarding 'p6doc' executable since it is no longer in the repo? If so, I'm happy to do the ticket cleanup.
2020-10-31T19:35:58Z #raku <[Coke]> jjmerelo - I will also save the list of tickets for the new repo so whoever owns it can extract the bug info if they need it.
jmerelo .tell melezhik Hacktoberfest is over... We still need the testing, too. 10:25
tellable6 jmerelo, I'll pass your message to melezhik
nwc10 I live about 1km outside of the boundaries of Vienna.
so I'm in Lower Austria, but only just
nine is in Upper Austria. 10:26
and all the new restrictions are country-wide, so he's go the same rules as I have.
jmerelo .tell Coke please go ahead. The new repo is Raku/rakudoc, so shouldn't be a big problem to move.
tellable6 jmerelo, I'll pass your message to [Coke]
lizmat is in lower lands
jmerelo nwc10: for the time being, it's city-wide lockdown and curfew here in Granada. But all bets are off, since there were 2000 new cases in the province just last Friday (~1million people) 10:27
We still have 51 META.info-described modules in the ecosystem. Maybe we should have some deprecation cycle for that... 10:43
I seem to remember samcv was doing PRs with this? Maybe zoffix?
samcv jmerelo, hmm? 10:45
jmerelo hi, samcv ! Didn't you start an issue to substitute old META.infos? 10:46
samcv we're talking about the module META files?
jmerelo samcv: ah, no, only the license field... Anyway. Yes, we're talking about that.
samcv ah. yeah i did the license field thing
jmerelo samcv: here github.com/Raku/ecosystem/issues/324 Well, that's on top of some using META.info And that one is not finished either. 10:47
I just re-released PSGI and didn't realize about the license field. That should have been tested... 10:48
Geth rakudo: ee089234fc | (Elizabeth Mattijsen)++ | src/core.c/ShapedArray.pm6
Expose the Array::ShapedArray role

So that we can actually dispatch on that, like we can on the native version, the array::shapedarray role.
10:50
10:53 squashable6 left 10:54 squashable6 joined 11:03 jmerelo left
nine Finally...we're mere days away from that super important event that will have worldwide implications and make my life better! 11:09
Speaking about the avaiability of Zen 3 CPUs of course ;)
I wonder if I will see build times halved compared to what I get on the 1800X
gfldex I'm trying to buy a board to hold these. Should arrive December 2nd. 11:10
nine Oh....availability that bad or will it only be released by then? 11:11
gfldex I got one aleady but it was dead on arrival. It refused to go past POST. The replacement takes forever. I guess they are all sold out. 11:12
And Amazon tends to be good at selling things. 11:13
nine Right now Geizhals only lists a single passively cooled X570 board...and it's 800 Euros
gfldex They only spin the fan when pci 4.0 is used. 11:14
11:41 finsternis joined 11:42 sxmx left
lizmat m: my @a; dd @a[;] 11:49
camelia Unable to call postcircumfix @a[ Nil ] with a type object
Indexing requires a defined object
in block <unit> at <tmp> line 1
lizmat that feels like an LTA error, feels to me to be a special case of the zen slice 11:50
gfldex m: my @a; dd @a[]; 11:52
camelia Array @a = []
gfldex m: my @a; dd @a[Empty]; 11:53
camelia ()
gfldex lizmat: [;] looks to me like an empty SemiList. I can't see any use case where it would be useful to behave different from Empty. 11:55
m: my @b; my @a; dd @a[||@b]; 11:56
camelia ()
lizmat yeah, making it behave like Empty would be trivial
gfldex [;] feels equivalent of [@b] where @b ~~ Empty 11:57
lizmat afk for a few hours& 12:05
12:08 Altai-man joined 12:11 sena_kun left
samcv Is there any reason I can't bump nqp and raku when I merge in the unicode 13 changes? 12:12
Been a while since I did that ;)
Everything looks good. Reviewed the issues with the grapheme cluster break roast tests. Was a good opportunity to add new changes which fudge tests based on the exact text of the unicode test file, instead of line number. Since line numbers can change between versions, but the text of the test doesn't 12:13
so should make it easier in the future
bartolin lizmat: maybe you have a quick idea how to fix this? github.com/rakudo/rakudo/issues/3999 12:15
tellable6 2020-10-27T01:47:51Z #raku-dev <[Coke]> bartolin there are some jvm specific opcodes at gist.github.com/coke/f87ab505c6cb8...b5cb6a8990 like "attrhintfor" that could use some dos.
bartolin .tell [Coke] (undocumented opcodes specific for jvm) I can't promise anything (short of time, not familiar with the more unusual opcodes), but I've added that to my list 12:22
tellable6 bartolin, I'll pass your message to [Coke]
13:54 camelCaser left 13:55 camelCaser joined 13:59 frost-lab left
lizmat samcv: afaik, it's ok to bump 14:03
nwc10 but not to hug. even with masks :-/
lizmat
.oO( doing the bump, that's soooo 80s :-)
14:04
14:12 domidumont left 14:31 domidumont joined
Geth rakudo: 86a208e2ef | (Elizabeth Mattijsen)++ | lib/Test.rakumod
Use nqp::can to test for test-assertion

In some cases, specifically on the JVM, frames can be NQP code objects. So instead of using .? dispatch, we're just checking if it can do the
  'is-test-assertion' method. Which it only can if the code actually had
the "is test-assertion` trait applied with a true value.
This should fix #3999, Christian Bartolomäus++ for spotting.
14:53
linkable6 RAKUDO#3999 [closed]: github.com/rakudo/rakudo/issues/3999 [JVM] Many spectest failures after addition of is-test-assertion() to Test.rakumod
Geth rakudo: 35213f66a3 | (Elizabeth Mattijsen)++ | 2 files
Make sure %h{|| foo} and @a[|| bar] are handled correctly
15:12
nine Looks like 2 of the remaining 3 issues actually come from a workaround I did in MoarVM. Reverting that gives me 3 other errors of the same kind instead 15:14
I even added a "FIXME this is just a trial" comment back then...
Geth roast: 12096fdd54 | (Elizabeth Mattijsen)++ | 2 files
Add tests for %h{|| foo} and @a[|| bar] cases
roast: 9f069e6379 | (Elizabeth Mattijsen)++ | S24-testing/line-numbers.t
Make any errors a little bit clearer
15:16
nine In a sense I'm glad that this popped up again, as I don't remember what I thought at that time or how that's supposed to be a proper fix (which apparently it isn't anyway). I would have had a real hard time writing a commit message 15:17
lizmat yeah, sometimes things come back to haunt you 15:18
nine So all I have to do is now find a proper fix :)
lizmat earlier today I was looking at something in Main from 4.5 years ago, and was thinking: what the hell was I thinking then? 15:19
nwc10 well, obviously, then you were young and inexperienced :-) 15:21
nine aren't we all? 15:22
[Coke] . 15:40
tellable6 2020-11-01T10:26:35Z #raku-dev <jmerelo> Coke please go ahead. The new repo is Raku/rakudoc, so shouldn't be a big problem to move.
2020-11-01T12:22:52Z #raku-dev <bartolin> [Coke] (undocumented opcodes specific for jvm) I can't promise anything (short of time, not familiar with the more unusual opcodes), but I've added that to my list
16:09 sena_kun joined 16:10 Altai-man left
nine All tests successful. 16:22
I wasn't that far off at all! A small adjustment actually made my workaround...work :) So now all tests and all spectests pass (including stress test). make install succeeds but the modules it installs are not yet usable 16:23
Well some are, but not NativeCall 16:24
I'm not terribly surprised. E.g. install-core-dist.raku both uses CompUnit::Repository::Staging and also installs it (which causes it again to get precompiled). 16:26
nwc10 I'm supposed to say "Woohoo!" aren't I? 16:33
lizmat Woohoo! :-) 16:36
nine :)
timotimo does that last comment mean it'd be precompiling NativeCall in the same process twice, and that's where trouble could arise? 16:47
nine No, that was why reproducible.t failed. I solved that by just precompiling in an external process. NativeCall fails to load because NativeCall::Types gets a dependency on NativeCall. Which means that some object from CompUnit::Repository::Installation::install leaks into the precomp file. I had to fix similar issues for CURF, so I'm quite confident that this won't become a deal breaker 16:51
I already know pretty well how to debug issues like these (there were a lot of them). But not today :) Just a couple more minutes till my strength training starts and then its time for a relaxing evening... 16:52
lizmat nine++ 16:53
Xliff nine: So will the P6M merge errors be mitigated? 16:58
I'm thinking R#3922 and R#3075 16:59
linkable6 R#3922 [open]: github.com/rakudo/rakudo/issues/3922 P6M Merging error when using non-packaged our-scoped variables with a prefix.
R#3075 [open]: github.com/rakudo/rakudo/issues/3075 [precompilation] Conflicting use statement does not produce an error, but fails dynamic compilation.
18:00 domidumont left 18:01 domidumont joined
nine Xliff: it's possible there will be progress on that front 18:07
Geth rakudo: 424ce71a99 | (Elizabeth Mattijsen)++ | src/core.c/native_array.pm6
Simplify native shaped array .shape method

nqp::dimensions already returns a clone, so we can just shift off of that without needing indices. This makes calling .shape about 10% faster, but this is generally drowned out in accessing native shaped arrays.
18:51
rakudo: 532b5423f0 | (Elizabeth Mattijsen)++ | 3 files
Expose Array::Shaped1|2|3Array roles

Just as their native counterparts, so that can be used for dispatching if necessary.
18:54
19:15 domidumont left 19:40 sxmx joined 20:05 sxmx left 20:08 sxmx joined 20:11 sena_kun left 20:34 squashable6 left 20:37 squashable6 joined 20:46 melezhik joined
melezhik .tell jmerelo: basically Rakudist is all about testing Raku modules against different Rakudo distros / OS, so if you still need testing , maybe be you could outline a problems and put details to github.com/melezhik/RakuDist/issues ? 20:48
tellable6 melezhik, I'll pass your message to jmerelo
Geth rakudo: vrurg++ created pull request #4000:
Don't inherit mixin class language revision
21:22
vrurg .ask Altai-man Could you please do a blin run on R#4000? 21:24
linkable6 R#4000 [open]: github.com/rakudo/rakudo/pull/4000 Don't inherit mixin class language revision
tellable6 vrurg, I'll pass your message to Altai-man
21:25 Kaiepi left
Geth ¦ rakudo: vrurg self-assigned use v6.e.PREVIEW does not allow mixin of custom roles github.com/rakudo/rakudo/issues/3998 21:28
21:56 Kaiepi joined 22:29 vrurg left 22:30 vrurg joined 22:34 Xliff left 22:51 squashable6 left 22:53 squashable6 joined