01:30 Kaiepi left 01:31 Kaiepi joined 01:34 sena_kun left 01:39 hungrydonkey joined 01:47 sena_kun joined 02:23 lucasb left 02:32 hungryd59 joined 02:35 hungrydonkey left 02:44 hungryd59 left, hungrydonkey joined 02:52 hungryd89 joined, hungrydonkey left 03:06 hungryd89 left 03:30 hungrydonkey joined 03:34 sena_kun left 03:48 sena_kun joined
Geth rakudo: vrurg++ created pull request #3481:
Renaming of perl6 to rakudo
04:20
05:29 hungrydonkey left 05:33 hungrydonkey joined 05:34 sena_kun left 05:50 sena_kun joined 07:33 sena_kun left 07:49 sena_kun joined 07:55 domidumont joined
|Tux| Rakudo version 2020.01-196-g73c5a25e8 - MoarVM version 2020.01.1-39-g657b536cf
csv-ip5xs0.707 - 0.719
csv-ip5xs-205.979 - 6.068
csv-parser22.985 - 23.456
csv-test-xs-200.369 - 0.370
test7.164 - 7.439
test-t1.786 - 1.797
test-t --race0.799 - 0.825
test-t-2029.349 - 29.492
test-t-20 --race8.988 - 9.294
09:04
09:30 unclechu joined 09:34 sena_kun left 09:48 sena_kun joined 09:52 rba[m] joined, AlexDaniel` joined, CIAvash joined 09:53 pmurias joined
lizmat Files=1302, Tests=110832, 211 wallclock secs (28.46 usr 8.12 sys + 2946.55 cusr 276.24 csys = 3259.37 CPU) 10:17
10:30 pmurias left
Geth rakudo: d49a418e2d | (Elizabeth Mattijsen)++ | src/core.c/Cool.pm6
Make sure that Cool.starts-with passes on :i, :m
10:47
roast: 5987cb99d0 | (Elizabeth Mattijsen)++ | S32-str/starts-with.t
Add proper tests for Cool.starts-with

By using a Match object, which should be quite common.
10:51 pmurias joined 10:54 hungryd7 joined, hungrydonkey left
Geth rakudo: 5a716a39f6 | (Elizabeth Mattijsen)++ | src/core.c/Cool.pm6
Make sure that Cool.ends-with passes on :i, :m
11:02
roast: ad1ac0d342 | (Elizabeth Mattijsen)++ | S32-str/ends-with.t
Add proper tests for Cool.ends-with

By using a Match object, which should be quite common.
11:03
11:06 Kaiepi left, domidumont left 11:07 Kaiepi joined
Geth rakudo: 7cb04813e5 | (Elizabeth Mattijsen)++ | src/core.c/Cool.pm6
Make sure that Cool.substr-eq passes on :i, :m
11:16
roast: 7bd47162ab | (Elizabeth Mattijsen)++ | S32-str/substr-eq.t
Add proper tests for Cool.substr-eq

By using a Match object, which should be quite common.
11:17
roast: 5ee3e41e38 | (Elizabeth Mattijsen)++ | 2 files
Move some support code to the loop it belongs
11:19
11:34 sena_kun left
Geth rakudo: 0761a882ab | (Elizabeth Mattijsen)++ | src/core.c/Cool.pm6
Remove unreachable candidates

Since the nameds are optional, the candidates without nameds will never be reached. So they can be removed. The alternative would be to create candidates with required nameds, but that would lead to a great many candidates, and since most of the work isn't done here anyway, and one should use the method way in general anyway if one is interested in performance, leaving it like this seems to be the best compromise.
11:44
11:48 sena_kun joined 12:05 patrickb joined
Geth rakudo: b140d7a96f | (Elizabeth Mattijsen)++ | src/core.c/Cool.pm6
Fix another Cool thinko with sub index

If the sub is called with a Str, it will already dispatch directly to Str.index, so no need to coerce to Str. If called with another Cool object, then it will take the long route through Cool.index.
12:11
rakudo: c2e608b276 | (Elizabeth Mattijsen)++ | src/core.c/Cool.pm6
Add support for :i :m to sub indices

Also, by not just taking the capture, make it about 3x as fast.
12:15
12:39 hungrydonkey joined
Geth rakudo: a6acfd0a4b | (Elizabeth Mattijsen)++ | src/core.c/Cool.pm6
Make sure that Cool.substr-eq passes on :i, :m
12:42
12:42 hungryd7 left
Geth roast: b548876f78 | (Elizabeth Mattijsen)++ | S32-str/indices.t
Add proper tests for Cool.indices

By using a Match object, which should be quite common.
12:43
roast: 627b2fd211 | (Elizabeth Mattijsen)++ | S32-str/index.t
Make output of matches better palatable
12:44
12:59 domidumont joined
nine [ 498s] t/02-rakudo/12-proto-arity-count.t (Wstat: 256 Tests: 352 Failed: 1) 13:02
[ 498s] Failed test: 82
build.opensuse.org/package/live_bu...eweed/i586
13:13 domidumont left
Geth rakudo: aefc851bab | (Elizabeth Mattijsen)++ | src/core.c/Cool.pm6
Fix sub indices proto

Spotted by Stefan Seifert++. Note to self: always run "make test"!
13:19
lizmat afk for a few hours&
13:33 Kaiepi left, Kaiepi joined 13:34 sena_kun left
pmurias vrurg: re renaming classes in the jvm backend to we get any significant benefit in doing that? 13:36
13:47 lucasb joined 13:49 sena_kun joined
vrurg pmurias: we must be getting away from perl6 completely. No other benefits. 14:11
pmurias I think moving the files and renaming packages should work, I'm a bit hesitant to move stuff on the jvm backend when it's work on much 14:20
patrickb rba: DRBD sounds like it might be a solution for how to synchronize files between the different servers. 14:22
vrurg pmurias: I don't think it makes sense renaming org/perl6 subdir while it's still org.perl6 inside. I leave it on you or anybody else to rename this part then. 14:23
patrickb rba: Basic principle: A single server has RW access to the filesystem, all others are read only. Files are automatically synced.
vrurg patrickb: while you're here. Where was RAKULIB discussed? I can't find anything on this. 14:25
rba patrickb: Ok. Did you use DRBD before? And do you know if this can be run in a docker container?
patrickb rba: colabti.org/irclogger/irclogger_lo...-02-11#l79 14:26
rba: It's part of the linux kernel and rather low level. DRBD creates a virtual block device. So it would have to happen on the host. 14:27
I haven't used it before 14:28
vrurg patrickb: thanks! 14:34
15:26 domidumont joined 15:33 sena_kun left 15:38 hungrydonkey left 15:42 hungrydonkey joined 15:49 sena_kun joined 16:04 travis-ci joined
travis-ci Rakudo build failed. Elizabeth Mattijsen 'Add support for :i :m to sub indices 16:04
travis-ci.org/rakudo/rakudo/builds/650411840 github.com/rakudo/rakudo/compare/b...e608b2764b
16:04 travis-ci left
[Coke] Looks like that's an actual test failure. 16:08
16:12 pmurias left 17:00 domidumont left 17:02 travis-ci joined
travis-ci Rakudo build failed. Elizabeth Mattijsen 'Make sure that Cool.substr-eq passes on :i, :m' 17:02
travis-ci.org/rakudo/rakudo/builds/650421012 github.com/rakudo/rakudo/compare/c...acfd0a4b39
17:02 travis-ci left 17:14 dogbert17 left 17:33 sena_kun left 17:45 domidumont joined 17:48 sena_kun joined 17:52 domidumont left 17:55 domidumont joined 18:03 hungrydonkey left 18:22 domidumont left
nine patrickb: DRBD is awesome, but it's no file distribution system. It's mirroring a block device. You cannot mount that ddvice on the secondary at all, not even read only 18:45
File systems need to write to e.g. replay a log on mount. 18:46
lizmat [Coke] I fixed the problem, Travis is just lagging...
patrickb nine: Sad! It sounded so good... :-(
nine It's also resrricred to 1 standby unless you use the proxy. 18:47
DRBD shines in failover cluaters, i.e. active server dies, the other takes over 18:48
Itvdoes support active/active configurations but you need a cluster file system to make use of that. And those of course bring their own complexities 18:49
Good old rsync or git are plenty for distributing static assets to readbonly mirrors. If you really want to go fancy (because you have millions of files) you can use btrfs send/receive 18:51
patrickb And all at once lsyncd looks a lot more attractive... 18:53
nine: My hope was to make it do the syncing automatically. Calling rcync every five minutes via cron might do the job good enough... 18:54
nine There's nothing wrong with old fashioned proven solutions in system administration :) Don't try to solve issues you do not yet have 18:55
moritz there's also unison for bidirectional syncing (though doesn't work well for Really Big directories) 19:00
Geth rakudo: 16962c8417 | (Elizabeth Mattijsen)++ | 3 files
Use native ints for the Date attributes

Marginally makes Date handling faster. At the expense of handling 1858-11-17 slower, as that is day 0 for daycount, and since daycount is now a native, 0 now indicates not being set (rather then undefined).
19:04
19:18 patrickb left
lizmat sena_kun: how badly will Comma break if we remove the (Foo) from $?FILE ? 19:24
sena_kun lizmat: I am not sure if horribly much, but even if putting Comma aside, I see such a removal as a breakage... not sure how bad one, but I imagine people can have regexes parsing it and then boom they'll break 19:27
lizmat but that was what lucs was trying to do, really :-) 19:28
sena_kun I am thinking about a possible warning, but OTOH that'll probably be rather annoying... and some people never update their stuff as well. 19:29
lizmat also: in the test suite, there is a lot of "use lib $?FILE.IO.parent(2).add: 'packages/Test-Helpers';"
that assumes $?FILE is (Foo) free
sena_kun yeah
lizmat as soon as the script would be precompiled, it would break :-()
19:34 sena_kun left 19:36 domidumont joined 19:37 domidumont left 19:38 hankache joined 19:47 sena_kun joined
sena_kun phew, finally I'm here, stupid connection... 19:47
lizmat: from my undelivered messaged: Comma won't be affected by this. do you suggest to put the fix into next release? 19:48
releasable6: status
releasable6 sena_kun, Next release in ≈7 days and ≈23 hours. 4 blockers. 0 out of 205 commits logged
sena_kun, Details: gist.github.com/b9a405faf4ad73be1c...277889650f
lizmat sena_kun: yeah, I think that would be appropriate...
Geth rakudo: 67022d4c67 | (Elizabeth Mattijsen)++ | src/core.c/Date.pm6
Make Date.Str (default formatter) about 250x as fast

It dawned on me that using sprintf (and its known overhead) was really overkill for stringifying a Date, as we know that the values in the object strictly have a limited value range. So use a combo of nqp::list_s, nqp::push_s, nqp::tostr_I and nqp::join for this amazing performance improvement.
I guess I should really get back to looking at sprintf's implementation again.
19:49
sena_kun stares at blockers
lizmat: can you whip up a patch this weekend?
then I can run blin as usual and see how it goes
lizmat sena_kun: will start looking at that now
sena_kun lizmat: there is still a whole Saturday, so no need to hurry 19:50
lizmat hmmm.... $?FILE already has the (Foo) *during* precompilation 19:55
lucs lizmat: FWIW, having 'no precompilation' in my file makes $?FILE report the wrong file :/ 20:00
lizmat what does it report? "-e" ? 20:01
lucs No, it reports that the file is Bar.pm6 (Bar.pm6 uses Baz.p6, which uses Foo.pm6) 20:03
lizmat could you add that to the ticket ?
lucs I'll try to build a mwe.
lizmat ok
please also add how the files are organized and where the "no precompilation" lives 20:04
lucs Yep. 20:05
Ok, got one, will write it up... 20:09
[Coke] mwe? 20:12
lucs Minimal Working Example. 20:13
(I think LaTeX people use that term -- I like it.)
lizmat: github.com/rakudo/rakudo/issues/25...-586456595 20:18
Maybe it should be MFE (failing) :) 20:20
lizmat stackoverflow.com/questions/602225...-their-own refers to an MRE 20:29
sena_kun: after looking at this for the better part of an hour, I still do *not* have a clue where the (Foo) comes from 20:30
20:32 hankache left
lizmat nine: any ideas? 20:33
maybe it is a property of CompUnit/Repository/FileSystem ?
nope, it happens with zef-installed modules as well 20:35
lucs Oh, I'll use MRE from now on.
nine It gets added durjng precompilation by CURI and CURF 20:37
[Coke] ah. mwe is like mvp. 20:39
lucs v?
(and p?)
[Coke] minimum viable product
lucs Ah
nine github.com/rakudo/rakudo/blob/6702...n.pm6#L283 20:41
20:41 hankache joined
nine lizmat: ^^^ 20:41
lizmat nine++ 20:42
20:56 patrickb joined 20:57 patrickb left 21:05 hankache left 21:06 patrickb joined
Geth rakudo: 2e69d7a888 | (Elizabeth Mattijsen)++ | 2 files
Remove extraneous information from $?FILE

  - specifically for precompiled files
  - fixes R#2539
  - tried making it an IO::Path
But that didn't stick for some reason, will need some further investigation. But since users of $?FILE are not expecting it to be an IO, and already coerce it to an IO (like in many places in roast), this appears to be a good intermediate solution.
21:10
linkable6 R#2539 [open]: github.com/rakudo/rakudo/issues/2539 [detrap] $?FILE shouldn't add garbage to end of filename.
21:13 pmurias joined
nine lizmat: please revert that. This will get you backtraces without module names, i.e. the whole point for why it's "source-name" not "source-file" is lost 21:14
lizmat ok, so the problem is really that it winds up in $?FILE ? 21:15
it *is* spectest and test clean
sena_kun lizmat: yes, backtraces should have module names 21:16
Geth rakudo: 4829711aa9 | (Elizabeth Mattijsen)++ | 2 files
Revert "Remove extraneous information from $?FILE"

This reverts commit 2e69d7a88879ccb75cc9f20b5f52d0bf15c86417.
lizmat well, it not really module names, it's the name of the file of the compunit
they usually are the same, but not necessarily 21:17
sena_kun yes, compunit names. :)
21:17 domidumont joined
nine lizmat: the proper way to fix this is by splitting the source-name into a path and a long name on the receiving side, I think in Perl6::World::current_file 21:21
The file name and comp unit names are wildely different for installed modules, i.e. the "my backtrace is full of SHAs and I have no idea what modules those are" problem 21:23
lizmat so we really want a $?FILE and a $?ORIGINAL-FILE ? 21:24
21:26 domidumont left 21:33 sena_kun left
nine I can't think of a use for $?ORIGINAL-FILE outside of error messages 21:41
lizmat if World.current-file returns the part before the ' (', should the stack trace still be ok ?
because that appears to depend on $?FILES ? 21:42
nine Can't say for sure anymore. It's been too many years 21:43
lizmat yeah, looks like 21:45
21:49 sena_kun joined 22:16 MasterDuke left
Geth rakudo: 4e12365b19 | (Elizabeth Mattijsen)++ | 3 files
Remove extraneous information from $?FILE

  - specifically for precompiled files
  - fixes R#2539
  - only changes the $?FILE in the file
  - does *not* touch the internal name of the compunit
  - so that stack traces will still show the complete information
This is a *HACK*. It needs a better solution.
22:17
linkable6 R#2539 [open]: github.com/rakudo/rakudo/issues/2539 [detrap] $?FILE shouldn't add garbage to end of filename.
lizmat nine: better like that ?
I wonder if this actually fixes pmurias' attempt at running the spectest from precompiled test-scripts 22:20
cause $?FILE.IO.parent.add... should now also work in precomped files 22:21
lizmat calls it a day... 22:27
23:33 sena_kun left 23:43 lucasb left 23:49 sena_kun joined