Geth roast/revert-608-problem-solving-142: 9763b3048b | (Vadim Belman)++ (committed using GitHub Web editor) | 3 files
Revert "Test for export of long names"
02:00
roast: vrurg++ created pull request #611:
Revert "Test for export of long names"
roast: 9763b3048b | (Vadim Belman)++ (committed using GitHub Web editor) | 3 files
Revert "Test for export of long names"
02:01
roast: 93a61e6f04 | (Vadim Belman)++ (committed using GitHub Web editor) | 3 files
Merge pull request #611 from perl6/revert-608-problem-solving-142

Revert "Test for export of long names"
|Tux| Rakudo version 2019.11-403-g5d5701171 - MoarVM version 2020.01-1-g105d09d62
csv-ip5xs0.706 - 0.726
csv-ip5xs-206.260 - 6.712
csv-parser23.590 - 23.637
csv-test-xs-200.433 - 0.433
test7.082 - 7.384
test-t1.752 - 1.760
test-t --race0.798 - 0.825
test-t-2029.668 - 29.700
test-t-20 --race8.967 - 9.429
10:53
lizmat Files=1294, Tests=109644, 211 wallclock secs (28.20 usr 8.29 sys + 2950.20 cusr 274.81 csys = 3261.50 CPU) 11:48
Geth rakudo: 960b0a1fe2 | (Elizabeth Mattijsen)++ | src/core.c/Str.pm6
Fix Str.is-identifier and make it at least 6x as fast

It didn't quite handle all of the identifier semantics yet, like being able to using numbers (as long as they're not the first character of an alphanumeric part), and underscores being allowed anywhere.
Geoff Broadwell++ for spotting that.
12:02
lizmat m: my %s is SetHash[Str,Int,Num]; dd %s # how bad do we think it is that it is silently ignoring additional parameters ? 12:55
camelia SetHash[Str].new()
jnthn Not ideal, that's for sure 13:07
jdv79 nine: now i get this: nopaste.xyz/?36eec0cbc2e7ccab#z1EP...iEZuMUzSs= 13:08
but i'm gonna upgade my laptop as soon as i get a chance since fedora 25 is eol...
Guest38485 jdv79: your MoarVM version implies that you're not using nine's fix from yesterday 13:27
sena_kun I'll bump now 13:28
Guest38485 ++sena_kun
sena_kun if nobody says "No, don't!"
oki
Geth rakudo: 2261ef11fd | (Elizabeth Mattijsen)++ | src/core.c/allomorphs.pm6
Make IntStr.raku about 2x as fast

  - use low level concat
  - evade .Int, nqp::tostr_I will do the right thing
  - evade .Str, just fetch the attribute directly
  - at 2/3rds of the number of allocations
13:39
rakudo: e262da1d11 | (Elizabeth Mattijsen)++ | src/core.c/allomorphs.pm6
Make NumStr.raku about 60% faster

  - use low level concat
  - evade .Int, just fetch the attribute directly
  - evade .Str, just fetch the attribute directly
  - at 70% of the number of allocations
13:52
jdv79 oh, i thought it was bumped in. it wasn't? i just ran git pull and rebuilt. 14:15
Geth nqp: 47238e2ce2 | Altai-man++ | tools/templates/MOAR_REVISION
[MoarVM Bump] Brings 3 commits

MoarVM bump brought: github.com/MoarVM/MoarVM/compare/2...g105d09d62 105d09d62 Fix NativeCall libraries losing state during repossession b49cb0fa7 Bump VERSION for release 83393ef5a Update release name in ChangeLog
14:16
rakudo: 2f94362d82 | Altai-man++ | tools/templates/NQP_REVISION
[NQP Bump] Brings 3 commits

NQP bump brought: github.com/perl6/nqp/compare/2019....g47238e2ce 47238e2ce [MoarVM Bump] Brings 3 commits f00c47590 [JVM] Fix readfh for (buffered) input from TTY 413f71ef0 Merge pull request #594 from Altai-man/master
MoarVM bump brought: github.com/MoarVM/MoarVM/compare/2...g105d09d62 105d09d62 Fix NativeCall libraries losing state during repossession b49cb0fa7 Bump VERSION for release 83393ef5a Update release name in ChangeLog
jdv79 saw some moarvm git noise and assumed... i'll check again in 8h again...
Altai-man_ I just did a bump
couldn't do it yesterday because there was a broken test in spec
Geth rakudo: a4b62fb139 | (Elizabeth Mattijsen)++ | src/core.c/Str.pm6
Make Str.raku about 7x as fast

  - for pure alphanumeric strings
  - at about 15% of allocations
  - bypassing the char-by-char checks completely
14:18
lizmat afk for a few hours&
|Tux| Rakudo version 2019.11-406-ge262da1d1 - MoarVM version 2020.01-1-g105d09d62
csv-ip5xs0.707 - 0.715
csv-ip5xs-206.374 - 6.493
csv-parser23.091 - 23.135
csv-test-xs-200.422 - 0.423
test7.287 - 7.306
test-t1.736 - 1.785
test-t --race0.861 - 0.863
test-t-2029.743 - 29.977
test-t-20 --race9.205 - 9.434
14:33
nine I don't know what argument I could bring against "I as the sole author of the tool only I have used so far will be more responsive and reliable than this corporate backed team of 17 people who tend to software that has been in mission critical use for 12 years". Logic doesn't seem to apply... 15:00
github.com/perl6/problem-solving/i...-571573363
[Coke] I've worked with teams that do more R&D style development and deployment, but we've had to productize not only the app but the support as well. 15:01
MasterDuke nine: any further comments on github.com/MoarVM/MoarVM/pull/1224 ? 15:04
[Coke] ah, didn't realize this was not a $dayjob conversation at first. 15:10
|Tux| Rakudo version 2019.11-408-ga4b62fb13 - MoarVM version 2020.01-1-g105d09d62
csv-ip5xs0.704 - 0.732
csv-ip5xs-206.369 - 6.461
csv-parser22.783 - 22.803
csv-test-xs-200.421 - 0.431
test6.257 - 7.569
test-t1.724 - 1.739
test-t --race0.849 - 0.883
test-t-2029.095 - 30.188
test-t-20 --race9.143 - 9.205
15:41
(again because of the .Str speedup)
japhb |Tux|: test-t seems strangely resistant to optimization these days; very little seems to affect it much. 16:10
|Tux| it might be micro-speedgains. I expect huge wins in optimizing breaks 16:11
breaks being next, last and the like 16:12
|Tux| => PM
Geth rakudo: 29260acb19 | (Elizabeth Mattijsen)++ | 2 files
Make Srt.is-identifier to Pair local sub

This makes Pair.raku about 10% slower. Fixes #3408
17:49
lizmat do we have a way to stub an enum ?? 18:47
japhb You mean BOOSTRAP style, like the Bool magic? 18:53
lizmat well, like a class really 18:54
m: class A { ... }; class A { has $.b = 42 }; dd A.new.b
camelia 42
lizmat thing is: I want to add a new enum to the setting, but I need it *before* we actually have a Pair class 18:55
travis-ci Rakudo build failed. Elizabeth Mattijsen 'Make Srt.is-identifier to Pair local sub 19:04
travis-ci.org/rakudo/rakudo/builds/633882099 github.com/rakudo/rakudo/compare/a...260acb19d3
lizmat a flapper in t/02-rakudo/repl.t 19:07
restarted job
Geth rakudo: 27d1c4b5d5 | (Elizabeth Mattijsen)++ | src/core.c/Enumeration.pm6
Fix typo in comment
19:56
rakudo/index-improved: 561d1ba2f1 | (Elizabeth Mattijsen)++ | src/core.c/core_epilogue.pm6
Allow Str.(r)index searches on character classes

This adds 2 enums: UnicodeClass and NotUnicodeClass, and a range of values matching the nqp::const::CCLASS_xxx constants.
It also adds some candidates to Str.index and Str.rindex, that will dispatch on specifying a value of either of these two enums. This ... (11 more lines)
20:01
rakudo: lizmat++ created pull request #3409:
[WIP] Allow Str.(r)index searches on character classes
20:03
¦ problem-solving: lizmat assigned to jnthn Issue Expose nqp::findcclass and nqp::notfindclass functionality github.com/perl6/problem-solving/issues/146 20:04
nine MasterDuke: ACK 20:08
And yes, I went through all the new commits. Your work is just great! 20:10
Xliff o/ 20:21
Here's another open ended question / notice / WTF from me.
So when you get a chance, please take a look and respond via .tell 20:22
So I've created a parallel build program here: github.com/Xliff/p6-GtkPlus/blob/g...-build.pl6 20:23
It does a few things: 1) builds a dependency tree out of all modules it finds in a project (including all dependencies), 2) computes an optimal bottom-up build order, 3) attempts a parallel build using that optimal build order by: a) starting up parallel rakudo builds up to max cores and b) waiting to add then next item into the build scheduler when it's dependency list is empty 20:25
Here's the latest build log from that script for p6-GIO: github.com/Xliff/p6-GIO/blob/maste...s-20200107 20:26
HRM! 20:27
Scratch that... looks like the log was truncated! :(
Ah. Now untruncated! :D 20:31
If someone needs me to explain that format, then feel free to ask questions. 20:32
If you look at this line: github.com/Xliff/p6-GIO/blob/maste...0107#L2171
You will notice that the precomp only takes 1.6 seconds, but the actual parse time is almost 40 seconds. Are some of my rakudo threads hanging due to some form of thread locking?
That's it. Just thought I'd share some of the things I've noticed in my projects, lately. 20:33
Thanks!!
If you do want to see this for yourself, you can use p6-GLib instead of p6-GIO 20:35
git clone github.com/Xliff/p6-GLib.git
cd p6-GLib
zef install --deps-only . 20:36
Xliff scripts/dependency-build.sh 20:36
nine Xliff: well as you discovered yourself, yes rakudo locks the precomp directory 20:40
So you'll probably not get as much parallelism out of it as you'd wish.
The good news is, that this could be changed quite easily. Instead of having a lock file per precomp store, we could take a lock on the individual precomp files. That shouldn't even be hard to do, it just never made it to the top of my todo list. 20:41
Xliff makes anime eyes at nine. 20:42
Seriously, I'm seeing quite an improvement on compile times with this lock.
Without it, things will seriously scream and I can start working on getting things into release form .
AlexDaniel moritz: hello! I'm moving some repos and at the same time trying to clean some things up. Can you move github.com/perl6/rakudo.de to your own github account? I don't think we need it 20:43
nine Xliff: sorry, eyes won't work, I'll not be the one to do this. You seem to have the motivation, the capability, the test case and even the time to do it! 20:44
Xliff: github.com/rakudo/rakudo/blob/mast...y.pm6#L231 and github.com/rakudo/rakudo/blob/mast...e.pm6#L193
Xliff Yeek! 20:45
Will take a look then. Someone is gonna have to help me with a test case that's suitable for spectest and roast!
nine Nothing about this is harder than the parallel build thing you already did :) 20:45
Xliff Hah.
nine And I'll be happy to answer any questions about this code and assist. 20:46
Xliff Is there a way to rakudobrew your own commit hash?
That will make my testing easier.
nine I'd just not use rakudobrew
Xliff :( 20:47
I'd rather not have to use a VM. I might do it under another account then. *sigh*
nine Why? Just clone the rakudo repo and install it into a different prefix
I'm off to bed now. I backlog #raku-dev and #moarvm 20:50
AlexDaniel tony-o: there are two repos for pluggable, which one is maintained? github.com/perl6/perl6-pluggable 20:51
what's better, a repo called `examples` or `raku-examples` ? 20:54
Xliff nine: Good point. I think I will start on this, soonish. 20:55
AlexDaniel same question for modules, `lwp-simple` or `raku-lwp-simple`?
moritz AlexDaniel: (rakudo.de move) done 21:02
AlexDaniel thanks!
Geth nqp: 0d91c6b561 | (Elizabeth Mattijsen)++ | docs/ops.markdown
Show what the CLASS_xxx constants apply to

As far as could be determined from the MoarVM codebase.
21:42
Geth rakudo: 5d8ef435d7 | (Elizabeth Mattijsen)++ | src/core.c/Pair.pm6
Simplify identifier logic

CLASS_WORD actually is CLASS_ALPHANUMERIC + underscore.
21:56
Blin: Altai-man++ created pull request #18:
Use latest sources for obtaining module lists
22:14
Blin: 6f57b722c5 | Altai-man++ (committed using GitHub Web editor) | bin/blin.p6
Use latest sources for obtaining module lists
22:19
Blin: 06717ac0ae | (Aleks-Daniel Jakimenko-Aleksejev)++ (committed using GitHub Web editor) | bin/blin.p6
Merge pull request #18 from Altai-man/patch-1

Use latest sources for obtaining module lists
stoned75 hi. I have a trivial fix for Blin/issues/17 23:07
Altai-man_ stoned75, can you help me out then? :) 23:15
stoned75 I can't make a PR: if it's processed and if things are configured as stated in doc/README.md, a new version of the container image should end up on docker's hub 23:17
Altai-man_ stoned75, so casting the image author is the right thing? 23:19
stoned75 sorry, I'm not sure what you mean 23:20
Altai-man_ stoned75, as far as I understand, you are suggesting the image should be updated, and the author can do it, right? 23:20
stoned75 ok I did get you right :) 23:21
whell docker/README.md implies that docker's hub had been setup to automagically build image from commit into Blin repository
Altai-man_ \o/
stoned75 s/whell/well/ 23:22
ah. I can't push a branch to perl6/Blin. lemme push it somewhere else ;-) 23:25
oh Blin repo had already moved to Raku org ? 23:27
Altai-man_ yes 23:28
stoned75 nice
Geth Blin: stoned++ created pull request #19:
Fix Blin container image usage
stoned75 I the meantime I'm uploading an image I had built with this fix on docker's hub, if you'd like to try 23:29
but this image is huge... so it will take a while 23:30
alternatively you could build it yourself 23:31
grab my patch, apply it, and run from the top of Blin repo: podman build -t myblin -f docker/Dockerfile .
Altai-man_ stoned75, will see, if the PR will be accepted and it'll fix the issue, that'd be awesome. thanks!
stoned75 then test it with something like podman run --rm -ti -v /var/tmp/Blin-volume:/mnt:z myblin --nproc=8 HTTP::UserAgent 23:32
sure 23:33
MasterDuke nine: i wish i could say thanks, but 99% of what i did was just accepting the suggestions of gcc and clang 23:34