ugexe ah 00:03
samcv gonna try getting up to date and then hopefully this segfault will go away ack 00:04
ugexe github.com/rakudo/rakudo/blob/nom/...le.pm#L214 # if i change this to always use `self!get-line-slow-path()` it croaks during tools/install-dist.pl about invalid utf8. Is it expected that switch doesn't act the same functionally? 00:06
samcv ack no. segfault still here.. ugh 00:07
not even using any async things
gonna pull out gdb
eveo ugexe: reading the code, making it always slow-path would lose all the last after you finish reading the file. I presume $!decoder.is-empty is false and it won't ever consume more chars 00:11
ugexe grr, im not able to wrap my head around that 00:18
eveo ugexe: oh, OK, I misread the else branch 00:26
ugexe: tho not entirely
ugexe: Here you read into the decoder: github.com/rakudo/rakudo/blob/nom/...le.pm#L220 Since you're only using slow-path, you'll be reading too much; you'd read even when we can still produce lines with .consume-line-chars. So eventually the file is read entirely, and the "if $buf.elems {" conditional will be always false. So you always go into the else branch, which does do 00:27
.consume-chars, but it also includes the :eof thing, which ends up calling decodertakeline with :eof set github.com/rakudo/rakudo/blob/nom/...der.pm#L33 and I'm guessing that makes the decoder think there aren't any more lines and it eofed?
Geth rakudo/nom: bb2bafa63c | (Nick Logan)++ (committed using GitHub Web editor) | src/core/IO/Handle.pm
Fix max byte count
 00:34
ugexe that seems like it makes sense 00:36
MasterDuke c: HEAD~80..HEAD~75 my int $i = 0; while ($i = $i + 1) <= 1000000 { }; say now - INIT now 00:40
committable6 MasterDuke, ¦58c01c5: «0.5323520» ¦6805f5c: «0.5637109» ¦ccfa5e5: «0.5386451» ¦1ac7996: «0.0095736» ¦0a10082: «0.007038» ¦7fa8568: «0.0083396» 00:41
eveo ugexe: how come?
Well, the "fix" bit. How come old value was broken?
ugexe everywhere else 100000 is used except that one 00:42
eveo Ah
MasterDuke the slowdown happens with this commit github.com/rakudo/rakudo/commit/cc...15f4103752 "Stop the optimizer from using what I assume are old semantics for multi dispatch" 00:58
eveo m: my int $i = 0; while ($i = $i + 1) <= 1000000 { }; say now - INIT now 00:59
camelia 0.55452843
eveo m: my int $i = 0; while ($i = $i + my int $ = 1) <= 1000000 { }; say now - INIT now
camelia 0.0894072
eveo Yeah, it's correct.
It was using the wrong multi before 01:00
buggable: toast 01:03
buggable eveo, Between 2017.05-482-g932b10f and 2017.05: 1 (0.12%) modules got burnt; 3 (0.36%) got unsucced; 200 (24.07%) out of 831 modules appear unusable. See toast.perl6.party/ for details.
eveo And the one burn is a floppy module's tests
one more unsucc is a deleted module, and last one.... is Inline::Perl5 with "Invalid json? File: /home/cpan/.zef/store/Inline-Perl5-0.27.tar.gz/Inline-Perl5-0.27/Build.pm"
nine: ^ so looks like .27 still got that issue? 01:04
gonna run a stresstest on HEAD on Windows and I think then I'm ready to release
m: my int $i = 0; while (++$i) <= 1000000 { }; say now - INIT now 01:09
camelia 0.0921624
travis-ci Rakudo build failed. Zoffix Znet 'Merge pull request #1103 from stmuk/nom 01:12
travis-ci.org/rakudo/rakudo/builds/244014523 github.com/rakudo/rakudo/compare/f...35c14eb387
buggable [travis build above] ✓ All failures are due to timeout (0), missing build log (0), GitHub connectivity (0), or failed make test (3). Across all jobs, only 04-nativecall/06-struct.t test file failed.
eveo fuck 01:13
Man, the Proc piping issue irks me. Sucks to make a release with it in it. Especially on the release where I claimed we used Toaster so everything is peachy 01:16
eveo pokes at it some more, in hopes of finding a fix
Geth rakudo/nom: 44acdd2c26 | (Zoffix Znet)++ | src/core/IO/Path.pm
Revert "Optimize newline translation in IO::Path.slurp"

This reverts commit 6788833bb831da6a3540755ce2ea8855a92b59ca.
Based on travis results, this commit may be causing a nativecall struct test failure. Can't repro this locally. Throwing this out of the release, just in case.		 01:18
ugexe fwiw osx seems to have the io pipe problem a lot less than the linux system i test on 01:19
only 3 failures out of 100 for the gist you posted 01:20
eveo Yeah, tho 3 out of 100 may as well be 100 out of 100 if you're using it in a proper program 01:21
ugexe but it hints towards the problem since osx has a decently smaller default pipe buffer size than linux 01:23
eveo Ah 01:24
Geth rakudo/nom: cd547021bd | (Zoffix Znet)++ | src/core/CompUnit/Repository/Installation.pm
Use .subst instead of fast-pathing newline translation

Reverting this outta the release for same reasons as in
  github.com/rakudo/rakudo/commit/44acdd2c26		 01:25
ugexe should all the .tap's used in Proc.pm be .act's? 01:28
eveo no idea 01:31
eveo millimeters closer to solution 01:35
god, windows stresstest is a bloodbath. I guess July will be the month of fixing Windows 01:41
ohmaigawd 01:56
I think I see it!
eveo compiles 01:57
\o/ |o| /o/ -o- |o| \o\ 01:59
Yeaaaah...! Fixed it :D
And the fix has fewer letters than the number of hours I spent hunting this bug :P
ugexe noice 02:00
Geth rakudo/nom: 239c505ea5 | (Zoffix Znet)++ | src/core/Proc/Async.pm
Fix regression in piping output of one Proc to another

In bound STDIN, we write to the proc, but forget to `await` that write, so we end up done-ing the supply before it finished writing, which closes stdin and intros a race where data piped to STDIN gets lost.
Fixes flopping of t/spec/S32-io/pipe.t spectest.		 02:10
eveo windows spectest hangs too. Last test files in output in both cases are socket tests. 02:14
BenGoldberg wonders if the Promises returned by .write should be automatically awaited if sunk. 02:20
eveo meh 02:23
It's bad enough async react is written as `start react {}` would suck to uglify it further with `$ = start react {}` 02:26
maybe a pragma
BenGoldberg I'm not suggesting all promises, just the ones from .write, since it's the kind of thing that people overlook. 02:47
eveo But then you're introing a special-case 02:52
And just creating confusion on when users are supposed to await for promises instead of just sinking them 02:54
m: class YourMom {}.WHY 03:13
camelia ( no output )
eveo m: class YourMom {}.WHY.say
camelia No documentation available for type 'YourMom'.
Perhaps it can be found at docs.perl6.org/type/YourMom
»
eveo heh
So JVM now has an implemented Proc::Async? 03:22
j: $*PERL.compiler.version.say
camelia v2017.05.486.g.239.c.505.ea
eveo j: Proc::Asyns.new
j: Proc::Async.new
camelia Could not find symbol '&Asyns'
in block <unit> at <tmp> line 1
Cannot resolve caller new(Proc::Async: ); none of these signatures match:
(Proc::Async $: *@ (Any $path, *@args), *%_)
in block <unit> at <tmp> line 1
eveo j: with Proc::Async.new('ls') { .stdout.tap: &say; await .start } 03:23
camelia Inline-Perl5
Perlito
Tst.pm
bin
dalek-queue
evalbot
evalbot.log
lala.out
lib
log
mbox
nqp-js
p1
p2
p6eval-token
perl5
rakudo-j-1
rakudo-j-2
rakudo-j-inst
rakudo-j-inst-1
rakudo-j-inst-2
rakudo-m-1
rakudo-m-2
r…
eveo cool
Geth rakudo/nom: 03c9bd68a6 | (Zoffix Znet)++ | docs/ChangeLog
Log all changes to date
 03:24
eveo It's hammer time 03:25
NeuralAnomaly: status 03:28
NeuralAnomaly eveo, [✔] Next release is today. Since last release, there are 68 new still-open tickets (0 unreviewed and 0 blockers) and 0 unreviewed commits. See perl6.fail/release/stats for details
eveo NeuralAnomaly: cut the release 03:29
travis-ci Rakudo build failed. Zoffix Znet 'Merge pull request #1104 from ugexe/proc-merge
travis-ci.org/rakudo/rakudo/builds/244090549 github.com/rakudo/rakudo/compare/9...7496671c7b
buggable [travis build above] ✓ All failures are due to timeout (0), missing build log (0), GitHub connectivity (0), or failed make test (3). Across all jobs, only 04-nativecall/06-struct.t test file failed.
eveo Come on, robot. It's me, lovable eveo!
NeuralAnomaly: status 03:30
NeuralAnomaly eveo, [✘] Next release is today. Since last release, there are 68 new still-open tickets (0 unreviewed and 0 blockers) and 1 unreviewed commits. See perl6.fail/release/stats for details
eveo NeuralAnomaly: status
NeuralAnomaly eveo, [✔] Next release is today. Since last release, there are 68 new still-open tickets (0 unreviewed and 0 blockers) and 0 unreviewed commits. See perl6.fail/release/stats for details
eveo NeuralAnomaly: cut the release
NeuralAnomaly eveo, Will do! If you're feeling particularly naughty, you can watch me at perl6.fail/release/progress or go look at some cats icanhas.cheezburger.com/lolcats
eveo, ♥♥♥♥♥♥ Prep done
Geth nqp: 836ff40a0a | (Zoffix Znet)++ | tools/build/MOAR_REVISION
bump MoarVM version to 2017.06
nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2......2017.06
bc4ffab33a | (Zoffix Znet)++ | VERSION
¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2......2017.06
NeuralAnomaly eveo, ♥♥♥♥♥♥ nqp tests OK 03:40
eveo, ♥♥♥♥♥♥ nqp release tarball tests OK 03:49
eveo, ♥♥♥♥♥♥ nqp release DONE
eveo, ☠☠☠☠☠☠ Rakudo: generate release announcement 03:50
eveo, ☠☠☠☠☠☠☠☠☠☠ ABNORMAL EXIT!
eveo booo
system perl6 was set on 2016.05 03:52
NeuralAnomaly: steps 03:53
NeuralAnomaly eveo, all pre nqp r post pre-r6 pre-blank-slate nqp-clone nqp-bump-vers nqp-build nqp-test nqp-tar nqp-tar-build nqp-tag nqp-tar-sign nqp-tar-copy r-clone r-prep-ann r-bump-vers r-build r-p5 r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp
eveo NeuralAnomaly: run r-prep-ann r-bump-vers r-build r-p5 r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp
Geth rakudo/nom: fe86fe1de6 | (Zoffix Znet)++ | docs/announce/2017.06.md
Generate release announcement for 2017.06
 03:54
rakudo/nom: version bump brought these changes: github.com/perl6/nqp/compare/2017.......2017.06
818a530f34 | (Zoffix Znet)++ | tools/build/NQP_REVISION
rakudo/nom: version bump brought these changes: github.com/perl6/nqp/compare/2017.......2017.06
c252b39e2f | (Zoffix Znet)++ | VERSION
¦ rakudo/nom: version bump brought these changes: github.com/perl6/nqp/compare/2017.......2017.06
NeuralAnomaly eveo, ☠☠☠☠☠☠☠☠☠☠ ABNORMAL EXIT! 04:02
eveo, ☠☠☠☠☠☠ Rakudo: make stresstest (master)
eveo NeuralAnomaly: run r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp 04:03
NeuralAnomaly eveo, ☠☠☠☠☠☠ Rakudo: make stresstest (master) 04:05
eveo, ☠☠☠☠☠☠☠☠☠☠ ABNORMAL EXIT!
eveo fucking 'ell 04:06
First time seeing t/spec/S29-os/system.t flop
NeuralAnomaly: run r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp
man t/spec/S17-procasync/stress.t just doesn't wanna pass :/ 04:07
passes when I run it on its own :/ 04:08
NeuralAnomaly eveo, ☠☠☠☠☠☠ Rakudo: make stresstest (master)
eveo, ☠☠☠☠☠☠☠☠☠☠ ABNORMAL EXIT!
eveo NeuralAnomaly: run r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp 04:09
again failed :| 04:10
samcv eveo, so it seems i did more work and it seems rakudobrew builds with debug=3 by default. and so it is causing segfaults for me unless i build without debugging. not sure what debug level causes my script to fail 04:11
NeuralAnomaly eveo, ☠☠☠☠☠☠☠☠☠☠ ABNORMAL EXIT!
eveo, ☠☠☠☠☠☠ Rakudo: make stresstest (master)
samcv but no failure if i tell rakudobrew to bulid without debugging
eveo NeuralAnomaly: run r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp
Gonna by-pass tests if this run fails. Both failing files run fine when I run them manually with t/fudgeandrun 04:12
Hm. Wonder if the not-a-tty is a cause here tho
failed again. don't seem to be just a flopper
samcv do you have the logs eveo
eveo samcv: for what?
samcv the failing tests
eveo perl6.fail/release/progress
samcv thanks 04:13
eveo It doesn't say what exactly failed
NeuralAnomaly eveo, ☠☠☠☠☠☠☠☠☠☠ ABNORMAL EXIT!
eveo, ☠☠☠☠☠☠ Rakudo: make stresstest (master)
ugexe it passes for me locally on osx 04:14
samcv eveo, let me try running it on my system
since i seem to be having some weidhness
eveo NeuralAnomaly: steps 04:15
NeuralAnomaly eveo, all pre nqp r post pre-r6 pre-blank-slate nqp-clone nqp-bump-vers nqp-build nqp-test nqp-tar nqp-tar-build nqp-tag nqp-tar-sign nqp-tar-copy r-custom r-clone r-prep-ann r-bump-vers r-build r-p5 r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp
eveo NeuralAnomaly: run r-custom
NeuralAnomaly eveo, ♥♥♥♥♥♥ Rakudo: custom command
eveo, ☠☠☠☠☠☠☠☠☠☠ ABNORMAL EXIT!
eveo NeuralAnomaly: run r-custom
NeuralAnomaly eveo, ♥♥♥♥♥♥ Rakudo: custom command 04:16
eveo, ☠☠☠☠☠☠☠☠☠☠ ABNORMAL EXIT!
ugexe passes on a linux vm as well
eveo NeuralAnomaly: run r-custom 04:17
NeuralAnomaly eveo, ☠☠☠☠☠☠☠☠☠☠ ABNORMAL EXIT!
eveo, ♥♥♥♥♥♥ Rakudo: custom command
samcv i'm gonna run a whole stresstest on my system after rakudo finishes compiling... 04:18
to make sure --debug=3 on moarvm doesn't cause any of the spectest tests to fail
eveo I don't get it... it says "not ok 36 - run sets $cwd and $env" for an is-deeply test but I don't see the failure diagnostics message anywhere :/ 04:24
samcv i get failing tests
t/spec/S17-procasync/stress.t
eveo Is it actually failing or just a flopper?
samcv i will run it again
eveo Seems to fail consistently in the bot when run as stresstest, but standalone it's fine
samcv nope it's failing 04:25
eveo What's the full output?
samcv gist.github.com/193454e9bf2f728f9a...28fced57ec
eveo samcv: is that with debug stuff? 04:27
samcv building moar with debugging yes
which is the default rakudobrew builds... so
tho idk how much we care about that?
but still is something to think about 04:28
passes fine with --debug=0 just tested 04:32
Geth roast: c749b6b084 | (Zoffix Znet)++ | S29-os/system.t
Fudge new test for release

The test passes, but not when running via release bot. Probably something to do with how $*EXECUTABLE is set or something. I changed the command executed and I see :cwd works fine. So it's something with the test itself.
The test is new, added just a few days ago, and isn't part of 6.c		 04:36
eveo NeuralAnomaly: run r-stress r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp
bah. Just doesn't want to pass while part of stresstest :/ Works fine when I t/fudgeandrun it, even via the bot 04:37
the proc stresstest that is 04:38
NeuralAnomaly eveo, ☠☠☠☠☠☠ Rakudo: make stresstest (master)
eveo, ☠☠☠☠☠☠☠☠☠☠ ABNORMAL EXIT!
samcv eveo, i get many failed tests 04:39
t/spec/S17-procasync/stress.t t/spec/S32-io/IO-Socket-Async-UDP.t t/spec/S32-io/IO-Socket-Async.t
t/spec/S32-io/IO-Socket-INET.t
t/spec/S32-io/socket-recv-vs-read.t
with debug set
i'm gonna rerun it again though
err. those tests individually i mean
eveo NeuralAnomaly: run r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp 04:41
samcv t/spec/S32-io/IO-Socket-Async-UDP.t fails # failed to resolve host name 04:42
NeuralAnomaly eveo, ♥♥♥♥♥♥ Rakudo stresstest (6.c-errata) OK 04:43
eveo I can repro the "process not opened for write" thing with this snippet run a few times: gist.github.com/zoffixznet/8b4cfcc...7be599b2de
samcv ok all those failed to resolve host name except the first one
somehow debug must make things trigger more. flappers it seems 04:44
or make them happen consistently
at least that's what appears to be happening :|
TimToady waves from not-quite-DC 04:55
eveo \o 04:56
k, no idea where the race is at in proc async. Might even be in moar 05:05
eveo continues with release
NeuralAnomaly: run r-stress-v6c r-tar r-tar-build r-tar-p5 r-tar-stress r-tag r-tar-sign r-tar-copy post-scp
NeuralAnomaly eveo, ♥♥♥♥♥♥ Rakudo stresstest (6.c-errata) OK 05:07
eveo, ♥♥♥♥♥♥ Rakudo release DONE 05:15
eveo, ♥♥♥♥♥♥ Post: upload tarballs to rakudo.org and rakudo.perl6.org
eveo, ????????????????????????????????????????????????????????????
eveo, The release of **Rakudo #112 2017.06** has now been completed
eveo, ????????????????????????????????????????????????????????????
NeuralAnomaly celebrates with an appropriate amount of fun
eveo *phew* 05:16
That was a ton of work.
Toaster++ this would've been another lemon of a release without it
Geth rakudo/nom: 3d2a521c25 | (Zoffix Znet)++ | 3 files
Revert "Revert "Change the way we code-gen simple for loops.""

This reverts commit 932b10f6a909cbe6a0bb50891f82694e9348ec7c.		 05:30
rakudo/nom: 5a80412cb3 | (Zoffix Znet)++ | src/core/CompUnit/Repository/Installation.pm
Revert "Use .subst instead of fast-pathing newline translation"

This reverts commit cd547021bdf5fb47794354d1aea102d95942101b.
rakudo/nom: c4e14731f6 | (Zoffix Znet)++ | src/core/IO/Path.pm
Revert "Revert "Optimize newline translation in IO::Path.slurp""

This reverts commit 44acdd2c266966ef61af37bec2cf379e10851448.
roast: 8ca80ecbb4 | (Zoffix Znet)++ | S29-os/system.t
Revert "Fudge new test for release"

This reverts commit c749b6b08456d97b415b779d23021143adf27a3a.
rakudo/nom: ee4c3d3226 | (Zoffix Znet)++ (committed using GitHub Web editor) | docs/release_guide.pod
2017.06 is now in the past
 05:31
eveo NeuralAnomaly: status
NeuralAnomaly eveo, [✘] Next release will be in 3 weeks and 6 days. Since last release, there are 1 new still-open tickets (0 unreviewed and 0 blockers) and 3 unreviewed commits. See perl6.fail/release/stats for details
eveo .tell jnthn got two new tickets you may be interested in: "for loop no longer sinks stuff": rt.perl.org/Ticket/Display.html?id=131593 and "some data race in Proc::Async write" probably worth waiting until the piping is reworked as you planned to do next week. Ticket: rt.perl.org/Ticket/Display.html?id=131592 05:35
yoleaux eveo: I'll pass your message to jnthn.
eveo Alright. Time for a break from Perl 6. 05:36
eveo leaves to learn C for a couple of weeks.
The ZofBot is ever vigillant, so just say my name if you need me
nine .tell eveo Inline::Perl5's tar file should be fixed in 0.28 06:30
yoleaux nine: I'll pass your message to eveo.
[Tux] lizmat, noted 08:23
yoleaux 17 Jun 2017 20:53Z <lizmat> [Tux]: CSV can become a bit faster if you replace all m{} by //
[Tux] now fighting code in order to track down a "Attempt to free an unreferenced scalar: SV 0x..." in perl5 08:24
lizmat Files=1204, Tests=62514, 216 wallclock secs (13.21 usr 5.03 sys + 1310.41 cusr 132.51 csys = 1461.16 CPU) # pre-release 08:32
[Tux] This is Rakudo version 2017.06-4-gee4c3d322 built on MoarVM version 2017.06 08:34
csv-ip5xs 2.477
test 12.618
test-t 4.032 - 4.183
csv-parser 12.785
if using options, like in m:m{...}, that isn't possible, right? 08:36
timotimo for things like ignorecase and ignoremark you can put that after the initial / in the / / 08:40
stuff that influences how the match will be done, like :g or :ov or :ex or :n or :th for example, won't work inside the regex 08:41
[Tux] m{} → // ⇒ 4.034 | 4.057 | 4.062 | 4.067 09:00
so, no noticable speed gain
timotimo i was also wondering what about // is faster than m{} 09:05
[Tux] and going back to m{} last timeing was 4.007, so - as I prefer reading explicit m{} - I will keep the m{} notation 09:10
I won't argue that it might speed up other modules or scripts, but it does not speed up CSV 09:11
timotimo mhm 09:12
[Tux] anyway, that shows a consisten (very) low 4.0, so I am reasonable happy 09:17
dogbert17 with some luck it will go under 4 next week 09:24
lizmat m: for ^100000 { 42 ~~ /^ <[0..9]> / }; say now - INIT now 09:27
camelia 1.1691020
lizmat m: for ^100000 { 42 ~~ m/^ <[0..9]> / }; say now - INIT now
camelia 9.2692342
lizmat timotimo: ^^^
timotimo huh 09:28
lizmat m// calls Str.match, // goes straight to Regex.ACCEPTS afaict 09:29
m: for ^100000 { "42" ~~ /^ <[0..9]> / }; say now - INIT now # proper string 09:30
camelia 0.9959404
lizmat m: for ^100000 { "42" ~~ m/^ <[0..9]> / }; say now - INIT now # proper string
camelia 1.267983
lizmat hmmm
not quite as pronounced
timotimo m: for ^100000 { 42.Str }; say now - INIT now 09:34
camelia 0.0334138
lizmat timotimo: it's more about candidate selection I think 09:40
not about the actual coercion
travis-ci Rakudo build errored. Zoffix Znet 'Revert "Change the way we code-gen simple for loops." 09:41
travis-ci.org/rakudo/rakudo/builds/244097742 github.com/rakudo/rakudo/compare/8...2b10f6a909
buggable [travis build above] ✓ All failures are due to timeout (0), missing build log (2), GitHub connectivity (0), or failed make test (0).
lizmat 42 ~~ m// has to go through Cool.match
42 ~~ // probably uses native auto-coercion 09:42
m: my int $a = 42; my str $b = $a; dd $b 09:43
camelia "42"
dogbert17 lizmat: do you know anything about a method called 'tree'? 10:30
it's declared in Any.pm but does not seem to be used anywhere in the rakudo source 10:32
timotimo thrown out with the glr i suspect 11:23
lizmat dogbert17: seem to be tests for it in t/spec/S02-lists/tree.t, perhaps you can derive from that what it's supposed to do 12:43
dogb dogbert17: I haven't got a clue :-)
afk again&
dogbert17 perhaps jnthn or Zoffix knows what tree is used for 13:10
timotimo back when we had parcels which would "lazily flatten or itemize depending on context/operation", you had to use something like .lol or .tree to build a tree structure from nested parcels 13:12
otherwise it'd just become one flat list with .List or whateve
dogbert17 timotimo: interesting, there seems to be some connection with lol (I assume that is not the game :-) 13:13
timotimo right 13:14
lol is probably like .tree(1)
dogbert17 timotimo: the reason for asking is that there's a doc issue on .tree being documented but when going through the src it wasn't used so I'm wondering if it should be deprecated/removed instead
*not being documented 13:15
timotimo: wrt to parcels, so lists had a tendency to lose their structure before? 13:17
timotimo yup
the GLR was what got us rid of that
dogbert17 cool, so perhaps it should go ... 13:18
timotimo you used to have to work for it if you wanted structure
and it was often bothersome to figure out what would happen in any given situation
dogbert17 wasn't involved with Perl 6 in those ancient times :)
timotimo the GLR made things a whole lot more regular
stop it, you're making me feel old :)
dogbert17 :) 13:19
I'll ignore tree for now (doc wise)
timotimo bark up another method :)
dogbert17 will do, btw do you want to investigate a SEGV? 13:21
timotimo i can try 13:22
dogbert17 what happens if you run this a few times: 'perl6 t/spec/S17-procasync/stress.t'
preferably with 2017.06 13:23
timotimo let me build some new commits
dogbert17 :)
travis-ci Rakudo build canceled. Zoffix Znet 'Revert "Change the way we code-gen simple for loops."
travis-ci.org/rakudo/rakudo/builds/244097742 github.com/rakudo/rakudo/compare/8...2b10f6a909
buggable [travis build above] ☠ Did not recognize some failures. Check results manually.
BenGoldberg m: my %_ = named_slurpy => 42; say [||] %_<<optional named_names pos_slurpy pos_onearg named_slurpy>>; 13:33
camelia 5===SORRY!5=== Error while compiling <tmp>
Cannot use placeholder parameter %_ outside of a sub or block
at <tmp>:1
------> 3my %_7⏏5 = named_slurpy => 42; say [||] %_<<opti
BenGoldberg m: my %meh = named_slurpy => 42; say [||] %meh<<optional named_names pos_slurpy pos_onearg named_slurpy>>;
camelia 42
timotimo may want [//] instead if falsy things are allowed to be passed 13:43
dogbert17: is sigpipe interesting at all? 13:48
dogbert17 you got a broken pipe? 13:50
timotimo stepping through the thing in gdb is annoying 13:51
so many broken pipes
dogbert17 if you make the nursery smaller it will SEGV for sure with: 13:52
Program received signal SIGSEGV, Segmentation fault.
0xb7c74c67 in gc_mark (tc=0x804c468, st=0x8096da8, data=0xb28a178, worklist=0xb225cc8) at src/6model/reprs/MVMCallCapture.c:54
54 if (flag_map[flag] & MVM_CALLSITE_ARG_NAMED) {
timotimo oh, in a call capture, eh?
dogbert17 looks like it, should I throw up a gist or do you want to replicate it? 13:53
timotimo i'll try to replicate it 13:54
how small is your nursery?
dogbert17 32768 * 8 13:55
travis-ci Rakudo build passed. Nick Logan 'Fix max byte count' 13:58
travis-ci.org/rakudo/rakudo/builds/244106917 github.com/rakudo/rakudo/compare/9...2bafa63c1a
timotimo yup, i got it 14:01
dogbert17 yay
timotimo i wonder how big it'll inflate when run with rr record :)
i mean, i reproduced it
dogbert17 do you think this commit bu nine might be helpful: "When compiling moarvm with --debug, the MVM_dump_p6opaque(tc, obj) function is available for calling from e.g. a debugger session." 14:02
timotimo forgot to turn optimize down 14:03
not sure in this case
travis-ci Rakudo build passed. Zoffix Znet 'Revert "Optimize newline translation in IO::Path.slurp" 14:15
travis-ci.org/rakudo/rakudo/builds/244116948 github.com/rakudo/rakudo/compare/b...acdd2c2669
dogbert17 timotimo: here's som valgrind output, perhaps it might prove useful: gist.github.com/dogbert17/a0406d85...42d4d46e0c 14:24
timotimo i wonder if all of this happens inside the same piece 14:26
this callsite is set in MVM_proc_args_init and at some point it gets corrupted in some way
dogbert17 so the valgrind output does not give any further clues? 14:30
timotimo hmm 14:37
so something was holding on to the pointer but that pointer wasn't pointed out to the gc 14:38
dogbert17 is this some MVM_root issue?
timotimo or worklist adding in a gc_mark
dogbert17 I might have been wasting time here, the failing test has the following text "run()ning two procs and passing the :out of one to the :in of the other doesn't segfault" 14:43
now, I believe this is something that will be looked into next week so perhaps we can ignore it :)
timotimo it'd be kind of cool if i could trigger a heap snapshot from inside gdb 14:45
dogbert17 is that possible?
timotimo the only problem is that the dump code is in nqp 14:46
it has to do a little bit of work that we use nqp for
we can rewrite it in C, then it could be done exclusively from inside gdb
it'll also have to mark all threads as blocked because gdb is still keeping them stopped 14:47
dogbert17 so it's nothing which is available today then 14:51
travis-ci Rakudo build passed. Zoffix Znet 'Use .subst instead of fast-pathing newline translation 14:53
travis-ci.org/rakudo/rakudo/builds/244118538 github.com/rakudo/rakudo/compare/4...547021bdf5
timotimo right 14:55
i ate ice cream sittign at my laptop and now one of my keys is sticky :|
clean-up-goop to the rescue!
dogbert17 :) 14:57
timotimo ick 14:58
it might have gotten too hot or something 14:59
it didn't stay in one piece when i tried to drop it back into its cage 15:01
i.e. some stayed stuck on my hand :<
dogbert17 can't your cats take care of it :)
timotimo i don't think it'd be healthy for them 15:03
also it smells rather pungent :)
ugh, i can't set a breakpoint for "the object in question is added to a worklist" because the problem is that nothing is adding it to any worklist 15:09
with valgrind active i could hunt for things that point at the object
dogbert17 wonder what kind of object it is 15:10
travis-ci Rakudo build failed. Zoffix Znet 'Fix regression in piping output of one Proc to another 15:14
travis-ci.org/rakudo/rakudo/builds/244125228 github.com/rakudo/rakudo/compare/c...9c505ea53b
buggable [travis build above] ☠ Did not recognize some failures. Check results manually.
timotimo gcc does have a find method 15:16
it wants a start and end point, though
dogbert17 tricky 15:21
timotimo i mean i could look through the two nurseries, definitely 15:29
but the gen2, that's a bit trickier
in theory we could build a little c function that scans through all gc-managed spots 15:31
dogbert17 sounds difficult 15:35
timotimo not so bad
it's a lot of work to do manually
travis-ci Rakudo build passed. Zoffix Znet 'Log all changes to date' 15:39
travis-ci.org/rakudo/rakudo/builds/244134019 github.com/rakudo/rakudo/compare/2...c9bd68a6e9
dogbert17 where are the easy bugs 15:51
timotimo yeah 15:59
dogbert17 m: my $sh = <a>.SetHash; for $sh.values { $_ = 0; $_ = 1 } # is this easy? 16:03
camelia ( no output )
dogbert17 I get an error when I run this 16:04
travis-ci Rakudo build passed. Zoffix Znet '[release] bump VERSION to 2017.06' 16:07
travis-ci.org/rakudo/rakudo/builds/244136794 github.com/rakudo/rakudo/compare/0...52b39e2f84
dogbert17 for some reason this test fails: github.com/MoarVM/MoarVM/blob/mast...MHash.c#L7 16:08
Geth nqp: dmaestro++ created pull request #364:
Issue/353 sprintf octal padding		 16:10
timotimo i bet we have to set a whence on the scalar container when we've removed the value out of the set 16:11
so that when you put a value back in, it knows how to store it back at the same key 16:12
dogbert17 but why does it 'work' on Camelia but not my poor 32 bit vm?
timotimo luck?
dogbert17 hmm 16:13
timotimo oh, it does keep the values stored in there
dogbert17 #0 get_string_key (tc=0x804c468, key=0x809fab0) at src/6model/reprs/MVMHash.c:7
#1 0xb7c5e18d in bind_key (tc=0x804c468, st=0x80969e8, root=0x80a2f30, data=0x80a2f40, key_obj=0x809fab0, value=..., kind=8) at src/6model/reprs/MVMHash.c:87
#2 0xb7c54a98 in MVM_repr_bind_key_o (tc=0x804c468, obj=0x80a2f30, key=0x809fab0, val=0x80a2f48) at src/6model/reprconv.c:481
#3 0xb7c91a15 in add_knowhow_how_method (tc=0x804c468, knowhow_how=0x809fa90, name=0xb7d2ca85 "new_type", func=0xb7c90719 <new_type>) at src/6model/bootstrap.c:264
timotimo huh, really? 16:15
dogbert17 yup
timotimo that's in the code you pasted above? 16:16
dogbert17 m: my $sh = <a>.SetHash; for $sh.values { $_ = 0; $_ = 1 } # this code
camelia ( no output )
timotimo you got a tiny nursery perhaps? 16:17
dogbert17 yes, from out previous experiments :)
*our
should I make it normal again?
timotimo i still have a tiny one, as well. i don't get a crash from that code 16:18
dogbert17 does valgrind say anything?
timotimo no, only the destruction parts 16:20
i might look into that, but i'm kind of distracted and such
dogbert17 thx 16:21
timotimo they don't really help any
dogbert17 who, the cats? 16:22
timotimo just to make valgrind not output anything when nothing happens
dogbert17 :)
timotimo no, the destruction parts that i could fix
dogbert17 'p IS_CONCRETE(key)' returns 1 16:24
travis-ci Rakudo build passed. Zoffix Znet 'Revert "Revert "Optimize newline translation in IO::Path.slurp"" 16:31
travis-ci.org/rakudo/rakudo/builds/244147543 github.com/rakudo/rakudo/compare/c...e14731f6e1
dogbert17 when the code fails 'p REPR(key)->ID' == 5 whatever that is 16:42
ah 5 means P6opaque 16:43
travis-ci Rakudo build failed. Zoffix Znet '2017.06 is now in the past' 16:56
travis-ci.org/rakudo/rakudo/builds/244147611 github.com/rakudo/rakudo/compare/c...4c3d322654
buggable [travis build above] ☠ Did not recognize some failures. Check results manually.
ugexe hmm on osx I'm getting segfaults when i run zef such that it has lots of dependencies to fetch/extract/test (each dependency launches a process for each phase - fetch, extract, test) 17:38
Segmentation fault: 11
always that
the stage that it happens seems to change if I delete precomp, but if i rerun a failed run it will segfault in the same spot 17:39
i assume it has to do with handles somehow... what should I try to get some more info? 17:40
my wild guess is it could be related to all the handle/proc workarounds in zef are leaving handles open or something 17:43
timotimo Semantics Fault 18:07
ugexe windows proc regression: `perl6 -e "shell qq|perl -E \"say 1\"|;"` 19:04
seems to be related to argument passing, since anything without a space works 19:05
in the -E "<here>"
otherwise its 19:06
Can't find string terminator '"' anywhere before EOF at -e line 1.
The spawned command 'perl -E "say 1"' exited unsuccessfully (exit code: 255)
geekosaur that sounds to me more like CMD.EXE or the C runtime that mangles the passed command line into something programs can deal with doesn't handle \" 19:07
timotimo wasn't there just a fix for that? 19:11
geekosaur I had thought so and was trying to recall if so... 19:13
ugexe a fix for proc::async yeah, but from what i understand that fixes some things and breaks others
but non-async run/shell handled both cases in 2017.05 19:14
geekosaur um, that would not be related 19:15
hm, or maybe it is.
oh, no, we don't do `` any more so that was metaquoting 19:16
does that invocation work at a cmd prompt?
ugexe perl -E "say 1"? yes 19:22
geekosaur I meant the nested one 19:23
oh wait, that qq|| does affect things 19:24
(I hate that syntax specifically because it buries the quoting...)
ugexe well im confused - i showed the error message
i ran that command on 2017.05 and 2017.06, and it doesn't work on 2017.06
geekosaur yes, what I am seeing is something stopped at " ... "
ignoring that one thinks \" should not be the closing " 19:25
ugexe it doesn't work inside a script as `shell('perl -E "say 1"')` either 19:26
which is what Inline::Perl5 does in its Build.pm
im just quoting that way so it can be pasted as a one liner
github.com/nodejs/node-v0.x-archiv...t-10549609 19:45
TimToady just used .tree in a program a few weeks ago, and would be sad if it went away 20:52
lizmat TimToady: could you document its function ? 20:53
brrt has some trouble figuring out what a method called .tree does
lizmat that would help preserve it :-)
.oO( save the trees! )
rightfold m: say [1, 2, 3].tree.perl.say
camelia $((1, 2, 3).Seq)
True
timotimo i didn't know we kept tree, but i guess i can see its lasting appeal :) 20:54
TimToady I used it on a tournament structure, where we have three-dimensional arrays of rounds, locations, and teams 20:55
to make a double round robin, you use: 20:56
if $double { @rounds.append: @rounds.tree(*.self, *.reverse, *.reverse) }
that reverses the locations and the teams without reversing the order of the rounds 20:57
basically, it's a method for when each level of a tree structure want to have its own filter 20:58
timotimo whoa, i did not know it does that? 20:59
TimToady m: say (((1,2,3),(4,5,6)),((7,8,9),(10,11,12))).tree(*.self, *.reverse) 21:04
camelia (((4 5 6) (1 2 3)) ((10 11 12) (7 8 9)))
TimToady m: say (((1,2,3),(4,5,6)),((7,8,9),(10,11,12))).tree(*.self, *.reverse, *.rotate)
camelia (((5 6 4) (2 3 1)) ((11 12 10) (8 9 7)))
TimToady m: say (((1,2,3),(4,5,6)),((7,8,9),(10,11,12))).tree(*.self, |(*.reverse xx 10)) 21:05
camelia (((6 5 4) (3 2 1)) ((12 11 10) (9 8 7)))
TimToady m: say (((1,2,3),(4,5,6)),((7,8,9),(10,11,12))).tree(*.self, |(*.reverse xx *))
but that ooopsies 21:06
camelia (timeout)
brrt wow. i wonder when the perl6 community will wrap its head arround that one
dogbert17 thought it was a remnant from earlier times 21:07
doc'ing this will be interesting 21:12
lizmat gets shivers that remind her of her APL days 21:14
dogbert17 m: (1, (2, 3, 5), 4).tree(*.self, *.map({$_ + 1})).say # test 21:21
camelia (1 (3 4 6) 4)
dogbert17 interesting 21:22
Geth rakudo/nom: 327c84098a | (Elizabeth Mattijsen)++ | src/core/Str.pm
Make Str.subst(Str,Str) upto 3x faster

  - based on "foo".subst("o","z")
  - upto 3x faster for Str,Str (substitute only once)
  - upto 2x faster for Str,Str,:g (all possible substitutions)
Gains caused by creating a Str:D,Str:D candidate that catches all nameds and then checks for :g/:global using nqp::ops, instead of always trying to unpack all possible nameds.		 21:30
timotimo sweet 21:32
json::fast probably already uses nqp::subst directly
lizmat timotimo: there's an nqp::subst ?? 21:33
m: use nqp; nqp::subst()
camelia ===SORRY!===
No registered operation handler for 'subst'
timotimo oh, then it must use the method form and benefit greatly :)
tadzik aww yiss :) 21:43
timotimo well, maybe 21:44
i don't remember my last benchmark
so don't know how high subst ranks
dogbert17 repos/Text-CSV6/lib/Text/CSV.pm: $t .= subst (/( $q | $e )/, { "$e$0" }, :g); 21:49
repos/Text-CSV6/lib/Text/CSV.pm: $t .= subst (/ \x[0] /, { $e ~ 0 }, :g) if $!escape_null;
Geth rakudo/nom: 19be872211 | (Elizabeth Mattijsen)++ | src/core/Rakudo/Internals.pm
Optimize TRANSPOSE(-ONE) a bit further

At the moment, having a native str in a signature prevents the method from being JITted. So changed the sig to Str:D's instead, and get about a 10% improvement in Str.subst(Str,Str).		 21:50
dogbert17 the above is what [Tux] uses
lizmat dogbert17: yeah, that's not Str,Str :-) 21:51
dogbert17 oh no :-)
I remember that he uses split quite a bit, but that has already been optimized no? 21:52
lizmat dogbert17: split on multiple strings, yes :-) 21:53
dogbert17 repos/Text-CSV6/lib/Text/CSV.pm: ?? $str.split (@re, :v, :skip-empty) 21:55
repos/Text-CSV6/lib/Text/CSV.pm: for $spec.split (";") -> $r {
so the first one is covered then
timotimo wow 21:59
lizmat dogbert17: the second as well, as that is a simplified case of the other 22:04
dogbert17 cool 22:06
travis-ci Rakudo build passed. Elizabeth Mattijsen 'Make Str.subst(Str,Str) upto 3x faster 22:07
travis-ci.org/rakudo/rakudo/builds/244319041 github.com/rakudo/rakudo/compare/e...7c84098aec
eveo perl6-debug-m complains --nqp-lib=blib is an invalid option. I see perl6-gdb-m has the same option and it works fine. Any idea why debug-m don't like it? 22:08
yoleaux 06:30Z <nine> eveo: Inline::Perl5's tar file should be fixed in 0.28
timotimo does it come in too late? 22:13
i'm not even sure what --nqp-lib was added for
eveo looks to be in the same place as in perl6-gdb-m. In fact, if I change perl6-debug.moarvm to perl6.moarvm, like it's in gdb one, it works. Gonna see what makes perl6-debug.moarvm; probably different path in compiler that don't take that option or something 22:15
timotimo i see there's a little piece somewhere that adds that 22:16
i'll make it compatible 22:17
eveo Yeah. I just copied it
timotimo ah
i would have done the same
i wonder why it isn't inside Perl6::Compiler though
or maybe the HLL::Backend?
eveo ¯\_(ツ)_/¯ 22:18
Geth rakudo/nom: ee6e5df466 | (Zoffix Znet)++ | src/perl6-debug.nqp
Remove trailing whitespace
 22:28
rakudo/nom: 1ab21744e7 | (Zoffix Znet)++ | src/perl6-debug.nqp
Make perl6-debug take --nqp-lib option

Same as normal perl6
travis-ci Rakudo build passed. Elizabeth Mattijsen 'Optimize TRANSPOSE(-ONE) a bit further 22:35
travis-ci.org/rakudo/rakudo/builds/244322930 github.com/rakudo/rakudo/compare/3...be87221160
eveo hah, debugger is pretty damn cool. 23:32
eveo is using it for the first time now :)
Geth rakudo/nom: 6d4691fbd9 | (Zoffix Znet)++ | 2 files
Unbust the perl6-debug-m debugger

The .define_slang stuff seems to be what's busting it. Unsure why. The breakage appears to occur whenever strings are used (so something to do with the switch to Quote lang?). For example -e 'say 42' won't crash, but -e 'say "x"' will. Even -e '"x"' crashes. Looking at the AST of the last one, shows (NQPMu) in the tree, where the working debugger's AST shows a QAST::Want for the WVal(Str)/SVal.		 23:44
eveo TimToady: ^ maybe you'll be able to make sense out of why .define_slang busts the debugger (or if the define_slang is even needed there)
eveo goes back to taking a break from Perl 6 :) 23:45

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!