timotimo i might be onto something. 00:05
done 00:09
spec test now
Geth rakudo/nom: 46b11f54c0 | (Timo Paulssen)++ | src/Perl6/Actions.nqp
bring back for ^N { } optimization again
 00:14
eveo \o/ 00:16
timotimo++
timotimo it'll only take a minute before the whateverables pick that commit up and i can bench it "officially"
c: 46b11f54c0 say "i'm here" 00:19
committable6 timotimo, ¦46b11f5: «i'm here»
timotimo bench: 46b11f54c0,abfb52be1d for ^1_000_000 { }
benchable6 timotimo, starting to benchmark the 2 given commits
timotimo, benchmarked the given commits, now zooming in on performance differences
timotimo, ¦46b11f5: «0.2715» ¦abfb52b: «0.5507»
timotimo it's not the 2.5x we had before 00:20
it's 2x, though
maybe it's too little to accurately measure
bench: 46b11f54c0,abfb52be1d for ^1-_000_000 { }
benchable6 timotimo, starting to benchmark the 2 given commits
timotimo, benchmarked the given commits, now zooming in on performance differences
timotimo, ¦46b11f5: ««run failed, exit code = 1, exit signal = 0»» ¦abfb52b: ««run failed, exit code = 1, exit signal = 0»»
timotimo nooooo
bench: 46b11f54c0,abfb52be1d for ^10_000_000 { }
benchable6 timotimo, starting to benchmark the 2 given commits
timotimo, benchmarked the given commits, now zooming in on performance differences 00:21
timotimo, ¦46b11f5: «1.6712» ¦abfb52b: «4.4582»
timotimo m: say 4.4582 / 1.6712
camelia 2.667664
timotimo that's the factor i was hoping for
AlexDaniel timotimo: fwiw, you can also push stuff into a branch 00:22
timotimo: and whateverables will pick it up too
timotimo: you'll have to address the commit by its hash though (not by branch name)
timotimo resolved the ticket 00:24
thanks, AlexDaniel
we can now get less noisy measurements from simple for loops like that again
AlexDaniel \o/ 00:25
samcv anyone else getting failures on GraphemeClusterBreak.t? 01:09
eveo said he was, but i'm not getting any. unless i forgot to bump something. but i remember bumping nqp and moarvm hm.
yeah i did bump 01:10
timotimo t/spec/S15-nfg/GraphemeBreakTest.t (Wstat: 0 Tests: 2411 Failed: 0) 01:12
TODO passed: 639-640, 2363-2367, 2369-2371, 2382, 2384-2386
2395, 2397
eveo samcv: that's not the file I referenced. The failures are in 6.c-errata 01:33
#perl6 <yoleaux> 12 Jun 2017 18:48Z <eveo> samcv: on 6.c-errata t/spec/S15-nfg/grapheme-break.t now fails 3 Regional_Indicator tests. Are the tests wrong? 01:34
eveo returna to sleeping
ugexe bisectable6: my $list := eager gather LABEL: for 1..4 -> $index { take $index; next LABEL; }; say $list.perl 03:32
bisectable6 ugexe, Bisecting by exit code (old=2015.12 new=46b11f5). Old exit code: 0
ugexe, bisect log: gist.github.com/704d8c9ddfc94e22b5...3379336683
ugexe, (2017-06-13) github.com/rakudo/rakudo/commit/46...c9c78623c7
ugexe timotimo: ^ 03:33
AlexDaniel well, at least it is fast :) 03:45
[Tux] This is Rakudo version 2017.05-405-g46b11f54c built on MoarVM version 2017.05-76-gba6725f3 06:21
csv-ip5xs 2.834
test 12.995
test-t 4.341 - 4.359
csv-parser 13.272
timotimo ugexe: i don't get it, the exit code and output are all exactly identical for every step in the bisect log? 07:43
m: my $list := eager gather LABEL: for 1..4 -> $index { take $index; next LABEL; }; say $list.perl 07:44
camelia labeled next without loop construct
in block <unit> at <tmp> line 1
timotimo i should be able to figure out the labeledness of the for and bail if it exists 07:45
that was probably also broken a long time ago when that opt was still working 07:46
c: releases my $list := eager gather LABEL: for 1..4 -> $index { take $index; next LABEL; }; say $list.perl
lizmat is looking at RT #131561 07:50
yoleaux 12 Jun 2017 22:53Z <eveo> lizmat: toaster found a regression in .MixHash coercer. Would you know how to fix this? rt.perl.org/Ticket/Display.html?id...xn-1465951
timotimo commit: releases my $list := eager gather LABEL: for 1..4 -> $index { take $index; next LABEL; }; say $list.perl 07:51
i killed committable? :( 08:00
Geth rakudo/nom: bf95bcb6c0 | (Elizabeth Mattijsen)++ | src/core/Rakudo/QuantHash.pm
Fix for RT #131561

R:Q.ADD-PAIRS-TO-MIX no longer has a SANITY step that weeded out the 0 weight elements before. Instead, now do this on the fly. Perhaps a final iteration on the resulting hash would be more efficient in some cases, but this feels better from a memory usage point of view.		 08:10
timotimo m: my @in = 1, 2, 4, 8; do LABEL: for @in -> $index { say $index; next LABEL; }; 08:15
camelia 1
labeled next without loop construct
in block <unit> at <tmp> line 1
timotimo ^- golfed, doesn't include a range any more
also, the optimization already doesn't occur when there's a label
lizmat timotimo++ *phew* :-) 08:18
hmmm... 08:19
timotimo: yesterday I did a base run of a piece of code: "my $s = ^1000 .Set; for ^100000 { my $b }"
at ran for 200 msecs, now it runs 1017 msecs :-(
oops, scratch that 08:20
fluke
now it still runs at 200 msecs
timotimo phew 08:26
but what does that even do? call .sink on that set a bunch of times?
er, "my $b"
that has nothing to do with $s at all?
lizmat indeed, it's the base run 08:27
a real run would have been something like "my $s = ^1000 .Set; for ^100000 { my $b = 42 ∈ $s }"
Geth rakudo/js: 66 commits pushed by (Zoffix Znet)++, (Julien Simonet)++, MasterDuke17++, (Stefan Seifert)++, (Elizabeth Mattijsen)++, (Samantha McVey)++, pmurias++
review: github.com/rakudo/rakudo/compare/f...ccac70c9c9		 08:32
nqp/master: 6 commits pushed by pmurias++ 08:33
lizmat timotimo: and I'm not really awake yet, the base run is now 176 msecs, down from 200
I hadn't recompiled yet :-(
DrForr I'll have to rerun my Perl6::Parser test suite and check out improements. 08:37
Geth roast: 4e394a9410 | (Elizabeth Mattijsen)++ | 2 files
Add tests for RT #131561
 08:43
lizmat afk& 08:44
travis-ci NQP build failed. pmurias 'Test that we have the correct ctx in exit_handler' 08:58
travis-ci.org/perl6/nqp/builds/242329563 github.com/perl6/nqp/compare/cabb9...2c044c433c
tadzik clearly we don't %) 09:03
stmuk msg? 09:23
yoleaux 12 Jun 2017 17:50Z <eveo> stmuk: you have a PR in this repo. Also, looks like the Issues tab is disabled: github.com/stmuk/tap-harness6-prove6
stmuk eveo: thanks for useful reminder ... that module only existed as a work around because TAP was in core, can be deleted and I'll update R* 09:28
Geth rakudo/proc-using-proc-async: dc0e39b621 | (Jonathan Worthington)++ | 3 files
Allow chaining .out/.err of Proc to .in of another

Done a bit differently from how it was before, but hopefully this way also works out on Windows (the way Proc used to do this failed).		 09:35
rakudo/proc-using-proc-async: 45fd8fafb3 | (Jonathan Worthington)++ | tools/build/NQP_REVISION
Bump NQP_REVISION.

To ensure we have all MoarVM and JVM Proc::Async improvements.		 09:37
jnthn spectests the merge
Geth star: 2ff3371bc9 | (Steve Mynott)++ | 4 files
Now TAP isn't in core remove tap-harness6-prove6 & replace with tap-harness6
 09:40
rakudo/nom: 15 commits pushed by (Jonathan Worthington)++
review: github.com/rakudo/rakudo/compare/b...ac31c5df46		 09:43
¦ rakudo/nom: version bump brought these changes: github.com/perl6/nqp/compare/2017....9-g9a2c044
jnthn There we go 09:47
jnthn tries to find the toaster URL 09:48
Latest MoarVM commit should unburn Log::Async 09:54
llfourn jnthn++ 09:55
toast.perl6.party/
jnthn Yeah, found it :)
One of those cases where I knew what patch to write but wanted to find a module that needed it :) 09:56
Just to be sure the fix did what I expected :)
llfourn mm this toasting is pretty handy for testing 09:57
jnthn Yeah, nice to be able to easily see what got busted since last release 09:58
eveo buggable: speed 4
buggable eveo, ▂▁▁▁ data for 2017-06-12–2017-06-13; range: 4.341s–4.427s; 2% faster
llfourn rebuilds rakudo to see if Proc::Async changes has fixed his OSX issues 10:00
jnthn llfourn: What was the issue, ooc? 10:01
llfourn looks for RT
jnthn: rt.perl.org/Public/Bug/Display.html?id=122722 10:03
or at least that's the reference I put in my code 10:04
I needed to put a sleep 0.1 to make it work: github.com/spitsh/spitsh/blob/mast...i.pm6#L445
jnthn Hm, I thought that issue was long fixed...
llfourn hmm could be 10:05
jnthn github.com/rakudo/rakudo/blob/nom/...nc.pm#L258 10:06
Note the @!promises there 10:07
llfourn hmm well it looks like it still a bug
because I removed that sleep and it no longer works
jnthn So the fix seems equivalent to the one proposed in the patch
oh 10:08
But that's a write/close race
Is $docker an async proc there?
llfourn yeah
jnthn Try changing
$docker.write($shell.encode('utf8'));
to
await $docker.write($shell.encode('utf8'));
llfourn that worked :o 10:09
thanks!
jnthn :)
llfourn I didn't realise .write returned a promise
\o/
jnthn But that RT I think can be closed as it's a differnet problem and we have a fix in place
nwc10 for me, the build on master/master/nom is broken 11:08
explosions of the form
Internal error: inconsistent bind result at gen/moar/BOOTSTRAP.nqp:3319 (blib/Perl6/BOOTSTRAP.moarvm:)
and
Unhandled exception: Dynamic variable $*ERR not found
pmurias is building a jvm nqp to fix the nqp-j test breakage... 11:09
samcv .tell eveo the tests may me wrong (probably are) I will check on them tomorrow :) 11:40
yoleaux samcv: I'll pass your message to eveo.
eveo nwc10: are you sure you did the full build and not just make install?
yoleaux 11:40Z <samcv> eveo: the tests may me wrong (probably are) I will check on them tomorrow :)
eveo like perl Configure something something 11:41
nwc10 full build from clean 11:44
Geth roast: 969dc08c1e | (Elizabeth Mattijsen)++ | S03-operators/subset.t
Add tests for Map vs Map for the (<=) family
 11:55
rakudo/nom: e1563a7644 | (Elizabeth Mattijsen)++ | src/core/set_operators.pm
Fix various issues with Map (<=) Map

The truthiness of the values in the Maps weren't checked, when they should have been, because they are checked when coercing to a .Set		 11:57
lizmat m: dd { a => 0 } (<) { a => 1 } 12:01
camelia Bool::False
lizmat m: dd { a => 0 }.Set (<) { a => 1 }.Set # more work to do
camelia Bool::True
eveo :/ just got a JSON error when viewing ecosystem-api.p6c.org/projects.json 12:02
Looks like the atomic write thing isn't atomicing
lizmat jnthn: spectest clean on MacOS (apart from passing GraphemeBreakTests) 12:04
eveo or maybe it's something else; the update runs at 48th minute, so it'd be weird if it finished on 02
m: $*PERL.compiler.version.say 12:08
camelia v2017.05.421.gac.31.c.5.df.4
eveo nwc10: sound like you're building the wrong thing. Off an unupdated fork or something. git pull and build works for me. git clone github.com/rakudo/rakudo/ .; perl Configure.pl --gen-moar --gen-nqp --backends=moar; make; make test; make install; works for me. And I see the doc travis run also built the latest commit and other than some ecosystem glitching dthe rakudo built fine: 12:10
travis-ci.org/perl6/doc/jobs/242368688
jnthn nwc10: Odd, I've no idea what that could mean. :S 12:13
nwc10 seems to be working on the new commit from lizmat 12:14
eveo Something's busted with ecosystem. doc fails to install modules and my trying to get Inline::Perl5 installed dies with: ===> Extracting: File::Which
Extracting with plugin: Zef::Service::Shell::git+{<anon|89153280>}
===> Extraction [OK]: File::Which to /home/cpan/.zef/store/perl6-file-which.git
nwc10 might *just* be a parallel build race
eveo No meta file? Path: /home/cpan/.zef/store/Inline-Perl5.git/e1563a764462da392c993ee65b538877b423908e
nwc10 but no, does not make sense
eveo hm... well, it's not lying. There is in fact no meta file in that dir :/ 12:16
jnthn lizmat: Glad to hear it's clean for you also :D 12:17
eveo Not clean for me :) It's failing to install Inline::Perl5, so I can't run full stresstest :) 12:18
jnthn news.perlfoundation.org/2017/06/gra...erl-6.html # for anyone keen to see me spending lots of time on Perl 6 during the summer :)
lizmat jnthn: noted for next weeks P6W 12:20
nwc10 no, still goes boom
CORE.d.setting.moarvm 12:21
fails with a lot of backtrace, that starts
Stage parse : Internal error: inconsistent bind result
(well, fails during stage parse)
lizmat afk for a while&
jnthn nwc10: That's really odd, I can't understand how that could happen but only locally 12:22
for you, and not for everyone
nwc10 yes, it is most strange
jnthn Can you paste me the backtrace?
nwc10 paste.scsys.co.uk/564428 12:23
jnthn What on earth... 12:24
Geth nqp: 8856086363 | pmurias++ | src/vm/js/nqp-runtime/sixmodel.js
[js] Fix MODE_BIGINT with type object
 12:25
nqp: 1be5e4889c | pmurias++ | t/nqp/071-setboolspec.t
Test MODE_BIGINT with type object
nqp: a33cbeb3f4 | pmurias++ | src/vm/jvm/runtime/org/perl6/nqp/runtime/Ops.java
[jvm] Implement nqp::captureposarg_{i,n,s}
timotimo eveo: you can reboot committable, right? 12:28
eveo yeah 12:31
c: HEAD say 42
committable6 eveo, ¦HEAD(e1563a7): «42»
eveo k, can't install any module 12:35
eveo nukes all and retries 12:37
lizmat eveo: something like No meta file? Path: /Users/liz/.zef/store/perl6-LN.git/e1563a764462da392c993ee65b538877b423908e 12:39
?
eveo lizmat: yup.
lizmat ok, confirmed here then
eveo I tried a bunch of modules, all end with that error
lizmat feels like jnthn's merge may be the cause ? 12:40
afk again&
eveo something's really wrong 12:44
lizmat back again 12:45
but that wrongness is *not* caught with make test nor make spectest (if Inline::Perl5 is installed)
eveo Here's part of output of my update-perl6 script that nukes everything and installs everything from scratch. First TAP::Harness test fails saying there was no, plan and then at the bottom it fails with "labeled next without loop construct": gist.github.com/zoffixznet/36c0fc4...41ce7ca934
lizmat feels like mangled precomp files ? 12:46
eveo re-runs update-perl6 to see if it spews the same errors
update-perl6 == rm -fr ~/.zef; rm -fr ~/.perl6; rm -fr ~/.rakudobrew/; git clone github.com/tadzik/rakudobrew ~/.rakudobrew; rakudobrew build moar; rakudobrew build zef; zef install TAP::Harness; zef --serial install Term::ANSIColor File::Temp Slang::Tuxic Testo Benchy 12:47
The answer is Yes. 12:52
travis-ci NQP build passed. pmurias '[jvm] Implement nqp::captureposarg_{i,n,s}' 12:53
travis-ci.org/perl6/nqp/builds/242398864 github.com/perl6/nqp/compare/9a2c0...3cbeb3f4b1
Geth rakudo/nom: 8d6a9114cb | (Zoffix Znet)++ (committed using GitHub Web editor) | src/core/IO/Pipe.pm
Fix typo in error messages
 12:55
eveo spies a change in CompWhatever visavis stderr handling 13:02
hm... but it likely not it, 'cause I've just built bf95bcb and installing Inline::Perl5 still dies, though now with the "labeled next without loop construct" error instead of "no meta?" 13:04
jnthn Darn, and all the spectests didn't catch this? :/ 13:07
eveo yeah 13:08
jnthn Oh, even more odd, I just did zef install . of some module (actually Log::Async that I fixed the precomp of earlier) and it just worked 13:09
jnthn runs zef install Testo, which we sure doesn't have installed yet
ah, and that reproduces it
While it was searching for a prereq... 13:10
eveo I see a change in this diff that old version did '-' if $RMD and piped if not for STDERR, but new version it .notes the STDERR stuff instead of printing if $RMD is there and doesn't collect $err at all. Building now with change but doubt it'll fix antyyhing: github.com/rakudo/rakudo/compare/b...6ed9cc28a4
jnthn eveo: Doesn't collecting it = it just inherits STDERR 13:13
eveo Um. Well, it does with the loop thing now instead of meta
*dies
jnthn I think the only thing for it will be to try and bisect it 13:15
eveo no $RMD => ($proc.err.slurp-rest(:close)) | (.note it)
$RMD => (inherit) | (inherit)
^ and the .note it thing should be $err ~= $_, no?
jnthn No, the inherit case is when it's not RMD 13:16
oh, han gon
darn 13:17
Grr, I think I mis-refactored it
Yeah, it should append to $err in there 13:18
eveo jnthn: ok, then I'll leave it with you. I gotta go do some things
eveo &
lizmat really afk& 13:24
jnthn I also managed to trigger inconsistent bind result when building an older commit... 13:28
ohh 13:35
I think that one was due to a commit in MoarVM master but where a revision bump wasn't done 13:36
Thus why only building master/master/nom showed it up 13:37
And it managed to break sufficiently little you could even spectest and stuff
Ooh, and I think I may be off the hook for the zef failure 13:42
I think 46b11f54c03511a96 is guilty 13:43
The commit before it seems to work, on that commit it busts
Trying on nom but with a revert of that 13:44
Yup, it works 13:46
Geth rakudo/nom: 4aa60ad040 | (Jonathan Worthington)++ | src/Perl6/Actions.nqp
Revert "bring back for ^N { } optimization again"

This reverts commit 46b11f54c03511a96d4db08213d55bc9c78623c7, which appears to break labeled loops (which in turn breaks `zef`; a little cocerning that we didn't fail a spectest, however).		 13:47
jnthn timotimo: See ^^, I dunno if it's obvious what's going wrong 13:49
travis-ci Rakudo build failed. Zoffix Znet 'Fix typo in error messages' 13:59
travis-ci.org/rakudo/rakudo/builds/242408838 github.com/rakudo/rakudo/compare/e...6a9114cb79
buggable [travis build above] ☠ Did not recognize some failures. Check results manually.
eveo "Internal error: inconsistent bind result" 14:00
jnthn Already fixed at MoarVM HEAD 14:01
eveo gonna bump then 14:02
jnthn But...the previous NQP_REVISION/MOAR_REVISION was *before* the MoarVM commit that broke things 14:04
eveo Ah OK
jnthn But yeah, can bump 'cus it fixes Log::Async :) 14:05
eveo OK, then I'll bump
timotimo jnthn: huh, but that optimization bails out when it sees a labeled loop already 14:11
jnthn timotimo: I wonder if the QAST construction busted?
Does it push the label into the right place?
timotimo committable6: releases my @in = 1, 2, 4, 8; do LABEL: for @in -> $index { say $index; next LABEL; };
committable6 timotimo, gist.github.com/59675e08c20984069d...7e84dacc46
timotimo look, it has been busted for like ever 14:12
eveo 0.o
bisect: my @in = 1, 2, 4, 8; do LABEL: for @in -> $index { say $index; next LABEL; }; 14:13
bisectable6 eveo, Bisecting by exit code (old=2015.12 new=4aa60ad). Old exit code: 0
eveo, bisect log: gist.github.com/fde16d8cf2219f7da2...654f9c4db2 14:14
eveo, (2016-07-04) github.com/rakudo/rakudo/commit/55...cf60ae3bb2
timotimo can you golf the case that's used inside zef so we can see my commit actually break ti? 14:16
because clearly something was working before that no longer works
ugexe zef basically just does 14:17
eager gather LABEL: for 1,2,3 -> $index { take $index; for 1,2,3 { next LABEL; }; };
the label being able to "next" the outter loop 14:18
timotimo committable6: releases eager gather LABEL: for 1,2,3 -> $index { take $index; for 1,2,3 { next LABEL; }; };
committable6 timotimo, ¦releases (19 commits): «»
timotimo ah, that's because latest fixes it again
eveo releqases only does releases don't it? 14:19
timotimo committable6: 46b11f5 eager gather LABEL: for 1,2,3 -> $index { take $index; for 1,2,3 { next LABEL; }; };
committable6 timotimo, ¦46b11f5: «labeled next without loop construct␤ in block at /tmp/U0wxNiZQ7A line 1␤ in block <unit> at /tmp/U0wxNiZQ7A line 1␤ «exit code = 1»»
timotimo okay, let me look at it more closely locally 14:20
it's not conditional on the optimizer running
eveo c: HEAD,HEAD~1 gather LABEL: for 1 {next LABEL} 14:21
committable6 eveo, ¦HEAD(4aa60ad): «» ¦HEAD~1: «labeled next without loop construct␤ in block <unit> at /tmp/3TZgREii9s line 1␤ «exit code = 1»»
eveo timotimo: ^ golfed a bit more
timotimo well, it's appending a label argument to the sink op
so that's wrong, but probably not the cause of the splosion 14:22
eveo Filed the other case that's been broken for a year as rt.perl.org/Ticket/Display.html?id...xn-1466171 14:26
hm, actually looking at it now seems to be the same cause 14:27
c: HEAD,HEAD~1 say (LABEL: for 1 {next LABEL})
committable6 eveo, ¦HEAD(4aa60ad),HEAD~1: «labeled next without loop construct␤ in block <unit> at /tmp/Mlcm95ifMC line 1␤ «exit code = 1»»
eveo c: HEAD,HEAD~1 @ = (LABEL: for 1 {next LABEL})
committable6 eveo, ¦HEAD(4aa60ad),HEAD~1: «labeled next without loop construct␤ in block <unit> at /tmp/0TVgWE1zL_ line 1␤ «exit code = 1»»
timotimo hmm. how about a "coverable" bot that just spits out the lines that were covered by a snippet you ran? 14:28
ugexe could just add a `use trace` to the snippet automatically 14:32
eh you're probably talking about something more useful, code covered in rakudo or something 14:34
eveo m: use nqp; my $i = 1;nqp::while($i--, nqp::handle(next, 'LABELED', 'meow')) 14:48
camelia ( no output )
eveo m: use nqp; my $i = 1;nqp::while($i--, nqp::handle(next, 'LABELED', 'meow'), :nohandler)
camelia ===SORRY!===
next without loop construct
eveo Thought I tried this thing already....
eveo re-tries
Ah, no it's not it 14:50
m: use nqp; my $i = 1;nqp::while($i--, nqp::handle(next, 'LABELED', 'meow', 'NEXT', say("Hi")), :nohandler) 14:51
camelia Hi
eveo that works fine
hoping to find this juicy bug before lunch 14:56
ZofBot: no moar dogfood for eveo! Bug sammich \o/
ZofBot eveo, cry
travis-ci Rakudo build passed. Jonathan Worthington 'Revert "bring back for ^N { } optimization again" 15:07
travis-ci.org/rakudo/rakudo/builds/242428275 github.com/rakudo/rakudo/compare/8...a60ad040bf
[Coke] 'zef info zef' shows Identity: zef:ver('0.1.7'):auth('github:ugexe') 15:14
zef --verbose upgrade zef shows: The following distributions are already at their latest versions: zef:ver('0.1.16'):auth('github:ugexe')
is it comparing those versions as "7" gt "16" ? 15:15
(is my zef too old to make my zef not too old?)
eveo m: say 0.1.7 before 0.1.16 15:16
camelia 5===SORRY!5=== Error while compiling <tmp>
Malformed postfix call
at <tmp>:1
------> 3say 0.1.7⏏057 before 0.1.16
eveo m: say v0.1.7 before v0.1.16
camelia True
eveo don't think I have a box with zef older than 0.1.16.... zef info zef tells me 0.1.16 15:18
[Coke] eh, I'll force it this once.
ugexe++ love that 'zef info zef' points me at the bugtracker. 15:20
ugexe [Coke]: i try to reproduce that problem here github.com/ugexe/zef/issues/172 but fail 15:29
i show how to go from no zef, to 0.1.15, to upgrading to 0.1.16 15:30
i'm not sure how the zef version would cause the issue, and rakudo wise the only thing i can think of is maybe an older version is not invaliding precomp properly? 15:31
[Coke] 2017.05-315-g160de7e6f 15:33
ugexe are you able to reproduce from a clean state? if i could just reproduce it myself...
[Coke] I can't reinstall zef ...7 at this point, can it? 15:34
*I?
ugexe yeah, one second
eveo oohhh
[Coke] ... eh, I suppose I could if I used github. 15:35
eveo I think I see the bug for the label thing
ugexe `zef install github.com/ugexe/zef/archive/v0.1.7.tar.gz` should do it
[Coke] but also, since I can't have a clean zef (it's always touching ~/.perl6, even if I have a rakudobrew in one place, and a separate rakudo install elsewhere, right?)
I also am on OS X, fwiw. 15:36
ugexe technically ~/.perl6 should be separated between compiler versions
yeah I tested on OSX as well
[Coke] perl Configure.pl --backends=moar --gen-nqp --gen-moar # error: pathspec '2017.05-152-gc8916bc5a' did not match any file(s) known to git. 15:39
... whoops, js branch.
eveo Dammit, Imma blind. The bug was right THERE all this time and I didn't even notice it 15:47
And I think I saw it before and mentioned it on some ticket already 15:48
yup. compiled and confirmed the fix works. Gonna rake through the codebase for any more instances of it, write tests, commit, and then revert back that for ^N{} perf opt 15:52
w00t found this bug's cousin hiding a hundred lines avove :) 16:05
[Coke] ugexe: if I'm starting from a clean slate, how can I use 'zef install <old version of zef' ? 16:07
wouldn't I install the old version *first* ?
eveo cd $(mktemp -d); git clone github.com/ugexe/zef/ .; git checkout v0.1.7; perl6 -Ilib bin/zef install . 16:08
[Coke] right. 16:10
(thank you for giving me a cut-and-paste!) 16:12
ugexe: same error. 16:13
eveo Any idea how to trigger this method? github.com/rakudo/rakudo/blob/nom/...ds.pm#L326
[Coke] I have zef info zef showing 0.1.7, and zef upgrade zef telling me it's already at the latest version (0.1.16)
eveo It's got the `next` bug in it and I'm trying to cover it, but can't figure out when it would be called
m: (FOO: for ^10 -> $_ --> Int { .say; next FOO; 42 }).iterator.sink-all 16:14
camelia 0
labeled next without loop construct
in block <unit> at <tmp> line 1
eveo ^ that ain't it
[Coke] ugexe: this with rakudo 2017.05-424-g4aa60ad04
so, let me know what I can run to get more debug output, etc.
eveo This seems to be it: FOO: ^10 .map(-> $_ --> Int { .say; next FOO; 42 }).iterator.sink-all 16:15
huh, nope. Not it. But it's at least hitting the 4th bug that's still somewhere there 16:19
Geth rakudo/nom: b0a1b6c31c | (Jonathan Worthington)++ | src/Perl6/Actions.nqp
Analyze methods for "can never use return" also.

This means that we can generate simpler code for such methods, which in turn makes many more methods possible to inline.		 16:26
jnthn m: say 2.579 / 2.767 16:27
camelia 0.932056
eveo k, it does seem to be it, but it doesn't wanna fix the same way as other spots
jnthn m: say 2.579 / 2.873
camelia 0.897668
jnthn m: say 2.579 / 1.134
camelia 2.274250
jnthn m: say 2.873 / 1.134 16:28
camelia 2.533510
jnthn Those are, respectively, speed gained on a for lines benchmark from the commit above, overall gain today with my commits over in MoarVM, how much faster Perl 5 is at us than this (assuming we do UTF-8 decoding and chomp), and how much faster it was than us before I did this work 16:29
eveo sweet 16:30
jnthn++
jnthn Takeaway: `my $fh = open "longfile"; my $chars = 0; for $fh.lines { $chars = $chars + .chars }; $fh.close; say $chars` runs in 90% of the time it used to after this work
Though I should also take a measurement from before my I/O refactors 16:31
Aww, it was actually faster before the refactors 16:35
2.421
But (a) we can still win that back easily, and (b) we're able to support user-space encodings now 16:36
Not that we implemented the latter
But the layer of indirection is there
eveo Ah, ok, it's crashing because `FOO: ^10 .map(-> $_ --> Int { .say; next FOO; 42 }).iterator.sink-all` is not a way to add a lable to that block 16:44
Still trying to hit this line: github.com/rakudo/rakudo/blob/nom/...ds.pm#L326 16:45
m: FOO: for ^3 { ^10 .map(-> $_ --> Int { .say; next FOO; 42 }).iterator.sink-all } 16:46
camelia 0
0
0
eveo dammit
That hits that block, but not the next codepath because it's not the right block that's labeled 16:47
oh 16:49
m: FOO: for ^10 -> $_ --> Int { .say; next FOO; }
camelia 0
labeled next without loop construct
in block <unit> at <tmp> line 1
eveo yey
So 4 bugs in total. A whole family of them 16:51
ZofBot: yum nom nom nom. DELISHUS!
ZofBot eveo, Just passing it the original thing
ugexe [Coke]: you can perl6 -Ilib bin/zef install $url as well 16:55
eveo ZOFFLOP: t/spec/S17-supply/supplier-preserving.t 16:56
[Coke] ugexe: I can duplicate the error now; what more debug output do you need? 16:59
ugexe all i really need to know is the commands since I already know what im looking for 17:00
Geth rakudo/nom: 3b67b4acea | (Zoffix Znet)++ | src/core/Any-iterable-methods.pm
Fix 4 cases of crashes with labeled `next`

A bunch of nqp::handle are missing a branch for `NEXT`, so when we try to `next SOMELABEL` in for/map loops, we get a crash, complaining we're trying to do labeled next without a label.
Fix by adding npq::null NEXT handlers.
Fixes RT#131567: rt.perl.org/Ticket/Display.html?id=131567
eveo ugexe: did you say you tried upgradeing 0.1.16 to 0.1.17 and failed to reporudce the issue. That sounds like there was a bug in upgrade feature in 0.1.7...0.1.16 versions that got fixed 17:02
Geth roast: 9f418777bc | (Zoffix Znet)++ | S04-statements/for.t
Cover crashes in labeled next in for/map

Rakudo fix: github.com/rakudo/rakudo/commit/3b67b4acea RT#131567: rt.perl.org/Ticket/Display.html?id=131567
ugexe from 0.1.15 to 0.1.16 yeah, but this also happened with 0.1.9 for someone else and I made no changes that would affect it as such 17:03
eveo does the version bumps 17:04
[Coke] ugexe: so, I installed a fresh perl6 into ./install - I then installed zef 0.1.7 from a git checkout (as eveo suggested), then tried with that installed zef to run 'zef upgrade'. Do you need more details? 17:06
ugexe [Coke]: can you try `perl6 -Ilib bin/zef info zef` too? 17:07
that should test if its the bin wrapper
which does a version sort thing
[Coke] given bin/zef is relative, what is my starting directory?
ugexe git clone $zef, cd zef, perl6 -Ilib bin/zef info zef 17:08
[Coke] the git checkout at 0.1.7 ?
ok, a fresh clone at HEAD.
ugexe yeah
[Coke] perl6 -Ilib bin/zef info zef > HEAD.out; zef info zef > installed.out; diff HEAD.out installed.out # no diff 17:10
(both show 0.1.7 as the installed version) 17:11
ugexe grr. sort is acting funny in zef somewhere then, but not when i try to reproduce.
[Coke] where in zef is the code doing the compare? 17:12
I can try to golf it to something that fails on OS X.
$latest-installed.dist.ver cmp $candi.dist.ver looks like the snippet from 0.1.7 17:17
lizmat Zoffix++ 17:20
Geth nqp: 4717849a94 | (Zoffix Znet)++ | tools/build/MOAR_REVISION
Bump MoarVM

MoarVM bump brought commits:
  github.com/MoarVM/MoarVM/compare/2...5-g21ee1a5
21ee1a5 Fix missed opportunity to run JITted code. 2e9aff7 Keep named-used marking if binding can de-opt. ... (7 more lines)		 17:21
¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...5-g21ee1a5
rakudo/nom: 97298aca4e | (Zoffix Znet)++ | tools/build/NQP_REVISION
Bump NQP

NQP bump brought commits:
  github.com/perl6/nqp/compare/2017....3-g4717849
4717849 Bump MoarVM a33cbeb [jvm] Implement nqp::captureposarg_{i,n,s} ... (15 more lines)
rakudo/nom: version bump brought these changes: github.com/perl6/nqp/compare/2017....3-g4717849
06c2e220e5 | (Elizabeth Mattijsen)++ | src/core/set_operators.pm
ugexe [Coke]: ok i see what the issue likely is
.candidates (a "best" match search) does sort, but .search (what backs zef info) does not 17:24
not sure why the order always works out for me with .search, but thats gotta be it 17:25
eveo I'm still getting failures with module install right now. After the bump: gist.github.com/zoffixznet/578ffa8...b1f86faddc 17:35
Complains about built.
This is without for ^100 {...} opt brought in (which I'll bring in back right now). 17:36
ugexe doh, the build stuff is a big ball of fuck-all workarounds
lizmat fwiw, I'm also seeing zef install problems: no meta file? 17:37
ugexe what if you run the command shown (above the first SORRY)
eveo lizmat: that should be fixed now. Are you sure you built all the latest commits? 17:38
ugexe the no meta file thing makes me think run/shell is not handling :cwd() the same as before
eveo lizmat: or maybe I'm misremembering. There was a bunch of bugs... and some got fixed :)
lizmat eveo: $ git describe
2017.05-428-g06c2e22
my last install was including a configure
perl Configure.pl --gen-moar --gen-nqp --backends=moar --make-install 17:39
eveo ah, k, it's the loop thing and the binding error that got fixed.
ugexe eveo: in that gist what happens if you run the command shown above the first SORRY?
eveo 1 sec
or rather 120secs. I'm stresstesting ATM 17:40
ugexe also may want to try it inside of *and* outside of '/home/cpan/.zef/store/Inline-Perl5.git/693e5764168c64584cc74a639ead44f7b9205e73' as cwd 17:41
ugexe cant test anything for a few hours 17:42
Geth rakudo/nom: 6900078434 | (Zoffix Znet)++ | src/Perl6/Actions.nqp
Revert "Revert "bring back for ^N { } optimization again""

This reverts commit 4aa60ad040bf57425ea28cba1f391c1d0e05907f.
The bug that caused the revert of this commit was fixed in:
  github.com/rakudo/rakudo/commit/3b67b4acea
eveo ugexe: which command? The /home/cpan/CPANPRC/rakudo/install/bin/perl6 -Ilib -I. -MBuild ? It doesn't seem to be a proper command 17:43
ugexe yeah looks like cwd is not getting set on spawned procs. the SORRY error shows /rakudo/lib and /rakudo at the head of the repo chain, but those should be pointing at that Inline::Perl5 directory
eveo k, need to escape the last arg manually
ugexe its a proper command, you just need to wrap the 2 items at the end with '' (after the -e) 17:44
eveo s: &run'
SourceBaby eveo, Something's wrong: ␤ERR: ===SORRY!=== Error while compiling -e␤Unable to parse expression in argument list; couldn't find final ')' ␤at -e:6␤------> put sourcery( &run<HERE>' )[1];␤ expecting any of:␤ infix␤ infix stopper␤
eveo s: &run
SourceBaby eveo, Sauce is at github.com/rakudo/rakudo/blob/3b67...oc.pm#L158
eveo Yeah, good spot ugexe++ :cwd isn't passeed 17:45
s: Proc, 'spawn'
SourceBaby eveo, Sauce is at github.com/rakudo/rakudo/blob/3b67...oc.pm#L118
eveo Yeah, see it. Gonna fix in a sec 17:46
ugexe i can replicate to see thats true, but not seeing where cwd is actually not being passed in those links though 17:49
eveo in !spawn-internal in $!proc.start 17:50
ugexe ah clearly 17:52
eveo looks like there's another bug hiding in IO::Pipe 17:56
"No such method 'quit' for invocant of type 'Channel'" 17:57
When trying to slurp off a pipe from run(:out).out
Surprised existing tests don't catch that :S
maybe it's just my test :/ 17:58
ugexe pipe used to provide its own slurp-rest but that got removed a few weeks ago i think 18:00
eveo It's something about CWD 18:02
cpan@perlbuild4~/CPANPRC/rakudo (nom)$ ./perl6 -e 'run(:out, $*EXECUTABLE, "-e", ｢say 42｣).out.slurp'
cpan@perlbuild4~/CPANPRC/rakudo (nom)$ ./perl6 -e 'run(:out, :cwd</>, $*EXECUTABLE, "-e", ｢say 42｣).out.slurp'
Unhandled exception in code scheduled on thread 3
No such method 'quit' for invocant of type 'Channel'. Did you mean any of these?
Oh right
There's no ./perl6 or whatever in that dir
(I think that is it)
Yeah. run(:out, "meow").out.close gives same result 18:03
ugexe $*EXECUTABLE should use the full path though, so it should affect that? 18:07
shouldnt^
eveo Yeah, it does. If I try to run it manually I get "Unhandled exception: failed to load library 'dynext/libperl6_ops_moar.so'" 18:08
ugexe obviously it does, i mean I dont understand why that would happen
eveo it needs something in %ENV*<PATH> to work
ugexe ah I think i've encountered proc or proc async issue where i had to do `my $env = %*ENV; run(..., :$env)` 18:09
even though that should have been getting set as a default
eveo heh, good thing you mentioned that, 'cause I now see run fails to pass :env too :) 18:10
now I just need to figure out a way to test this :/ 18:12
timotimo is that what you get when you run the perl6-m inside the rakudo source tree rather than the one udner install/bin? 18:13
BenGoldberg To test, have the child process attempt to print something from it's own env.
run(:out, $*EXECUTABLE, "-e", 'say $*ENV{ok} // "fail"', env => { ok => ok }), or somesuch. 18:14
eveo I meant test the :cwd 18:15
I need to shove something into %*ENV<PATH> to make it work
BenGoldberg Put two executables of different names in two different directories, and see which is run.
s/different names/the same name/ 18:16
eveo Yeah, executables isn't something I have lying around
BenGoldberg On *nix, a script with the exec bit set should work. 18:17
eveo roast is for all OSes 18:18
I don't know what it's problem is. I set PATH right. 18:19
Guess I should fix the error out of failed procs and see what the error is
BenGoldberg feels silly. What's wrong with run(:out, :cwd</>, $*EXECUTABLE, "-e", 'say $*CWD') ? 18:20
eveo I kinda thought Proc::Async had a Proc under the hood, which you get from Proc::Async.start, so now with Proc being implemented in terms of Proc::Async, it's kind of a twister...
BenGoldberg: it doesn't start
BenGoldberg Proc::Async and Proc had previously been totally independent of one another. 18:21
Proc had a bit more functionality, but for historical raisins, it was not possible to implement Proc::Async in terms of Proc. 18:22
eveo But Proc::Async gave you Proc from .start's Promise
BenGoldberg might be misremembering. 18:23
eveo ah, k, sorted it. $*EXECUTABLE has ""./perl6"" as path and not just .perl6 18:25
*perl6
Geth rakudo/nom: 69b1b6c808 | (Elizabeth Mattijsen)++ | src/Perl6/World.nqp
Suggest 'last' on compile-time error for 'break'
 18:31
BenGoldberg m: dd $*EXECUTABLE 18:32
camelia IO::Path.new("./rakudo-m-inst/bin/perl6-m", :SPEC(IO::Spec::Unix), :CWD("/home/camelia"))
BenGoldberg m: dd $*EXECUTABLE.absolute
camelia "/home/camelia/rakudo-m-inst/bin/perl6-m"
BenGoldberg eveo, Try using that ^ when doing the cwd test. 18:34
eveo BenGoldberg: it doesn't like it. Unhandled exception: failed to load library 'dynext/libperl6_ops_moar.so'\n at <unknown>:1 (/home/cpan/CPANPRC/rakudo/perl6.moarvm:<dependencies+deserialize>)\ 18:36
Geth rakudo/nom: e1aa16cdb3 | (Zoffix Znet)++ | src/core/Proc.pm
Fix regression in run/shell not propagating :cwd and :env

Fixes module install issues. Bug find: irclog.perlgeek.de/perl6-dev/2017-...i_14727506		 18:37
BenGoldberg Erm. Include LD_LIBRARY_PATH in the env? 18:38
lizmat tests
eveo I forgot to spectest that one actually :| 18:40
only ran the new test I wrote
lizmat eveo: LN fails to install because of test failures in Temp::Path 18:42
eveo ZOFVM: Files=1255, Tests=139011, 117 wallclock secs (20.52 usr 3.25 sys + 2294.07 cusr 173.19 csys = 2491.03 CPU)
lizmat eveo: with --force it installs ok 18:43
m: my $a = 42; $a = $a but role { method foo { "foo" } }; dd $a.foo # this works 18:44
camelia "foo"
eveo lizmat: what sort of failures?
lizmat m: my $a = 42; $a but= role { method foo { "foo" } }; dd $a.foo # why doesn't this ?
camelia 5===SORRY!5=== Error while compiling <tmp>
Cannot make assignment out of but because structural infix operators are too diffy
at <tmp>:1
------> 3my $a = 42; $a but=7⏏5 role { method foo { "foo" } }; dd $a.f
eveo I know Temp::Path's tests were segfaulting last night
lizmat Aborting due to test failure: Temp::Path:ver('1.001001')
is there an easy way to run the tests of an installed module ?
eveo zef look Temp::Path; zef --/tap-harness test . 18:45
zef look Temp::Path; zef --debug --/tap-harness test .
oh that still won't show all without a custom zef
zef look Temp::Path; zef --debug --/tap-harness --/prove test .
lizmat tests ok this way :-( 18:46
ugexe you can set tap harness verbosity with some env var too
er, perl5's prove via a TAP_HARNESS liek env var 18:47
eveo lizmat: sounds like this bug is still unfixed by the Proc refactor as was hoped. rt.perl.org/Ticket/Display.html?id...et-history 18:48
Geth roast: 31e2f175a4 | (Zoffix Znet)++ | S29-os/system.t
Test run propagates :cwd and :env

Rakudo fix: github.com/rakudo/rakudo/commit/e1aa16cdb3		 18:49
ugexe github.com/rakudo/rakudo/blob/nom/...ry.pm#L258 18:52
$err is not really used in a way that would ever do anything 18:53
eveo ugexe: irclog.perlgeek.de/perl6-dev/2017-...i_14725874
was about to add that, just testing the module install stuff is fixed... 18:54
ugexe heh i was just about to ask that same thing
eveo yey... upgrade-perl6 worked flawlessly this time :D 18:58
ZOFVM: Files=1255, Tests=139011, 113 wallclock secs (19.32 usr 3.19 sys + 2214.24 cusr 159.24 csys = 2395.99 CPU) 19:04
yes! the run time numbers are low again :D
Geth rakudo/nom: f7d183b91f | (Zoffix Znet)++ | src/core/CompUnit/PrecompilationRepository.pm
Fix STDERR handling when $RMD is not set

The Proc::Async refactor[^1] accidentally made STDERR .note its output to $*ERR, instead of capturing it and then testing.
The $err is already tested where .err.slurp-rest(:close) was being tested before. Fix by changing the code to capture STDERR to $err var.
  [1] github.com/rakudo/rakudo/commit/3a...27d3174d84
AlexDaniel timotimo: “i don't get it, the exit code and output are all exactly identical for every step in the bisect log?” yep, that's how it works if the offending commit is the last one
AlexDaniel timotimo: because the bisection itself does not need to test the endpoint 19:09
(if one of the starting points was marked as “bad” already, we don't need to test it again to get our result) 19:10
jnthn jnthn-- # bugs 19:11
AlexDaniel timotimo: but I created this: github.com/perl6/whateverable/issues/160
jnthn eveo++ # fixes and, importantly, the tests that one woulda wished were there in the first place... 19:12
eveo jnthn++ Proc refactor :)
AlexDaniel timotimo: and also this, but you'll have to elaborate more :) github.com/perl6/whateverable/issues/161 19:14
eveo jnthn: what should this be? github.com/rakudo/rakudo/compare/b...c52a4a0R61
jnthn: there's no method quit on a Channel, so run(:out, "meows").out.slurp(:close) no complains there's no quit on channel 19:15
s/no/now/
jnthn oh my, why didn't we call it quit there too 19:16
I think .fail
eveo k, I'll change 19:17
jnthn yes: github.com/rakudo/rakudo/blob/nom/...el.pm#L245
ugexe rt.perl.org/Public/Bug/Display.html?id=127682 appears to be resolved now 19:23
eh, proc::async uses :$cwd and :%ENV? 19:28
jnthn Yeah, I spotted that inconsistency
It was done to match with %*ENV 19:29
Then Proc picked %env
We should probably make %env work on Proc::Async too
ugexe then i added :CWD to Proc, and that got renamed to :cwd for consistency heh
jnthn :%env that is
heh
Yeah
fail
There's also some :$enc vs :$encoding too I think 19:30
Though we can probably alias those everywhere also
Gotta go for a bit; bbl
eveo IO::Handle.read requires a size; .readchars takes a positional, defaulting to $*DEFAULT-READ-ELEMS, and .Supply takes a *named* args size
jnthn eveo: At a guess, .Supply taking that probably wants to go away; in the longer term, asking for a .Supply of a handle should probably do something like dup it, toss it to the libuv event loop, and actually do real async I/O on it... 19:32
Geth rakudo/nom: 1a2ee934cb | (Zoffix Znet)++ | src/core/Proc.pm
Fix regression with pipes of failed Procs

The Proc refactor[^1] accidentally used .quit on a Channel, but the method is called .fail
Fixes crashes when sinking or trying to read off pipes of failed Procs.
  [1] github.com/rakudo/rakudo/compare/b......8d6a911
jnthn And then since you're not asking for anything, just getting stuff as it comes, a size doesn't make much sense
jnthn I'm fine with .read defaulting to $*DEFAULT-READ-ELEMS like readchars 19:33
OK, really afk :)
Geth roast: 56c5a98b68 | (Zoffix Znet)++ | S29-os/system.t
Test failed Proc sinking and pipe handling

Rakudo fix: github.com/rakudo/rakudo/commit/1a2ee934cb
rakudo/nom: b7150ae1ad | (Zoffix Znet)++ | src/core/IO/Handle.pm
Give IO::Handle.read default size

To make consistent with .readchars. Per irclog.perlgeek.de/perl6-dev/2017-...i_14728193		 19:40
roast: 4c64fbac15 | (Zoffix Znet)++ | S16-filehandles/io.t
Test IO:Handle.read has a default size

Rakudo impl: github.com/rakudo/rakudo/commit/b7150ae1ad Per irclog.perlgeek.de/perl6-dev/2017-...i_14728193		 19:41
rakudo/nom: aa9516be23 | (Zoffix Znet)++ | src/core/IO/CatHandle.pm
Make IO::CatHandle.read consistent with IO::Handle.read

Add same default as in
  github.com/rakudo/rakudo/commit/b7150ae1ad		 19:50
roast: 1d4232f0c2 | (Zoffix Znet)++ | S32-io/io-cathandle.t
Test IO::CatHandle.read has a default size

Rakudo impl: github.com/rakudo/rakudo/commit/aa9516be23
eveo & 19:51
timotimo AlexDaniel: left you a little piece of explanation on the coverable issue 20:05
travis-ci Rakudo build failed. Elizabeth Mattijsen 'Fix various issues with Map (<=) Map 20:10
travis-ci.org/rakudo/rakudo/builds/242389821 github.com/rakudo/rakudo/compare/a...563a764462
buggable [travis build above] ✓ All failures are due to timeout (0), missing build log (2), GitHub connectivity (0), or failed make test (0).
AlexDaniel timotimo: nice 21:08
lizmat good night, #perl6-dev! 21:45
timotimo nite liz 21:46
Geth roast: 686dddb63f | (Samantha McVey)++ | S15-nfg/GraphemeBreakTest.t
Update the GraphemeBreakTest.t untodoing now passing tests

Also fix an issue involving not detecting normalization.		 23:34
samcv so yeah the 6.c-errata GCB tests were wrong. also i didn' write the code that creates the tests 23:37
and i'm not sure if i should attempt to fix it, or just make a note in the generator file that not to run it again? cause 6.c-errata shouldn't be changing right? and not have new tests added? 23:38
hm
or remove the generator file since it's supposed to stay static anyway idk 23:39
japhb Anyone else having trouble installing Inline::Python? I see some build warnings ("pyhelper.c:326:5: warning: ‘py_retval’ may be used uninitialized in this function [-Wmaybe-uninitialized]"), but should that cause zef to think the build should [FAIL]?
Geth roast/6.c-errata: d40e03f035 | (Samantha McVey)++ | S15-nfg/grapheme-break.t
Fix three wrong tests in S15-nfg/grapheme-break.t
samcv anyway. i'm pushing the fixes to the tests themselves but haven't changed the generator script or anything else 23:40
eveo, get back to me on what you think the best thing should be to do. as you are the 6.c-errata person 23:41
eveo samcv: dunno, maybe add a "die 'Some wrong tests were removed and this generator script is broken. Don't run it in 6.c-errata or it will re-generate the broken tests'" at the top of generator script? So the code is there, readily available, if ever needed, but at the same time you can't "accidentally" run it and get wrong tests back. 23:46
japhb: what's the full output?
samcv sounds good
eveo I see Inline::Python ded on toaster. But it's cause the lib or whatever isn't installed toast.perl6.party/module?module=In...4-g56e71d5 23:52

Powered by Cro and the Raku® Programming Language.

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

Thank you!