dalek ast: 99eacc7 | usev6++ | S32-list/ (2 files):
Unfudge two tests for RT #128720 for rakudo-j
05:31
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=128720
TimToady jnthn: I think the problem with RT #128550 is that in charrange in src/vm/moar/QAST/QASTRegexCompilerMAST.nqp, the final case is using ordat, which I suspect is the wrong thing for a synthetic 06:35
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=128550
nine [Coke]: you can install your modules as root. That's just the same as when distros package the modules. We can still write to ~/.perl6 at runtime. 06:54
[Tux] This is Rakudo version 2016.07.1-145-g3284025 built on MoarVM version 2016.07-16-g85b6537 07:00
test 15.434
test-t 7.350
csv-parser 16.283
$ panda install p6doc FAIL 07:03
t/trailing_whitespace.t (Wstat: 256 Tests: 342 Failed: 1)
Failed test: 30
# Failed test 'must not have any trailing whitespace in doc/Language/functions.pod6'
I have a separate .pod file for the documentation, because it is huge. panda install . or panda install Text::CSV however does not install that pod, so p6doc cannot find it 07:38
are there separate actions I need to take (in META6) in order to have the documentation available on install? 07:39
gfldex [Tux]: you may be better of with zef instead of panda 07:43
[Tux] installs zef … 07:44
==> Successfully installed zef 07:46
but, as with p6doc, it failed to install the symlink, so it cannot find the command :/
gfldex p6doc is using CompUnit to look for modules. I don't know if zef is aware of .pod6 yet. 07:48
[Tux] I have .pod, no .pod6 07:50
gfldex if you want to unconfuse github (and windows) you may want to rename it to .pod6 07:51
[Tux] I can, but that does not change panda/zef behaviour in (not) installing it 08:01
should I mention it separately in META6?
gfldex ugexe: ^^^ 08:02
stmuk Tux: you may have to write a Build.pm to install your docs 08:30
you probably can add it to the META6 08:31
depends on the location of files 08:32
moritz really hopes there's a declarative way to install documentation 08:59
jnthn Morning, #perl6-dev 09:20
TimToady: Yes, anything using ordat really wants a careful looking at... 09:24
dalek p: 946b078 | jnthn++ | tools/build/MOAR_REVISION:
Bump MOAR_REVISION for cancelnotify op.
10:38
p: 035f4e8 | jnthn++ | src/vm/moar/QAST/QASTOperationsMAST.nqp:
Map nqp::cancelnotify op for MoarVM.
p: 714b8eb | jnthn++ | src/vm/jvm/ (2 files):
Add cancelnotify for JVM backend.
10:46
kudo/nom: 69d1528 | jnthn++ | tools/build/NQP_REVISION:
Bump NQP_REVISION for cancelnotify op.
11:01
kudo/nom: 0aedc6e | jnthn++ | src/core/Supply.pm:
Better support eventual cancellations.

Some things take time to cancel, which can lead to races if the cancellation frees up a resource (such as a network port). Therefore, we allow such cancellations to return a Promise, which is awaited.
rakudo/nom: 9cbd855 | jnthn++ | src/core/IO/Socket/Async.pm:
rakudo/nom: Make socket listener cancellation Promise-y.
rakudo/nom:
rakudo/nom: This means that the close operation will not return until the socket
rakudo/nom: really has been closed. This fixes the instability in the async
rakudo/nom: socket tests that led to a test failure around 1 in 20 times.
pmurias jnthn: is rebootstrapping stage0 a big deal, not sure if I should do it to get rid of the $CONTROL_RETURN hack 11:03
dalek p: 7a8be7c | (Pawel Murias)++ | src/vm/js/nqp-runtime/package.json:
[js] Update repository in package.json of nqp-runtime.
11:04
nqp: 05fe569 | (Pawel Murias)++ | src/vm/js/nqp-runtime/ (3 files):
nqp: [js] Drop es6-map dependency, we are targeting EcmaScript 6 now.
jnthn pmurias: Not really, just bumps up the repo size a bit 11:05
dalek ast: d9249ea | jnthn++ | S17-lowlevel/lock.t:
Correct test that did unprotected array writes.

It's only safe to grab an *already vivified* element of an array and write to it in a cross-thread way. Since @out's elements were never vivified, this led to concurrent binds - and thus resizes - to the array, and thus incorrect results and crashes.
The SEGV that sometimes happened needs addressing too; that already has an RT issue and will be covered by another test that actually wants to create that condition, rather than this one which did it by accident.
11:06
p: 0769714 | (Pawel Murias)++ | src/vm/js/Compiler.nqp:
[js] Remove old hack.
11:36
p: 1e6666d | (Pawel Murias)++ | src/vm/js/Compiler.nqp:
[js] Explain a HACK better.
travis-ci Rakudo build errored. Jonathan Worthington 'Better support eventual cancellations. 11:58
travis-ci.org/rakudo/rakudo/builds/151180613 github.com/rakudo/rakudo/compare/3...edc6e08fef
buggable travis-ci, one build failed due to the timeout. No other failures.
moritz RabidGravy: you should make change buggable to not be triggered by other bots 12:11
a typical approach is to require an explicit colon to trigger the bot 12:12
ilmari having it tell us why the failure happened is useful, though
timotimo i appreciate getting immediate feedback from a travis-ci link that it's a useless "error" warning 12:13
ilmari t/04-nativecall/16-rt125408.t seems to hang on JVM
unmatched} moritz: buggable is my bot and its explicit purpose is to announce spurious failures, so we don't have to do anything about them
ilmari but not with --gen-nqp=master
moritz unmatched}: oh, ok 12:14
geekosaur I think the confusing part is it looks like it is responding to travis-ci 12:15
perhaps just drop that part (maybe only for travis-ci)?
dalek kudo/nom: e829a98 | lizmat++ | / (2 files):
Add 29 digit ranges to magic auto-inc/dec

Inspired by www.reddit.com/r/perl6/comments/4w...r_tuesday/ and ff26dd8e5e64068c6a (aka RT #128868)
The one-liner in question that was used to find missing digit ranges:
  .say for ((^0xffff)>>.chr.map({"{.ord.fmt("0x%05X")} {.uniname} = $_" if .uniprop eq "Nd" && .uniname.contains("DIGIT ZERO") && (.ord + 9).chr.uniname.contains("DIGIT NINE")}))
12:31
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=128868
unmatched} travis-ci.org/rakudo/rakudo/builds/151180613 12:37
buggable ✓ [travis build above] One job failed due to the timeout. No other failures.
jnthn Neat :) 12:42
lizmat thanks to the wonders of a generate script, this was rather simple 12:44
I was tempted to turn the one-liner into a script that would generate the generate script :-)
JimmyZ jnthn: looks like missing a change : github.com/rakudo/rakudo/blob/9cbd...ync.pm#L74 # though I am sure. 12:51
*not sure
jnthn I'm not sure either...though still want to review the bunch of places we cancel 12:52
JimmyZ :)
jnthn That one isn't likely to be problematic, however
JimmyZ async is hard ,hah
jnthn The listen one was because you actually care about the port being freed up :)
Which is why we had occasionally IO-Socket-Async.t fails
[Coke] nine: I ended up giving up on the perl6 user entirely; Not worth figuring out right now why the panda install process isn't seeing the env vars I set. (and therefore failing to download or install anything) 12:55
geekosaur if you were using sudo, remember that it resets environment variables 12:57
nine Don't know which of the great hotels to book for YAPC? No problem! Just wait till 2 weeks before the conference and scramble to find one that has any room available. Works every time :/ 13:08
RabidGravy moritz, buggable ain't nothing to do with me :) 13:10
geekosaur it's the nick of the week formerly known as zoffix >.> 13:12
DrForr Sounds like the scheduling. 13:20
travis-ci Rakudo build passed. Elizabeth Mattijsen 'Add 29 digit ranges to magic auto-inc/dec 13:29
travis-ci.org/rakudo/rakudo/builds/151200213 github.com/rakudo/rakudo/compare/9...29a98c4282
lizmat jnthn: seems like the last updates borked --profile 14:16
$ perl6 --profile -e 'my $a' 14:17
Writing profiler output to profile-1470838613.57616.html
===SORRY!===
SC not yet resolved; lookup failed
jnthn lizmat: timotimo noticed that a day or so ago, so it's not the very latest updates, at least 14:19
lizmat well, let me put it this way: late yesterday evening I was doing profiles 14:20
now I'm not :-(
jnthn Hm, it was the heap profiler he noticed it in though. Odd...
lizmat so I think it's the last NQP_REVISION bump
which brought in a new Moar 14:21
jnthn doesn't see any likely candidates in the NQP or Moar recent changes :( 14:23
So it's gonna have to be a bissect
mst declares today 'honorary second monday' then 14:25
jnthn I'm currently in the middle of fixing an icky SEGV though, so it'll be a bit before I can do it.
[Coke] geekosaur: was using the "USER" directive in a dockerfile. no clue what that does under the covers. :) 14:26
man, I am distractable. 14:27
RT: 1329; @LARRY: 14; CONC: 25; GLR: 6; JVM: 66; LTA: 114; NEW: 858; NYI: 82; OSX: 5; PERF: 22; POD: 17; PRECOMP: 10; RFC: 31; SEGV: 32; STAR: 5; TESTNEEDED: 15; TODO: 10; UNI: 28; UNTAGGED: 495; WEIRD: 3
note that @LARRY/RFC/NYI/TODO probably -all- need devs to review them to make sure we still want 'em. 14:28
geekosaur (distractable. new bot? :p )
unmatched} :D
jnthn At random time intervals, posts a link to a photo of a kitten? :) 14:29
dalek kudo/nom: 7501ee5 | lizmat++ | src/core/Array.pm:
Make Array.splice(offset,size,@) about 10x faster

This concludes my work on Array.splice for now
lizmat jnthn: perl6 --profile -e '' # golfed 14:33
dalek kudo/nom: 5dd91d8 | lizmat++ | src/core/Backtrace.pm:
Backtrace::Frame.is-(hidden|routine) return Bool

Spotted by sena_kun++
15:08
ugexe gfldex, [Tux] : There is no mechanism for pod installation yet. I liked the idea of implementing install/uninstall hooks at the CUR level, making p6doc a type of CUR itself, and having p6doc use the install/uninstall hooks to extend the installation process 15:09
stmuk^ 15:10
dalek p: 61157e7 | (Pawel Murias)++ | src/vm/js/nqp-runtime/reprs.js:
[js] Use Object.setPrototypeOf instead of the deprecated way of setting __proto__.
15:32
p: 6d10691 | (Pawel Murias)++ | src/vm/js/Compiler.nqp:
[js] Fix typo.
p: aaac2d0 | (Pawel Murias)++ | src/vm/moar/QAST/QASTOperationsMAST.nqp:
Stop the compilation of the nqp::for op from changing the QAST.
p: 6ad5470 | (Pawel Murias)++ | src/vm/js/Operations.nqp:
[js] Remove a hack no longer needed after fixing a bug in the moar backend.
p: d218ad1 | (Pawel Murias)++ | src/vm/js/nqp-runtime/core.js:
[js] Remove polyfill TODO as we are targeting ECMAScript 6 now.
15:34
[Coke] Back to working on github.com/coke/perl6-profiler-html/issues - anything we'd like to see in the profiler html template? looking to minimize the generated html/js/css, convert to angular... 18:01
er, angular 2
timotimo ooooooh 18:07
please make the call graph deep-searchable
[Coke] timotimo: add an issue? ;) 18:08
timotimo ok 18:20
[Coke] OHNOES MORE TIX 18:23
timotimo oh well 18:24
[Coke] :) 18:25
I have to learn angular 2 for work, this will be a good thing to practice on.
timotimo yup 18:28
i'm going to refrain from making a ticket that reads "make it faster" 18:29
[Coke] ... that is implied, yes. 18:37
timotimo :)
[Coke] ... wow, angular 2 ain't small 18:47
timotimo maybe we should build the thing with WebAssembly right away 18:48
lizmat hopes that --profile will be operational soon 19:20
timotimo was it AlexDaniel who could run a custom bisect for me? 19:21
unmatched} Yeah 19:22
timotimo AlexDaniel: could you try to find out what commit made --profile give the error message it currently gives? 19:23
AlexDaniel timotimo: sure 19:43
timotimo thank you 19:46
dalek kudo/nom: 5719826 | lizmat++ | src/core/Any-iterable-methods.pm:
We don't need to pass on as named param internally
19:48
AlexDaniel timotimo: 69d15282129bfa08f57263fd446c34f43576844a 19:49
lizmat eh, that much was clear to me already :-)
AlexDaniel which was expected, I think
lizmat yeah, but good to know it *is* that rakudo commit :-)
timotimo huh, that's newer than i thought 19:50
potentially because i tend to work with nqp from master ... 19:51
lizmat timotimo: I run 10s of profiles avery day
*every
and it worked late last night :-)
timotimo oh 19:53
oh hey we can just bisect nqp and ignore rakudo
that could make it fast enough to work
AlexDaniel unfortunately I have nothing to help with that :( 19:54
timotimo no worries
i'll do it locally
AlexDaniel \o/ 19:55
lizmat timotimo: pretty sure it's the MOAR_REVISION bump :-)
timotimo *shrug* 19:56
actually ... 19:59
i've got it both crash and not crash without changing what moarvm i'm running
jnthn Pretty much all the recent MoarVM changes were concurrency fixes, so I'm very curious which change it'll turn out to be... 20:01
timotimo last step.
2f9c5d5a6a9c1b0243fd59a7b827ad1f99d113cb is the first bad commit 20:02
Set the cursor_type for all regexes compiled by NQP.
lizmat ah?
jnthn: sorry :-) 20:03
jnthn Ho...hmm.
timotimo well, something's got a reference now that it didn't use to have
jnthn So it was an NQP one...
git show 2f9c5d5a6a9c 20:04
timotimo indeed
jnthn ...wow I'm tired :P
AlexDaniel timotimo: are you sure?
timotimo kind of
AlexDaniel timotimo: can you try d68b9b375f7bd5d31da95a8b9d457827f0413422 too?
jnthn timotimo: ooc, is --profile busted in NQP by this change? 20:05
AlexDaniel which is a child of 2f9c5d
it says “Fix 2f9c5d5” so it's probably worth a try
timotimo OK 20:06
that one's also broken
jnthn: yes, i'm only testing NQP, not rakudo 20:08
food time! TTYL 20:11
dalek kudo/nom: 5b50e27 | lizmat++ | src/core/Any-iterable-methods.pm:
.squish|.unique|.repeated now pass on is-lazy

Previously, they would always return False
20:41
kudo/nom: c322e1e | lizmat++ | src/core/Any-iterable-methods.pm:
first() returns a scalar, which is always not lazy
20:44
lizmat wonders whether either a ":doesn't-slip' named parameter to .map would make sense 21:00
it would allow the iterating logic to not have to check for Slips
maybe it should be called :raw , so it could be used even if you did return a Slip and didn't want it slipped ? 21:03
this would make .WHICH making a lot faster in a lot of cases
lizmat gets some shuteye first 21:04
good night, #perl6-dev!
timotimo gnite liz! 21:05
unmatched} s: Date, 'new', \() 23:53
SourceBaby unmatched}, Sauce is at github.com/rakudo/rakudo/blob/c322...ish.pm#L34
unmatched} s: Date, 'new', \('2016-06-06')
SourceBaby unmatched}, Sauce is at github.com/rakudo/rakudo/blob/c322...ate.pm#L36
unmatched} :D