samcv .tell Zoffix dunno if you'd be interesed in this. should be an easy fix. But DateTime's - now doesn't work. though DateTime - now.DateTime works. so i guess it doesn't check if it can coerce to a DateTime object 00:23
yoleaux samcv: I'll pass your message to Zoffix.
Zoffix . 00:24
yoleaux 00:23Z <samcv> Zoffix: dunno if you'd be interesed in this. should be an easy fix. But DateTime's - now doesn't work. though DateTime - now.DateTime works. so i guess it doesn't check if it can coerce to a DateTime object
Zoffix samcv: that's cause there's an ambiguity: it could mean $date-time - now.DateTime (returning an Int of days) or $date-time.Instant - now (returning Duration object) 00:25
samcv ah ok 00:27
thanks
Geth rakudo/supervisor_fewer_allocations: 4ef28a7bcb | (Timo Paulssen)++ | src/core/ThreadPoolScheduler.pm6
calculate $smooth-per-core-util without .sum

we don't yet manage to throw the named args hash away for the sum method, so we save a bunch of allocations here.
01:09
rakudo/supervisor_fewer_allocations: 8f4a5a5333 | (Timo Paulssen)++ | src/core/ThreadPoolScheduler.pm6
replace for ^$foo with native int loop

creating the range and its iterator was the single most expensive bunch of allocations.
Moving the variables to the front also allows us to get away without taking a closure here.
rakudo/supervisor_fewer_allocations: d425421644 | (Timo Paulssen)++ | src/core/ThreadPoolScheduler.pm6
nqp-ify a little bit of math

manually constant-folding away a multiplication and calling nqp math ops directly saves us pretty much all boxing that used to happen here.
AlexDaniel reportable6: 2018-03-19T00:00:00Z 2018-03-26T00:00:00Z 01:16
reportable6 AlexDaniel, OK, working on it! This may take up to 40 seconds
AlexDaniel, gist.github.com/0ee777ff9662ddb627...f09d73162d 01:17
AlexDaniel I'm fine. Thanks
reportable6: 2018-03-19T00:00:00Z 2018-03-26T00:00:00Z 01:18
reportable6 AlexDaniel, OK, working on it! This may take up to 40 seconds
AlexDaniel reportable6: 2018-03-19T00:00:00Z 2018-03-26T00:00:00Z
reportable6 AlexDaniel, OK, working on it! This may take up to 40 seconds
AlexDaniel, gist.github.com/2494cd8ee2e36dfcb7...44eb4e4d14 01:19
Geth rakudo: db7361a385 | (Zoffix Znet)++ | 2 files
Fix bad error on Lock.protect with wrong args

Fixes github.com/rakudo/rakudo/issues/1644 R#1644
The LEAVE phaser gets executed when wrong args are given, which causes LTA error: "unlock requires a concrete object with REPR ReentrantMutex".
Fix by making the method a multi, moving arg matching out of the block with the LEAVE phaser.
01:29
synopsebot R#1644 [open]: github.com/rakudo/rakudo/issues/1644 Lock.protect({}) fails, but with surprising message
rakudo: timo++ created pull request #1653:
Supervisor fewer allocations
01:35
roast: d733c2a9c5 | (Zoffix Znet)++ | S02-types/num.t
No parsing glitches in the way Num is parsed

Closes github.com/rakudo/rakudo/issues/1626 Rakudo fix: github.com/rakudo/rakudo/commit/84...23678b5a6c
03:25
rakudo: 646f9211cf | (Samantha McVey)++ | 2 files
Add ShiftJIS decode/encode support for MoarVM

This makes the new shiftjis support in MoarVM accessible. It is accessible as windows-932 because that is the dialect of shiftjis that was implemented. It is not listed as shiftjis to maintain forward compatibility in case other shiftjis (or the original) are added in later.
03:28
roast: f693c5547f | (Samantha McVey)++ | 3 files
Add shiftjis (windows-932) encoding tests

Tests encode, decode, reading from a filehandle. Also tests that replacement's work properly on unencodable bytes encountered on decoding or unencodable Unicode encountered on encoding.
03:45
roast: 589e30a8e8 | (Zoffix Znet)++ | S32-num/stringify.t
No precision lost in Num stringification

Closes RT#127184: rt.perl.org/Ticket/Display.html?id=127184 MoarVM fix: github.com/MoarVM/MoarVM/commit/06...103a025472
03:49
synopsebot RT#127184 [open]: rt.perl.org/Ticket/Display.html?id=127184 [BUG] One digit short to correctly stringify a double
Geth roast: 01d02dafb6 | (Zoffix Znet)++ | S32-num/stringify.t
Test same Nums strigify the same

Part of RT#132330: rt.perl.org/Ticket/Display.html?id=132330 Fixes github.com/MoarVM/MoarVM/commit/067c0594103a025
   github.com/MoarVM/MoarVM/commit/8841c4241b4faa8
   github.com/MoarVM/MoarVM/commit/af2eb8a7f7d4344
   github.com/MoarVM/MoarVM/commit/4d3fc2818d0032b
   github.com/rakudo/rakudo/commit/8422d7b4e23678b
   github.com/rakudo/rakudo/commit/a2a2a745c4242d1
03:56
synopsebot RT#132330 [open]: rt.perl.org/Ticket/Display.html?id=132330 [MATH] Floating point anomalies
rakudo: f48636011c | (Samantha McVey)++ | t/spectest.data
Add shiftjis encoding test to spectest.data
roast: b6d5364fe0 | (Zoffix Znet)++ | S02-types/num.t
Nums that are close to each other parsed correctly as different

Part of RT#132330: rt.perl.org/Ticket/Display.html?id=132330 Fix: github.com/MoarVM/MoarVM/commit/067c0594103a025
   github.com/MoarVM/MoarVM/commit/8841c4241b4faa8
   github.com/MoarVM/MoarVM/commit/af2eb8a7f7d4344
   github.com/MoarVM/MoarVM/commit/4d3fc2818d0032b
   github.com/rakudo/rakudo/commit/8422d7b4e23678b
   github.com/rakudo/rakudo/commit/a2a2a745c4242d1
04:00
roast: 855af84f82 | (Zoffix Znet)++ | S02-types/num.t
no parsing glitches in the way Num is parsed (2)

Part of RT#132330: rt.perl.org/Ticket/Display.html?id=132330 Fix: github.com/MoarVM/MoarVM/commit/067c0594103a025
   github.com/MoarVM/MoarVM/commit/8841c4241b4faa8
   github.com/MoarVM/MoarVM/commit/af2eb8a7f7d4344
   github.com/MoarVM/MoarVM/commit/4d3fc2818d0032b
   github.com/rakudo/rakudo/commit/8422d7b4e23678b
   github.com/rakudo/rakudo/commit/a2a2a745c4242d1
04:03
roast: f4a6c635f4 | (Zoffix Znet)++ | S32-str/sprintf.t
sprintf maintains sane precision when stringifying nums

Part of RT#132330: rt.perl.org/Ticket/Display.html?id=132330 Fix: github.com/MoarVM/MoarVM/commit/067c0594103a025
   github.com/MoarVM/MoarVM/commit/8841c4241b4faa8
   github.com/MoarVM/MoarVM/commit/af2eb8a7f7d4344
   github.com/MoarVM/MoarVM/commit/4d3fc2818d0032b
   github.com/rakudo/rakudo/commit/8422d7b4e23678b
   github.com/rakudo/rakudo/commit/a2a2a745c4242d1
04:16
synopsebot RT#132330 [open]: rt.perl.org/Ticket/Display.html?id=132330 [MATH] Floating point anomalies
roast: d1faf1d049 | (Zoffix Znet)++ | S02-types/set.t
Nums that are close to each other remain distinct when put in sets

Part of RT#132330: rt.perl.org/Ticket/Display.html?id=132330 Fix: github.com/MoarVM/MoarVM/commit/067c0594103a025
   github.com/MoarVM/MoarVM/commit/8841c4241b4faa8
   github.com/MoarVM/MoarVM/commit/af2eb8a7f7d4344
   github.com/MoarVM/MoarVM/commit/4d3fc2818d0032b
   github.com/rakudo/rakudo/commit/8422d7b4e23678b
   github.com/rakudo/rakudo/commit/a2a2a745c4242d1
04:25
roast: 7fdd5372b8 | (Zoffix Znet)++ | S02-types/num.t
Num literals yield closest available Num to their nominal value

RT#128912: rt.perl.org/Ticket/Display.html?id=128912 Fix: github.com/MoarVM/MoarVM/commit/067c0594103a025
   github.com/MoarVM/MoarVM/commit/8841c4241b4faa8
   github.com/MoarVM/MoarVM/commit/af2eb8a7f7d4344
   github.com/MoarVM/MoarVM/commit/4d3fc2818d0032b
   github.com/rakudo/rakudo/commit/8422d7b4e23678b
   github.com/rakudo/rakudo/commit/a2a2a745c4242d1
04:37
synopsebot RT#128912 [open]: rt.perl.org/Ticket/Display.html?id=128912 [BUG] decimal->float bad rounding
nine ~~ 05:45
yoleaux 25 Mar 2018 23:33Z <Zoffix> nine: any build jobs that build MoarVM/nqp/rakudo separately from scratch and check exit code fail now after this message " No rule to make target 'gen/nqp-version', needed by 'blib/Perl6/ModuleLoader.moarvm'". You can repro with this set of commands: gist.github.com/zoffixznet/03fac7a...3525db42a8
02:21Z <Zoffix> nine: Did a bit of digging and wrote all my findings in github.com/rakudo/rakudo/issues/1654
nine .tell Zoffix gen/nqp-version gets created by the check_nqp_version target which is run by m-all, so a make, make test or make install will all create the file. But maybe you're running another make target that does not imply m-all? Btw. I've had a more bullet proof version already, but that wouldn't work on nmake... 05:47
yoleaux nine: I'll pass your message to Zoffix.
nine .tell Zoffix scrap that, I see this in your output: /home/cpan/perl5/perlbrew/perls/perl-5.24.0/bin/perl tools/build/check-nqp-version.pl /home/cpan/R/install/bin/nqp-m 05:53
yoleaux nine: I'll pass your message to Zoffix.
nine .tell Zoffix I assume you get output from /home/cpan/R/install/bin/nqp-m --show-config | grep source-digest
yoleaux nine: I'll pass your message to Zoffix.
Zoffix . 06:19
yoleaux 05:47Z <nine> Zoffix: gen/nqp-version gets created by the check_nqp_version target which is run by m-all, so a make, make test or make install will all create the file. But maybe you're running another make target that does not imply m-all? Btw. I've had a more bullet proof version already, but that wouldn't work on nmake...
05:53Z <nine> Zoffix: scrap that, I see this in your output: /home/cpan/perl5/perlbrew/perls/perl-5.24.0/bin/perl tools/build/check-nqp-version.pl /home/cpan/R/install/bin/nqp-m
05:53Z <nine> Zoffix: I assume you get output from /home/cpan/R/install/bin/nqp-m --show-config | grep source-digest
Zoffix $ /home/zoffix/R/install/bin/nqp-m --show-config | grep source-digest
nqp::source-digest=f535bf491741d4a8a99257997be7bfa950f04899
nine: ^ yeah I get that
I mean, this is on working version, after I ran `make clean` on a blown up install and re-ran the install 06:20
Zoffix nukes gen/nqp-version and tries again and then will run the --show-config after/if it blows up
Oh, I see the reply on the Issue now 06:22
But yeah after it blows up the show-config still shows stuff: $ /home/zoffix/R/install/bin/nqp-m --show-config | grep source-digest 06:23
nqp::source-digest=f535bf491741d4a8a99257997be7bfa950f04899
Yup, tossing `-j` from make avoids the problem :) 06:25
[Tux] Failed running /pro/bin/perl5.26.0 Configure.pl --backends=moar --gen-moar=master --gen-nqp=master --git-reference="/pro/3gl/CPAN/rakudobrew/git_reference" --make-install at /pro/3gl/CPAN/rakudobrew/bin/rakudobrew line 57. 07:14
main::run("/pro/bin/perl5.26.0 Configure.pl --backends=moar --gen-moar=m"...) called at /pro/3gl/CPAN/rakudobrew/bin/rakudobrew line 382
main::build_impl("moar-blead", undef, "") called at /pro/3gl/CPAN/rakudobrew/bin/rakudobrew line 115
Makefile:16: recipe for target 'moar' failed
and I am up-to-date with rakudobrew 07:15
Rakudo version 2018.03-81-gf48636011 - MoarVM version 2018.03-29-g621ca3c22
csv-ip5xs0.901 - 0.972
csv-ip5xs-208.658 - 8.876
csv-parser36.869 - 37.926
csv-test-xs-200.428 - 0.505
test9.100 - 9.429
test-t2.497 - 2.500
test-t --race1.057 - 1.067
test-t-2044.599 - 45.585
test-t-20 --race15.691 - 16.397
07:59
lizmat Files=1236, Tests=76229, 309 wallclock secs (14.87 usr 5.15 sys + 2132.87 cusr 206.87 csys = 2359.76 CPU) 09:57
Geth rakudo/master: 4 commits pushed by (Timo Paulssen)++, (Elizabeth Mattijsen)++ 10:03
rakudo: d0fbe7447a | (Elizabeth Mattijsen)++ | src/core/ThreadPoolScheduler.pm6
Inline .getrusage-total to reduce nul-load overhead

  - fewer frames entered / BootHashes allocated
  - fewer Ints allocated
   - assiging a native int from a sub returning a native int allocates an Int
11:08
pmurias are the semantics where storing a 2**40+2 into an int32 overflows but storing 2**80 throws dies intentional and wanted? 11:16
yoleaux 05:40Z <Zoffix> pmurias: Are you sure nqp::div_In to make Nums is the right approach? This should be True: use nqp; dd nqp::div_In(99989999999999991, 100_000_000_000_000_000) == nqp::div_In(9998999999999999, 10_000_000_000_000_000); But it's converting both of those Ints to doubles before doing the division and precision gets busted
lizmat notable6: weekly 11:43
notable6 lizmat, 3 notes: gist.github.com/227027f16364524b14...8c74052271
stmuk lizmat: note also rakudup.github.io/ rebranding 12:12
lizmat stmuk: noted :-) 12:14
reportable6: 2018-03-19T00:00:00Z 2018-03-26T00:00:00Z 12:26
reportable6 lizmat, OK, working on it! This may take up to 40 seconds
lizmat hmmm.. looks like I killed reportable6 ? 12:27
reportable6: 2018-03-19T00:00:00Z 2018-03-26T00:00:00Z
reportable6 lizmat, OK, working on it! This may take up to 40 seconds
lizmat, gist.github.com/438c5ac0b1da6c04be...0645eff5c5 12:28
lizmat whee! :-)
m: my $a = 42; $a ** $a ** $a # but but but 12:40
camelia WARNINGS for <tmp>:
Useless use of "**" in expression "$a ** $a **" in sink context (line 1)
Numeric overflow
in block <unit> at <tmp> line 1
AlexDaniel actually, maybe there are now less segfaults after R#1645 ? Maybe reportable won't die like this anymore. 14:25
synopsebot R#1645 [open]: github.com/rakudo/rakudo/issues/1645 [SEGV][regression][⚠ blocker ⚠] Very simple spesh-related segfault
AlexDaniel reportable6: uptime
reportable6 AlexDaniel, 1 hour, 4 minutes, and 41 seconds, 201.992188MiB maxrss. This is Rakudo version 2018.02.1-166-g51c4d4d89 built on MoarVM version 2018.02-171-geee5be412 implementing Perl 6.c.
AlexDaniel that rakudo is *relatively* old
Geth rakudo: 3e9b705ea4 | (Elizabeth Mattijsen)++ | src/core/Rakudo/Internals/HyperRaceSharedImpl.pm6
Streamline .hyper.grep batch producer

This makes the producer go from 810 msecs to 639 msecs for something like
   say ^Inf .hyper.grep( *.is-prime ).skip(9999).head
Instead of creating a new IterationBuffer for the result, the existing IterationBuffer is reused.
The --profile shows significantly fewer allocations.
14:44
Zoffix pmurias: when I noticed that behaviour last night, I was certain it's a bug. Our goal is to throw on assignment and to overflow on modifications (like $x++). It doesn't make much sense to me that int32 would be overflowing for values <64-bit and throwing for >64-bit values. 15:17
pmurias: there's this old, partially filed ticket about inconsistencies in native types: rt.perl.org/Ticket/Display.html?id...et-history 15:18
pmurias the current behavior to me seems to be accidental rather than intentional 15:20
jnthn oh heh, I missed that protect was fixed by switching it to a multi... 15:33
Wasn't mentioned on the ticket :P 15:34
oh, it is 15:35
Just very small so I missed it
Zoffix: Did github.com/rakudo/rakudo/issues/1644 not get a test for a reason? 15:36
Zoffix jnthn: it's in the same commit: github.com/rakudo/rakudo/commit/db...394a7cb3fe 15:39
jnthn: I didn't want to spec that some arguments give X::Multi::NoMatch
jnthn Ah, I see
OK, fair enough
I wasn't too keen on specing that either
Geth rakudo: 299dc6fc55 | (Jonathan Worthington)++ | 2 files
Fix Lock::Async.protect to match Lock.protect
15:42
roast: dogbert17++ created pull request #407:
Add missing test for RT 128758
16:17
synopsebot RT#128758 [open]: rt.perl.org/Ticket/Display.html?id=128758 Reduce with numeric ops does not numify things if only one arg is passed ([*] set(1,2,3))
Geth roast: c7d45f680a | (Jan-Olof Hendig)++ | S03-metaops/reduce.t
Add missing test for RT 128758
16:19
roast: 6b0fcc2b91 | (Aleks-Daniel Jakimenko-Aleksejev)++ (committed using GitHub Web editor) | S03-metaops/reduce.t
Merge pull request #407 from dogbert17/test-rt-128758

Add missing test for RT 128758
nqp: d9d648efbe | (Jonathan Worthington)++ | tools/build/MOAR_REVISION
Bump MOAR_REVISION for child process ID support
16:21
¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...0-g7198eb8
rakudo: 1321917d1b | (Jonathan Worthington)++ | 2 files
Provide child process ID via Proc::Async

The `ready` method now conveys the child process ID. A `pid` method is added as an alias for `ready`. Arguably, we could just tell people to use `ready`, but providing a `pid` alias feels like it will be a little more self-documenting. Sharing a `Promise` between the two as an implementation detail makes this cheap to do, and we never did commit to what the `ready` `Promise` evaluates to, so it's safe to make it end up evaluating to the PID too. So far its use has just been as a "the process is started" signal, and it will continue to serve as that just fine.
16:27
roast: cbb0422090 | (Jonathan Worthington)++ | S17-procasync/basic.t
Spectests to cover Proc::Async.pid
dogbert17 jnthn: did you just fix github.com/rakudo/rakudo/issues/1638 ? 16:29
jnthn dogbert17: Pretty much; left a comment 16:33
dogbert17 cool 16:34
AlexDaniel YAY! 16:47
I needed this :)
tbrowder_: by the way ↑ 16:48
timotimo here's a second bump coming already
Geth nqp: 833364342a | (Timo Paulssen)++ | tools/build/MOAR_REVISION
bump moar for getrusage api change
16:55
¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...gfdb5e4d6b
rakudo: c71fa17d8d | (Timo Paulssen)++ | 4 files
adapt to allocation-free getrusage api
AlexDaniel Meaningless benchmark of the day: role Foo {}; my $x = 42; for ^1000 { .say; $x = $x does Foo } 16:57
jnthn Heh, a 1000+ level deep inheritance chain. :) 17:03
dogbert17 and rakudobrew is also busted 17:12
Zoffix timotimo: is there an nqp bump missing? (Just looking at commits log; haven't tried building stuff) 17:15
timotimo that last commit also bumps nqp 17:16
Zoffix Ah :D
timotimo wakelift.de/p/1a39c0ea-adfb-45d2-b...380197d71/ - proofreaders welcome 17:17
dogbert17 timotimo: I don't get this sentence 'However, "zero allocations" was in very close now.' 17:21
Zoffix timotimo: "The math we were doing [...] were doing some boxing and unboxing"; s/were/was/; 17:22
AlexDaniel m: say 42
camelia 42
AlexDaniel e: say 42
evalable6 Cannot test c71fa17d8da36b8cfcbd85584cee32566f25f059 (Commit exists, but a perl6 executable could not be built for it)
Zoffix Also in the same sentence there's a comma splice. "boxing and unboxing, a call" -> "boxing and unboxing. A call"
AlexDaniel grumbles
timotimo look again 17:23
AlexDaniel removes the build in a hope that it's just a fluke
Zoffix oh no comma splice, nm... (I would've used `;` instead of commas to separate the outer list) 17:24
Oh yeah looks good now :)
timotimo++ good blog/work :)
make test fails t/06-telemetry/01-basic.t and bunch of other telemetry tests 17:25
fuller output gist.github.com/zoffixznet/961a4b7...d1509ab5a5 17:26
That's with master/master/master
AlexDaniel e: say 42
evalable6 AlexDaniel, No build for c71fa17d8. Not sure how this happened!
timotimo yes, i'm investigating, Zoffix
m: use nqp; my int @foo; nqp::bindpos_i(@foo, 1, 1); 17:28
camelia ( no output )
timotimo ok, got the fix
Geth rakudo: 3468f6908b | (Timo Paulssen)++ | lib/Telemetry.pm6
fix telemetry with `int @data`
17:30
timotimo lizmat: can you add wakelift.de/2018/03/26/tangentiall...ated-work/ to the weekly today? :)
tbrowder_ AlexDaniel: ditto YAY! 17:44
[Coke] timotimo++ 17:48
travis-ci Rakudo build failed. Timo Paulssen 'adapt to allocation-free getrusage api' 17:56
travis-ci.org/rakudo/rakudo/builds/358502929 github.com/rakudo/rakudo/compare/1...1fa17d8da3
buggable [travis build above] ☠ All failures are due to: failed make test (6 failures). Across all jobs, 4 unique test files failed.
timotimo ^- it's already compiling the new build that probably fixes this exact failure 17:57
AlexDaniel e: say 42 18:01
evalable6 42
AlexDaniel \o/
timotimo so travis, you're telling me the build has been running for 32 minutes and it hasn't even started a single sub-task yet? 18:03
oh, no, it just wasn't loading. refreshing the page fixed it 18:04
Geth rakudo: 1c4eebe52a | (Elizabeth Mattijsen)++ | src/core/Rakudo/Internals/HyperRaceSharedImpl.pm6
Streamline .hyper.grep batch producer

This makes the producer go about 2.5% faster for something like:
   say ^Inf .hyper.map( *.is-prime ).skip(99999).head
The --profile shows significantly fewer allocations.
18:33
lizmat oops, that was a pasto in the title :-( should have been .hyper.map :-( 18:46
m: use nqp; nqp::getrusage(my int @a); dd @a # feels to me we need a MoarVM bump ? 18:50
camelia array[int].new(0, 168415, 0, 42103, 78964, 0, 0, 0, 15067, 171, 0, 61600, 40, 0, 0, 0, 264, 12)
lizmat huh?
ah, I guess the bump was hidden somewhere
timotimo it was 18:59
:S
also, that's a much nicer way to spell that that i really should have thought of, too %)
travis-ci Rakudo build passed. Elizabeth Mattijsen 'Streamline .hyper.grep batch producer 19:20
travis-ci.org/rakudo/rakudo/builds/358544783 github.com/rakudo/rakudo/compare/3...4eebe52a04
lizmat and another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2018/03/26/...ly-perl-6/ 21:56
lizmat calls it a day 22:26
jnthn lizmat++ # weekly 22:55