travis-ci Rakudo build errored. Elizabeth Mattijsen 'Make creating a Promise a bit cheaper 00:46
travis-ci.org/rakudo/rakudo/builds/290151432 github.com/rakudo/rakudo/compare/c...80cdafe4a2
buggable [travis build above] ✓ All failures are due to: timeout (3 failures).
MasterDuke how is that these two signatures are ambiguous? 01:19
:(Match $: Regex:D \var, int \im, int \monkey, $, $, $, *%_)
:(Match $: Mu:D \var, int \im, int \monkey, $, int \a, \context, *%_)
BenGoldberg m: say i ** i 02:06
camelia 0.207879576350762+0i
yoleaux 18 Oct 2017 06:00 EDT <Zoffix> BenGoldberg: 'cause quietly is a control thing, not a routine :)
BenGoldberg m: say e ** (- pi / 2 ) 02:07
camelia 0.207879576350762
BenGoldberg m: say (i ** i).im 02:08
camelia 0
BenGoldberg m: say (i ** i).im.WHAT 02:09
camelia (Num)
travis-ci Rakudo build passed. Zoffix Znet 'Remove trailing whitespace' 04:49
travis-ci.org/rakudo/rakudo/builds/290239042 github.com/rakudo/rakudo/compare/e...d7c7287864
AlexDaniel` whew! One hell of a changelog 05:21
still some editing to do… 05:22
travis-ci Rakudo build passed. Zoffix Znet 'Add buffering test files to list of files to run' 05:57
travis-ci.org/rakudo/rakudo/builds/290240206 github.com/rakudo/rakudo/compare/1...50d0f02efa
[Tux] I'll try to remember to run the timings later today. System is rather busy and 3.253 - 3.311 is due to system load (I think) 07:00
yoleaux jnthn: write tests for `lazy/hyper/race for foo() { }` 07:50
AlexDaniel` jnthn: maybe RT #132287 is something to look at if you have time 08:58
synopsebot RT#132287 [open]: rt.perl.org/Ticket/Display.html?id=132287 [REGRESSION][CONC] stall/block in async heavy code
jnthn MasterDuke: First one is narrower on second parameter (Regex ~~ Mu) but wider on the 7th (Mu !~~ int), which means they are tied in the candidate sort 09:02
Geth nqp: bae6331d11 | pmurias++ | 2 files
[jvm] Implement nqp::eqatic
 09:11
nqp/master: 5 commits pushed by pmurias++
ilmari Tux++ 09:12
Geth nqp: 48be669296 | pmurias++ | t/nqp/059-nqpop.t
Test nqp::eqatic
 09:13
jnthn That op name always makes me think of a seahorse...
gfldex m: my @a = do lazy for ^3 { now - FIRST now }; @a».put; 09:24
camelia Use of Nil in numeric context
Instant:1508491513.633690
Instant:1508491513.639123
Instant:1508491513.640955
in block at <tmp> line 1
Use of Nil in numeric context
in block at <tmp> line 1
Use of Nil in numeric context
in block …
gfldex where does that Nil come from?
jnthn Does FIRST work as an r-value? 09:27
Geth nqp: 18a635e7ae | pmurias++ | src/vm/js/nqp-runtime/io.js
[js] Make open throw a nqp level exception on missing file
 09:42
nqp: cbf4654747 | pmurias++ | t/nqp/019-file-ops.t
Test catching an exception when nqp::open is passed a missing file
nqp: f49799985b | pmurias++ | src/QRegex/Cursor.nqp
Remove workaround for lack of nqp::eqatic on the js and jvm
 09:44
ilmari pmurias: does the js version handle nqp::eqatic("ss", "ß", 0)? 09:48
travis-ci NQP build failed. pmurias 'Remove workaround for lack of nqp::eqatic on the js and jvm'
travis-ci.org/perl6/nqp/builds/290375667 github.com/perl6/nqp/compare/cbf46...9799985bfe
ilmari JS doesn't seem to have .toFoldCase() 09:49
nqp: nqp::eqatic("ß","ss",0)
camelia ( no output )
ilmari nqp: nqp::say(nqp::eqatic("ß","ss",0)) 09:50
camelia 1
ilmari nqp-js: nqp::say(nqp::eqatic("ß","ss",0))
camelia /bin/sh: node: command not found
Zoffix [Tux]: it's possible the in-buffer thing made things slower 10:19
well, guaranteed even, just not sure by how much 10:20
Geth nqp: ba2a1863f9 | pmurias++ | src/QRegex/Cursor.nqp
Revert "Remove workaround for lack of nqp::eqatic on the js and jvm"

This reverts commit f49799985bfe62dbb421f78473a984fd6cb3da1e.		 10:21
Zoffix m: class Foo { has int $!x = 0x100000; method x ($v) { $v.abs }; method y { self.x: 0x100000 } }; my $x = Foo.new; $x.y for ^1000_000; say now - INIT now 10:22
camelia 0.48197415
Zoffix m: class Foo { has int $!x = 0x100000; method x ($v) { $v.abs }; method y { self.x: $!x } }; my $x = Foo.new; $x.y for ^1000_000; say now - INIT now
camelia 0.77565530
Zoffix m: say 0.77/.48
camelia 1.604167
Zoffix so buffer replenishing is 1.6x slower
pmurias ilmari: nqp::say(nqp::eqatic("ß","ss",0)) is broken on the js 10:23
how do I reboostrap the nqp-j? 10:24
Zoffix m: class Foo { has $!x = 0x100000; method x ($v) { $v.abs }; method y { self.x: $!x } }; my $x = Foo.new; $x.y for ^1000_000; say now - INIT now 10:25
camelia 0.5604632
Zoffix need to check whether removing `int` type constraint from in-buffer/out-buffer makes things better or worse
or alternatively, adding `int` type constraint on things where $!in-buffer/$!out-buffer is given 10:26
in IO::Handle
pmurias is normalizing to NFD and taking the first codepoint of that normalization a sane way to implement nqp::ordbaseat? 10:30
Zoffix m: $*IN.in-buffer = 0; say get 11:16
camelia Out of range: attempted to read 0 bytes from filehandle
in block <unit> at <tmp> line 1
Zoffix Just realized this... there's really no way to turn off in buffering, since we don't really have it. It just feeds the decode 11:17
I think I'm gonna revert it. It's really iffy
(no buffering for binary, except for slurp; always buffering for char mode, even tho the "buffer" is just our decoder, etc) 11:18
lizmat hhmmmm... the past 2 make spectests I've run, both hung, but on different files 11:39
once on S05-match/capturing-contexts.t (test 24) 11:40
and once on S26-documentation/10-doc-cli.t (test 1)
Am I the only one seeing this?
lizmat starts of another spectest run
*off 11:41
*sets
hmmm... next run completed ok :-) 11:47
jnthn pmurias: Yes, that's how MoarVM does it 12:00
lizmat hmmm... intriguing bug that I think Zoffix is going to like 12:11
put "sub MAIN(:$foo) { .say for lines }" into a file, e.g. foo.pl
then run the file with "perl6 foo.pl <some_other_file" 12:12
it will show the lines from some_other_file
now run it with:
then run the file with "perl6 foo.pl --foo <some_other_file"
and get: Failed to open file /Users/liz/Github/rakudo.moar/--foo: No such file or directory 12:13
Zoffix lizmat: not a bug and will be changed in 6.d: github.com/perl6/6.d-prep/blob/mas...nside-main
well, not a bug in a sense that nothing's broken. Everything works as advertized, except it's not useful, since @*ARGS are fed into MAIN 12:14
sub MAIN(:$foo) { .say for $*IN.lines } will workaround it in 6.c 12:15
lizmat ok, so I'll just @*ARGS.pop :-( 12:16
Zoffix or sub MAIN(:$foo) { $*ARGFILES = IO::CatHandle.new: $*IN; .say for lines } 12:17
(might need `my` on `$*ARGFILES`)
Weird. Stresstest hung. Killed it after 354s, but it still says "pass" 12:20
with 2 TODOs in t/spec/S12-construction/destruction.t passing too 12:21
lizmat just added a --hyper option to Tux's test-t, making it 2x as fast 12:22
[TuxCM] Whoosh ... 12:23
lizmat so what would be the best way to share an object between all elements of a batch
?
stmuk_ I can't install DBIsh on moar-blead
lizmat case in point:
Zoffix stmuk_: are you sure you got latest rakudo? I fixed a bug yesterday in NativeCall that prevented DBIish's installation 12:24
lizmat lines.hyper.map: { my $csv = Text::CSV.new; $csv.parse($_) }
I'd like the $csv to be shared by all calls from the same batch
stmuk_ This is Rakudo version 2017.09-464-gdd50d0f02 built on MoarVM version 2017.09.1-608-ge051ee3c
Zoffix m: $*PERL.compiler.version.say 12:25
camelia v2017.09.464.gdd.50.d.0.f.02
Zoffix stmuk_: what's the error?
lizmat doing: lines.hyper.map: { my $csv = once Text::CSV.new; $csv.parse($_) } 12:26
fails spectacularly: moar(28720,0x70000c501000) malloc: *** error for object 0x7f925bd014f0: double free
Zoffix stresstest hung again; htop shows it's S17-promise/lock-asyunc-stress1 and 2 that are still running. 12:27
Zoffix gives them a few more minutes
S17-promise/lock-async-stress1.t and S17-promise/lock-async-stress2.t I mean
zero CPU use tho 12:28
stmuk_ gist.github.com/stmuk/50bd08160f9d...aa65ae33ca 12:29
Zoffix "constant is-win = Rakudo::Internals.IS-WIN();" seriously? 12:31
Zoffix 's distaste for DBIish grown stronger now
jnthn lizmat: Make the instnace of the object outside of the map block 12:35
lizmat: But it will only work if Test::CSV is a threadsafe object
Which, if it's stateful on the inside, it won't be 12:36
I think for map I might set it up to closure-clone the block per worker though, so that once/state will be per batch
So those don't end up shared between threads 12:37
lizmat Text::CSV is not threadsafe 12:39
stmuk_ the gist updated with -v output
Zoffix stmuk_: how did you make it? Does zef have an option for that?
stmuk_ zef -v install 12:40
lizmat jnthn: "once my $csv = Text::CSV.new;" fails because at some point $csv is not initialized
and "my $csv = once Text::CSV.new" crashes with a double free error
jnthn You just said yourself that Test::CSV is not threadsafe 12:41
And it's long known that state/once ain't
lizmat "so that once/state will be per batch" is what you just said ?
jnthn Yes, that's a proposal
lizmat aaaah
ok
jnthn Though it might be a good one :) 12:42
lizmat well, that would make sense
yes, I think that would make sense :-)
and a good one :-)
Geth rakudo/nom: e0dabacad2 | (Zoffix Znet)++ | t/spectest.data
Remove in-buffer tests

  - The feature getting the axe for now
  - The tests were added a day ago; not part of 6.c-errata		 12:46
rakudo/nom: 7a2561c24b | (Zoffix Znet)++ | 2 files
Revert "Implement knob to weak IO::Handle input buffer"

This reverts commit 1e5d6f0d4eee9d3d6b02d064063ed2e633207bd5.
The current implementation of the feature is a bit iffy, so I'm reverting it for now. Problems:
... (7 more lines)
roast: a61110535d | (Zoffix Znet)++ | S32-io/in-buffering.t
Remove in-buffer tests

Tests were added a day ago and aren't part of 6.c-errata.
Feature currently reverted for reasons[^1]
  [1]
Zoffix [1] github.com/rakudo/rakudo/commit/7a2561c24b
timotimo pmurias: actually there's situations where the first piece of NFD is *not* the base character, because now unicode has "prepend" combining codepoints that go in front 12:48
i wonder if we can get "once" run in parallel to not segfault 12:53
Zoffix .ask [Coke] what was the judgement for "unneeded" in github.com/perl6/doc/commit/0b5126...e4b9da8d10 ? There were many IO::Handle tests unskipped and they create files on the filesystem. Seems skipping is advisable
yoleaux Zoffix: I'll pass your message to [Coke].
timotimo we usually want users to be incapable of segfaulting (or otherwise crashing) the VM without using NativeCall 12:54
jnthn timotimo: It's not gonna be once doing that, it'll be something inside of Text::CSV, and almost certainly will boil down to the usual suspects of VMArray and VMHash 12:55
Zoffix doing bumps. Gonna need them anyway soon for release testing and will make it easier to debug stmuk_'s bug 12:56
timotimo ah, that's fair then
lizmat looks into it further
stmuk not "my" bug now :) 12:57
Zoffix "nothing to commit, working directory clean" 12:58
Actually, we're already on moar-master
lizmat jnthn: would this workaround be ok: 12:59
outside of hyper: my @CSV[64]; # store for CSV objects / thread
inside hyper: my $csv = @CSV[$*THREAD.id] //= Text::CSV.new;
timotimo lizmat: why not map over a list of Text::CSV instances and just use $_
oh, because you don't know the number of threads it'll spawn 13:00
lizmat timotimo: que? normally, in test-t, there's only one Text::CSV object
in any case, this brings it down from 1.5 seconds to 1.1 seconds for me 13:01
timotimo rather than guess how many threads there'll be, i'd probably lock additions to that array? 13:02
lizmat timotimo: still wouldn't know which worker would need which object ?
timotimo because right now if you run it on a 128 core machine it'll crash again :)
lizmat well, yes
my @CSV[$*SCHEDULER.max_threads] # voila 13:04
timotimo: that should be save, no ? ^^^
*safe
timotimo sounds good 13:05
jnthn lizmat: Just fix it like I said ;) 13:07
github.com/rakudo/rakudo/blob/nom/...mpl.pm#L39
Here, declared my $mapper = $!mapper.clone();
Zoffix .tell AlexDaniel` when I stresstest on my 24-core VM, S17-promise/lock-async-stress.t and S17-promise/lock-async-stress2.t look to hang (waited up to 700s, where normal stresstest takes ~150s). Tho when I run them individually, they pass (each takes ~50s to run). Based on logs, 2 days ago this did not happen: irclog.perlgeek.de/perl6-dev/2017-...i_15317839
yoleaux Zoffix: I'll pass your message to AlexDaniel`.
jnthn And then call that inside of the block instead of the attr
AlexDaniel` . 13:08
jnthn oh, they have the & sigil
yoleaux 13:07Z <Zoffix> AlexDaniel`: when I stresstest on my 24-core VM, S17-promise/lock-async-stress.t and S17-promise/lock-async-stress2.t look to hang (waited up to 700s, where normal stresstest takes ~150s). Tho when I run them individually, they pass (each takes ~50s to run). Based on logs, 2 days ago this did not happen: irclog.perlgeek.de/perl6-dev/2017-...i_15317839
Zoffix AlexDaniel`: "look to hang" => I see them both in `htop` as the only remaining test files and there's zero CPU use
Zoffix reverts two Promise-affecting commits and tries again 13:10
This run: 540 wallclock secs before I killed it
timotimo is one of them mine? 13:11
Zoffix Two of them yeah: ee3f0f4fb8bf09c14c4ef26c9b9dc648985b19de 270e7c8a72351c8aa4edfc8f52463bbf51bf09c6 f980cdafe4a2ef103d32e25a9276d972c3cd57aa
Trying right now with those three reverted 13:12
lizmat Zoffix: if reverting those 3 fixes it, then I suggest with just reverting mine 13:13
I don't think timotimo's changes could affect anything
AlexDaniel` well if it works without these three then I guess it wouldn't be too hard to bisect 13:15
Zoffix yup, completed now in Files=1282, Tests=152759, 153 wallclock secs (21.52 usr 3.34 sys + 3283.92 cusr 168.13 csys = 3476.91 CPU) 13:17
Zoffix tries with just the f980cdafe4a2 remnoved
timotimo wow, that's fascinating
lizmat if that fixes it, I suspect some deep gremlins wrt "is default" :-( 13:19
timotimo ;(
AlexDaniel` heh, a lot of fun stuff going on here :) 13:21
by the way, everyone did a great job this month! Still working on the changelog… :) 13:23
Zoffix \o/
ZOFVM: Files=1282, Tests=152761, 153 wallclock secs (21.30 usr 3.30 sys + 3297.22 cusr 167.33 csys = 3489.15 CPU)
timotimo fwiw, i don't think i did a full spec test run after changing the default values for .keep() and .break() 13:24
that wasn't great
Zoffix spectest wouldn't help as these two are stresstests :)
timotimo right, but there could have been places that indirectly rely on the values 13:25
Geth rakudo/nom: d20e972a12 | (Zoffix Znet)++ | src/core/Promise.pm
Revert "Make creating a Promise a bit cheaper"

This reverts commit f980cdafe4a2ef103d32e25a9276d972c3cd57aa.
This commit appears to cause reproducible stresstest hangs in S17-promise/lock-async-stress.t and S17-promise/lock-async-stress2.t
IRC: irclog.perlgeek.de/perl6-dev/2017-...i_15329963
AlexDaniel` Zoffix++
what about that DBIish issue?
Zoffix I reproed it on its travis 13:26
Filed potential is default gremlins as RT#132326 13:28
synopsebot RT#132326 [new]: rt.perl.org/Ticket/Display.html?id=132326 Potential issue with `is default` and Promises
jnthn It'll probably be lazy attribute initialization 13:29
lizmat jnthn: but "is default" init isn't lazy ? 13:30
is it?
hmmm 13:31
jnthn Attribute scalar container creation is
Down in The Guts
But I'd like to kill that off
It makes every attribute access more expensive
lizmat so how would you like to fix that ? 13:32
jnthn I don't know yet.
lizmat ah, ok :-)
is cloning a container atomic ?
or could it be ? 13:33
timotimo cloning could cause a GC run
lizmat ah, ok, so cratch that then :-) 13:34
timotimo i don't think anything that allocates from gc memory can easily be made atomic
actually, what does atomic mean here ...
jnthn The problem isn't the cloning of the container, it's the time between checking if we have one, and installing the clone 13:35
lizmat that no other thread could access that container when half initialized
Zoffix AlexDaniel`: DBIish failures are segfaults. Full output: gist.github.com/zoffixznet/4ded7a0...1bd3d3975d 13:40
All three failing test files segfault
perl6 -Ilib t/40-sqlite-common.t # segfaults after 11th test 13:41
100% reliability
Zoffix & for a few hours
Geth rakudo/nom: d37a19ea83 | (Elizabeth Mattijsen)++ | src/core/Rakudo/Internals/HyperRaceSharedImpl.pm
Make 'once' work in hypered/raced blocks, jnthn++

  - by cloning the mapper for each batch		 13:43
Zoffix t/spec/S07-hyperrace/hyper.t also hung now. I killed stresstrest after 232s; passes on its own 13:47
Zoffix will run a few more stresstests later tonight 13:48
This is on 2017.09-467-gd20e972
(freshly cloned)
Zoffix &
lizmat Zoffix: just ran a stresstest, and no hang, but one fail: MISC/bug-coverage-stress.t (test 3) 13:51
AlexDaniel` alright 14:11
soooooo
c: 2017.09,HEAD gist.githubusercontent.com/AlexDan.../dbiish.p6
committable6 AlexDaniel`, Successfully fetched the code from the provided URL.
AlexDaniel`, gist.github.com/8646c6815619ad858b...ad44a4b30f
buggable New CPAN upload: Color-Names-1.001001.tar.gz by HOLLI cpan.metacpan.org/authors/id/H/HO/...001.tar.gz
AlexDaniel` sooo yeah 14:12
bisect: old2017.09 gist.githubusercontent.com/AlexDan.../dbiish.p6
bisectable6 AlexDaniel`, On both starting points (old=2015.12 new=d37a19e) the exit code is 1 and the output is identical as well
AlexDaniel`, Output on both points: «04===SORRY!04=== Error while compiling /tmp/xNaBDpld7m␤Malformed postfix call␤at /tmp/xNaBDpld7m:1␤------> 03old2017.08⏏0409 gist.githubusercontent.com/Al»
AlexDaniel` ooops
bisect: old=2017.09 gist.githubusercontent.com/AlexDan.../dbiish.p6
bisectable6 AlexDaniel`, Successfully fetched the code from the provided URL.
AlexDaniel`, Bisecting by exit signal (old=2017.09 new=d37a19e). Old exit signal: 0 (None)
AlexDaniel` why so slow? Because precomp 14:13
bisectable6 AlexDaniel`, bisect log: gist.github.com/68741e55bc632f1b2d...9513384dfd
AlexDaniel`, (2017-10-18) github.com/rakudo/rakudo/commit/ff...d3db2eae7d
AlexDaniel` let's double-check
timotimo AlexDaniel`: were you meaning to be in whateverable? ;) 14:14
AlexDaniel` commit: ff063e7b53^,ff063e7b53 gist.githubusercontent.com/AlexDan.../dbiish.p6
committable6 AlexDaniel`, Successfully fetched the code from the provided URL.
timotimo (i have no clue what exactly you're doing)
committable6 AlexDaniel`, gist.github.com/e25c6a14d6078e99f6...de494565b1
AlexDaniel` timotimo: not really, someone said that DBIish tests are failing
Geth: ver github.com/rakudo/rakudo/commit/ff...d3db2eae7d 14:15
Geth AlexDaniel`, version bump brought in these changes: github.com/perl6/nqp/compare/2017....9-g254d590
AlexDaniel` Geth: ver github.com/perl6/nqp/commit/254d59...708c68f9f0
Geth AlexDaniel`, version bump brought in these changes: github.com/MoarVM/MoarVM/compare/2...8-ge051ee3
timotimo right
AlexDaniel` nine: any ideas? ↑
Geth rakudo/nom: 870eaa316a | (Elizabeth Mattijsen)++ | 5 files
Optimize hyper/race a bit

  - use prefix ++ instead of postfix ++ in sink contexts
  - start at -1 instead of 0, allowing use of prefix ++		 14:17
AlexDaniel` oh, so we can try with MVM_JIT_DISABLE maybe?
commit: MVM_JIT_DISABLE=1 ff063e7b53^,ff063e7b53 gist.githubusercontent.com/AlexDan.../dbiish.p6 14:18
committable6 AlexDaniel`, Successfully fetched the code from the provided URL.
AlexDaniel`, gist.github.com/24319e448313c13c16...7d7961accb
AlexDaniel` yup.
nine AlexDaniel`: can you find out which of those commits is guilty? 14:20
yoleaux 18 Oct 2017 20:55Z <japhb> nine: Should nqp/rakudo bump their moarvm version to catch your nativecall jit improvements? I would think we should test more before the release, unless it needs more work to be stable again.
AlexDaniel` nine: only manually. Want me to do it? 14:21
nine AlexDaniel`: please. I'm a bit under the weather unfortunately 14:22
AlexDaniel` nine: alright, I'll get back to you in ≈1h
meanwhile here's a ticket to track the progress: RT #132328 14:23
synopsebot RT#132328 [new]: rt.perl.org/Ticket/Display.html?id=132328 [SEGV][REGRESSION] DBIish tests are failing spectacularly (JIT compilation of native calls)
buggable New CPAN upload: Color-Names-1.001002.tar.gz by HOLLI cpan.metacpan.org/authors/id/H/HO/...002.tar.gz 14:31
[Coke] Zoffix: the skips were unneeded, because that code all compiled. 14:34
yoleaux 12:53Z <Zoffix> [Coke]: what was the judgement for "unneeded" in github.com/perl6/doc/commit/0b5126...e4b9da8d10 ? There were many IO::Handle tests unskipped and they create files on the filesystem. Seems skipping is advisable
AlexDaniel` samcv: fyi there's an issue with moarvm RT #132328 14:35
synopsebot RT#132328 [new]: rt.perl.org/Ticket/Display.html?id=132328 [SEGV][REGRESSION] DBIish tests are failing spectacularly (JIT compilation of native calls)
[Coke] so for those few that were generating files, we can re-skip those. it would be nice if we could add a reason to the skip-test
Zoffix [Coke]: ahhh. Got it
[Coke] ... except they shouldn't be generating files.
because we're just checking to see if the code compiles with:
AlexDaniel` [Coke]: for now you can leave a regular pod comment maybe
Zoffix [Coke]: right, they'd generate files only during run time. I didn't consider that we're only compiling
[Coke] github.com/perl6/doc/blob/master/x...tion.t#L85 14:36
Zoffix lizmat: does MISC/bug-coverage-stress.t fail all the time?
[Coke] (it's code wrapped in an if False { ... } and a few other things, but unless it's got a BEGIN it's not going to run.
Zoffix Like if you run it on its own; t/fudgeandrun MISC/bug-coverage-stress.t
[Coke] AlexDaniel`: a pod comment that says what? 14:37
I would argue it's not needed.
AlexDaniel` [Coke]: with a reason for skip-test
[Coke] skip-test *takes an arg*.
please use that mechanism which is documented rather than adding in a comment that isn't attached to anything.
AlexDaniel` aaah sorry I misunderstood you 14:38
[Coke] github.com/perl6/doc/blob/master/EXAMPLES.md
\o/
AlexDaniel` runs back to his git bisect paradise
jnthn Righty, those tests I was gonna write... 14:41
[TuxCM] Before merge of Liz' work ... 14:42
This is Rakudo version 2017.09-467-gd20e972a1 built on MoarVM version 2017.09.1-608-ge051ee3c
csv-ip5xs 1.158 - 1.167
test 8.391 - 8.468
test-t 3.101 - 3.105
csv-parser 11.610 - 11.751
Liz, --hyper causes a core dump 14:47
double free corruption 14:48
Zoffix [TuxCM]: it's likely fixed on 468
Geth roast: 4a6da2ce57 | (Jonathan Worthington)++ | S07-hyperrace/for.t
Test `hyper for`, `race for`, and `for` semantics
 14:51
rakudo/nom: 527b88816a | (Jonathan Worthington)++ | t/spectest.data
Run S07-hyperrace/for.t
 14:52
AlexDaniel` nine: hello again! It's github.com/MoarVM/MoarVM/commit/4e...5861468dc1 (works just fine on its parent and segfaults on that commit)
trisectable6: where are you? :) 14:57
bisectable6 AlexDaniel`, On both starting points (old=2015.12 new=870eaa3) the exit code is 1 and the output is identical as well
AlexDaniel`, Output on both points: «04===SORRY!04=== Error while compiling /tmp/fG7RI7XGeI␤Unsupported use of ? and : for the ternary conditional operator; in Perl 6 please use ?? and !!␤at /tmp/fG7RI7XGeI:1␤------> 03where are you?08⏏04 :)»
AlexDaniel` …
Geth roast: 3659ede149 | (Jonathan Worthington)++ | S04-statements/for.t
Test lazy for loops
jnthn There we go, yesterdays commits testified :) 14:58
Well, my ones at least :-)
timotimo mine have testing, too 14:59
jnthn nice :)
timotimo well, i only made the promise stuff 15:00
i promised, and i delivered
so, what's the over/under of your promise delivery? :P
nine AlexDaniel`: I pushed a revert of that commit 15:05
AlexDaniel` nine: confirming that it works after a revert 15:10
Geth rakudo/nom: 7e9b9633d6 | (Jonathan Worthington)++ | src/core/RaceSeq.pm
Fix copy-pasta spotted by b2gills++
Zoffix bumps 15:11
AlexDaniel` Zoffix: thanks 15:13
OK I'll be away for an hour or two, .tell me if there are any issues meanwhile 15:14
fwiw I will run toaster tonight (unless somebody beats me to it)
Zoffix: ah, so what's up with IRC::Client? 15:15
any new info?
or old info maybe, I didn't really follow it
Zoffix AlexDaniel`: I didn't get a chance to investigate what's up with it. Planning to do that in 6h 15:16
AlexDaniel` Zoffix: OK, thank you
here's a wild guess, what if it's just output buffering? 15:18
no, probably not, nevermind
Zoffix ZOFFLOP: t/spec/S11-modules/require.t 15:19
ZOFFLOP: TODO PASS: t/spec/S12-construction/destruction.t
Geth nqp: 252fd89025 | (Zoffix Znet)++ | tools/build/MOAR_REVISION
Bump MoarVM
 15:20
¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...9-g1a9be0a
rakudo/nom: 6726d4af20 | (Zoffix Znet)++ | tools/build/NQP_REVISION
Bump NQP
rakudo/nom: version bump brought these changes: github.com/perl6/nqp/compare/2017....6-g252fd89
da977785c6 | (Jonathan Worthington)++ | src/core/Rakudo/Internals/HyperRaceSharedImpl.pm

This splitting up also removes a level of indirection for the common block case of hyper/race grep, and thus should speed things up a bit in that case also.
jnthn Alright, I think that's all I've time/energy for this week 15:29
lizmat Zoffix: MISC/bug-coverage-stress.t runs fine by itself 15:30
afk&
timotimo have fun on your vacay, jnthn :)
jnthn timotimo: Thanks! :) Will be away Sun-Wed afternoon
lizmat jnthn: have fun!
jnthn Thanks :)
Zoffix lizmat: weird. Don't see anything iffy about the test. It's not timing based. So it's possible the failure might indicate the bug it covers isn't fully fixed :/ github.com/perl6/roast/blob/master....t#L50-L67 15:46
Zoffix sticks it into while t/fudgeandrun MISC/bug-coverage-stress.t; do true; done while doing @other-things 15:53
ZOFFLOP: t/spec/S11-modules/require.t (test 11) 15:54
AlexDaniel` goddamn require.t 15:55
Zoffix heh, well got a failure in it, but in test 1, not 3. and it's a doesn't-hang() test so it might've just took too long 15:56
ZOFVM: Files=1283, Tests=152770, 153 wallclock secs (21.36 usr 3.42 sys + 3228.20 cusr 165.37 csys = 3418.35 CPU) 15:58
AlexDaniel` quotable6: ZOFFLOP 15:59
quotable6 AlexDaniel`, OK, working on it! This may take up to three minutes (4412528 messages to process)
AlexDaniel`, Found nothing!
AlexDaniel` *facepalm*
Zoffix still no failures in MISC/bug-coverage-stress.t; killing the VM 16:21
c: 2017.09,HEAD say so 'abc' ~~ /:r a [ b+ || bc ] $ /; 16:39
committable6 Zoffix, ¦2017.09: «True» ¦HEAD(da97778): «False»
Zoffix So here's the golfed IRC::Client issue.
I've no idea which version is correct
moritz: do you know if the above is meant to match or not? How does the ratches thing works?
samcv: or you, do you know? I think this change was introed by your commit 16:40
*ratched 16:41
ZofBot: you so ratchet
ZofBot Zoffix, The fiend hath prick'd down Bardolph irrecoverable; and his face is Lucifer's privy-kitchen, where he doth nothing but roast malt-worms
AlexDaniel` Actually the change was introduced by smls I think 16:46
Zoffix Oh
AlexDaniel` maybe github.com/perl6/nqp/commit/263257...cdfc9c7a64 or github.com/perl6/nqp/commit/ba165c...8993251155
RT #130117 16:47
synopsebot RT#130117 [resolved]: rt.perl.org/Ticket/Display.html?id=130117 [TESTNEEDED] [REGEX] Sequential alternation `||` does not respect `:ratchet`
AlexDaniel` RT #131973
synopsebot RT#131973 [resolved]: rt.perl.org/Ticket/Display.html?id=131973 [TESTNEEDED] [REGEX] Backtracking modifiers on individual atoms fail to override a regex-global `:ratchet` modifier.
Zoffix Ohh
Well, then to throw a spanner into ratchet.. what about this case: 16:48
c: 2017.09,HEAD say so 'abc' ~~ /:r a [ b**2 || bc ] $ /;
committable6 Zoffix, ¦2017.09,HEAD(da97778): «True»
Zoffix How come this backtracks?
c: 2017.09,HEAD say so 'abc' ~~ /:r a [ b**{1..2} || bc ] $ /;
committable6 Zoffix, ¦2017.09: «True» ¦HEAD(da97778): «False»
AlexDaniel` oh, that's my ticket even! I've been searching for it…
Zoffix Oh, never mind
Well, I don't get how this works, but at least 1..2 range has same behaviour as + 16:49
Gonna fix IRC::Client in $n hours 16:51
ZofBot: take care, bruhg
ZofBot Zoffix, ''Ark at 'im! Calls 'isself a barman and don't know what a pint is! Why, a pint's the 'alf of a quart, and there's four quarts to the gallon
AlexDaniel` Zoffix: ok, well what part exactly? b**2 attempts to match two b's, so it never succeeds (and therefore tries bc)
Zoffix AlexDaniel`: but does it need to backtrack to do the "therefore" part? 16:52
And if not, how come b+ does?
AlexDaniel` c: 2017.09,HEAD say so 'abc' ~~ /:r a [ b+ || bc ] $ /;
committable6 AlexDaniel`, ¦2017.09: «True» ¦HEAD(da97778): «False»
AlexDaniel` b+ matches one b, so it moves forward
and fails on $ because there's c 16:53
Zoffix Ah, I get it now. Thanks
Zoffix &
AlexDaniel` so it seems that IRC::Client is wrong and rakudo on HEAD is right, but how many people were relying on the previous behavior? :S 16:54
buggable New CPAN upload: Color-Named-1.001001.tar.gz by HOLLI cpan.metacpan.org/authors/id/H/HO/...001.tar.gz 17:21
Zoffix ZofBot: killall adobe 17:40
ZofBot Zoffix, We are the dead
Zoffix What a stupid update system. It sits and waits for you to click "Update" and doesn't let you launch without updating, which of course results in you going in to launch an app to do $work and instead you sit and wait half an hour for an update to finish 17:41
[Coke] Zoffix: sounds like playstation update process. :) 17:42
Zoffix :)
buggable New CPAN upload: Color-Named-1.001002.tar.gz by HOLLI cpan.metacpan.org/authors/id/H/HO/...002.tar.gz 17:51
New CPAN upload: Color-Names-1.001003.tar.gz by HOLLI cpan.metacpan.org/authors/id/H/HO/...003.tar.gz
New CPAN upload: Operator-grandpa-1.001001.tar.gz by HOLLI cpan.metacpan.org/authors/id/H/HO/...001.tar.gz
moritz m: say so 'abc' ~~ /:r a [b+ || bc]/ 18:47
camelia True
yoleaux 18:12Z <Zoffix> moritz: why doesn't this pick the longest match of "if else"? m: my $if-st = rule { 'if' | 'if' 'else' }; say "if else" ~~ $if-st;
18:22Z <Zoffix> moritz: even more weird here where overridden <ws> doesn't match, but a brand new token with same definition matches: irclog.perlgeek.de/perl6/2017-10-20#i_15331627
moritz Zoffix: if the default ws terminates LTM 18:48
s/if//
timotimo when's the "oh no we missed something we have to build a point release" thing going to start this time :P
Zoffix Ohhh. 18:49
moritz m: say so 'abc' ~~ /:r a [b+ || bc] $/
camelia False
Zoffix moritz: is the list of what terminates LTM available somewhere?
moritz Zoffix: S05 should contain it, just a sec 18:50
Zoffix I should take the time to read all the S* stuff...
moritz design.perl6.org/S05.html#Longest-...n_matching
Zoffix moritz++ thanks 18:51
moritz S05 is quite enlightinging... when you read it the third or fourth time :-)
also, I don't know if the new semantics above are correct
ah yes, they are
never mind
b+ matches one b
thus || does not execute the second branch 18:52
Zoffix Thanks.
moritz and since no backtracking is allowed, the match remains "ab", and $ fails
the list in S05 might be missing (left) recursion as an LTM terminator 18:54
bartolin pmurias: I had to bootstrap nqp-j some days ago. if I'm not mistaken, you'd have to change the mapping of 'eqat' first (in Compiler.nqp) and then run 'make && make j-bootstrap-files'. Only with the new jars available in src/vm/jvm/stage0 you can change the usage of 'eqat' through the rest of the code and run 'make' again. in my case I made two commits: 11ef6ea37d for the changed bootstrap files, 7ca352fdf5 for the other changes 19:16
bartolin messed it up the first time, but found the solution in the logs somewhere 19:17
buggable New CPAN upload: Color-Scheme-1.001001.tar.gz by HOLLI cpan.metacpan.org/authors/id/H/HO/...001.tar.gz 19:46
Zoffix :o weabot 19:48
xi- hi
Zoffix ZofBot: what sort of a bot a weabot is?
ZofBot Zoffix, As O'Brien passed the telescreen a thought seemed to strike him
Zoffix hi 19:49
xi- I've actually made that nick a bot using your IRC::Client library on another network
Zoffix ha cool :)
Zoffix glances at #perl6 21:40
ZofBot: pah! Humans and their weird conversations! 21:41
ZofBot Zoffix, They are a form of pressure that you cannot withstand, even if you wished to
Zoffix Is that how they make diamonds?
timotimo Zoffix: how important is <tbody>? 21:46
Geth rakudo: MasterDuke17++ created pull request #1198:
Optimize INTERPOLATE even more		 21:47
Zoffix timotimo: in what context? 21:50
It's an optional element.
timotimo ok, maybe the question should be: when is it a good idea or necessary to use tbody inside a table?
Zoffix BAHAHAA 21:51
I just spent like 40 minutes hacking in :out-buffer into IO::CatHandle until I realized: NONE of the output methods on it are implemented :P 21:52
timotimo oh! wow
Zoffix By design. They're NYI
timotimo well, how exactly is CatHandle supposed to write stuff?
when does it switch? etc
Zoffix timotimo: it doesn't write anything. It's read-only
timotimo oh
well, then :)
Zoffix Oh, right yeah, that's why it's NYI
No clear semantics of how that should be done 21:53
timotimo: for <tbody>... No idea. There's this page developer.mozilla.org/en-US/docs/W...ment/tbody
"In conjunction with a preceding <thead> and/or <tfoot> element, the <tbody> element provides additional semantic information for devices such as printers and displays."
timotimo ah, ok
right, could be a target for css selectors 21:54
if you have multiple tr and maybe even td in what acts as your table header
Zoffix ZOFFLOP: t/spec/S11-modules/require.tt 22:14
ZOFFLOP: TODO PASS (3-4) t/spec/S12-construction/destruction.rakudo.t
I meant descruction.t 22:15
spelt right
Geth rakudo/nom: 765dd6944f | (Zoffix Znet)++ | 2 files
Add IO::CatHandle.out-buffer

Throw NYI. To block out IO::Handle.out-buffer

Powered by Cro and the Raku® Programming Language.

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

Thank you!