samcv MoarVM is about ready for release. just gonna get the changelog looked over once more, but it's complete 02:33
Zoffix \o/ 02:41
Geth nqp: 9a43e6c444 | (Samantha McVey)++ | t/nqp/019-file-ops.t
Revert "This test is now passing on OS X." for release

Refudge this test since it is still failing on TravisCI sometimes. This reverts commit 6c982a88f1e056cb44721525d4fb10e176854c06.		 04:42
samcv MoarVM has been released 05:40
[Tux] Rakudo version 2018.03-266-g79ed89ba4 - MoarVM version 2018.04-1-g577857ae0
csv-ip5xs0.879 - 0.893
csv-ip5xs-208.708 - 8.718
csv-parser36.326 - 36.522
csv-test-xs-200.438 - 0.467
test8.896 - 9.518
test-t2.471 - 2.485
test-t --race1.024 - 1.117
test-t-2045.167 - 45.345
test-t-20 --race15.976 - 16.191
 06:11
lizmat Files=1239, Tests=76308, 320 wallclock secs (15.79 usr 5.55 sys + 2201.73 cusr 219.47 csys = 2442.54 CPU) 09:22
brrt lizmat: can i ask how you compute a benchmark like that 09:27
i want to try out a change with potentially significant impact
lizmat make spectest; sleep 18000; make spectest
the first make spectest is to make sure all precomps have been done
the sleep 18000 is to make sure the second one runs around 5am in the morning, when the machine is cold and there's not a lot happening on line 09:28
then copy the line from the output
brrt oh, that seems simple enough 09:29
gfldex brrt: a positive significant impact would be very welcome :) 09:41
yoleaux 12 Apr 2018 16:54Z <jmerelo> gfldex: can you please tell me something about tags and why they were added? github.com/perl6/doc/commit/1509c4...ce5f695fcb ?
brrt heh, can't promise that
it's a tradeoff
to be specific, i'm going to add '-fno-omit-frame-pointer' 09:56
and the point of doing so is that i'm going to add a stack walker to find the current position in the JIT frame
and the point of doing that is that the VM can figure out where we are in the current execution of a JIT frame, without compiling code for storing that position in a special field every so many instructions 09:57
and the point of doing that, is that this is currently an extremely common operation, and the operations that require this information to be availalbe are not nearly as common
so it's exchanging one very common operation, relatively expensive operation, with a cheaper, somehwat less common operation; and another operation that *might* be more expensive but is really very uncommon 09:59
AlexDaniel good morning o/ 10:26
will cut rakudo release shortly™
lizmat ++AlexDaniel 10:27
El_Che AlexDaniel: \o/ 10:38
chansen_ m: say 0x4E00.uniname; 11:06
camelia <CJK Ideograph-4E00>
chansen_ m: say 0x17000.uniname;
camelia <Tangut Ideograph-17000>
chansen_ Is not the correct Unicode name, the correct name for CJK Ideographs is "CJK UNIFIED IDEOGRAPH-%.4X" and Tangut TANGUT IDEOGRAPH-%.4X. Looks like names are derrived from the meta ranges in UnicodeData.txt which is incorrect: <www.unicode.org/versions/Unicode10...44161>. 11:07
Zoffix chansen_: would you file a ticket? github.com/rakudo/rakudo/issues/new 11:10
lizmat hopes that nine or ugexe or tonyo would be able to write a blogpost about the PTS before next weeks weekly 11:13
chansen_ Zoffix: done 11:25
lizmat m: use Test; my %h = a => Mu; is-deeply %h, %h # disappointing :-( 11:26
camelia Type check failed in binding to parameter '<anon>'; expected Any but got Mu (Mu)
in sub _is_deeply at /home/camelia/rakudo-m-inst-1/share/perl6/sources/C712FE6969F786C9380D643DF17E85D06868219E (Test) line 651
in sub is-deeply at /home/camelia/…
Zoffix chansen_++ 11:29
Thanks
lizmat: that's cause of `eqv` not taking a `Mu` and it doesn't take it cause it wants to autothread junctions
m: my %h = a => Mu; say %h eqv %h
camelia True
Zoffix huh
m: my %h = a => Mu; say $%h eqv $%h 11:32
camelia Type check failed in binding to parameter '<anon>'; expected Any but got Mu (Mu)
in block <unit> at <tmp> line 1
Zoffix s: &infix:<eqv>, do { my %h = a => Mu; \(%h, %h)} 11:33
SourceBaby Zoffix, Sauce is at github.com/rakudo/rakudo/blob/79ed...p.pm6#L446
Zoffix s: &infix:<eqv>, do { my %h = a => Mu; \($%h, $%h)}
SourceBaby Zoffix, Sauce is at github.com/rakudo/rakudo/blob/79ed...p.pm6#L446
Zoffix Ah, the non-conted version goes through the eqaddr fast-past
lizmat well, not sure if it's a bug or not, but is makes it impossible for me to use is-deeply in tests for the new Hash::Util CPAN5 module 11:37
Zoffix I'd say there's a bugglet that `is-deeply` (and likely other routines containerize) stuff. Filed as R#1771 11:38
synopsebot R#1771 [open]: github.com/rakudo/rakudo/issues/1771 [LHF] Test.pm6 subs need to avoid containerizing arguments
Zoffix As for testing..... Yeah, that's LTA.
... but is expected if we offer is-deeply as "it does eqv" 11:39
m: my %h = a => Mu; say $%h ~~ $%h
camelia Type check failed in binding to parameter '<anon>'; expected Any but got Mu (Mu)
in block <unit> at <tmp> line 1
Zoffix boo
lizmat guess so :-)
maybe it's just a case of DIHWIDT 11:40
Geth star: hankache++ created pull request #110:
Update README		 12:01
star: 9cfbaec014 | (Naoum Hankache)++ (committed by Zoffix Znet) | README.md
Update README (#110)

  * Update README
  * Update README: Add more links to guides		 12:24
rakudo: f3a56190d0 | (Aleks-Daniel Jakimenko-Aleksejev)++ | tools/create-release-announcement.p6
Change some urls in announcement generation

https and less redirects		 12:29
rakudo: e96a38d32a | (Aleks-Daniel Jakimenko-Aleksejev)++ | 2 files
Log all changes (+ announcement)

Deliberately not logged: aa974ecd 32774e7b 3b86ffd5 6be586c5 04014cf6 e31dd966 ee5150e4 77c3a9c5 ec5416a9 05796698 a3a7c142 d443076d 19b3e52f 3420b67b a3e4f9e3 b5685109 c43817ac 5cba7c40 4032e62e 82a0e706 7353e055 f1b9c7da 9ff65564 318a9b09 95481e26 72c7c5bf 92eaa11a a2a2a745 ... (11 more lines)
rakudo: eb67e6d051 | (Aleks-Daniel Jakimenko-Aleksejev)++ | docs/release_guide.pod
Actual date, claim next release
timotimo `for for ... { Package::foo }` - is the first for supposed to be outside of the ``? 12:50
AlexDaniel: ^
AlexDaniel I don't think so… 12:51
AlexDaniel looks 12:52
right, anything else?
timotimo we might have to change a few things in the release document 12:54
ah, no
Zoffix AlexDaniel: oh, I think that was meant to be `do for`
timotimo false alarm; the github source code highlighting was turning parts of the document italic that weren't supposed to be
AlexDaniel yea I've seen that too
github has three different markdown processors (wiki, comments, files), but maybe the diff one is different too? 12:55
and they're all incompatible in edge cases 12:56
Zoffix Or maybe just one `for`
or maybe for `for`
AlexDaniel that's what I changed it to now, yeah, but not committing just yet :) 12:57
Zoffix c: 4402839a,4402839a^ unit package A; our sub need() {}; for <s> { A::need }
right no bot. ok
AlexDaniel they pinged out :( 12:58
c: 4402839a,4402839a^ unit package A; our sub need() {}; for <s> { A::need }
committable6 AlexDaniel, ¦4402839: «» ¦4402839a^: «Cannot find method 'Nil' on object of type List␤ in block at /tmp/ymPtL_fYCy line 1␤ in block <unit> at /tmp/ymPtL_fYCy line 1␤␤ «exit code = 1»»
AlexDaniel c: 4402839a,4402839a^,HEAD unit package A; our sub need() {}; for <s> { A::need }
committable6 AlexDaniel, ¦4402839,HEAD(eb67e6d): «» ¦4402839a^: «Cannot find method 'Nil' on object of type List␤ in block at /tmp/jSl4R87q4N line 1␤ in block <unit> at /tmp/jSl4R87q4N line 1␤␤ «exit code = 1»»
Geth rakudo: coke assigned to samcv Issue Incorrect Unicode character names github.com/rakudo/rakudo/issues/1770
09e86787b4 | (Aleks-Daniel Jakimenko-Aleksejev)++ | 2 files
 13:00
nqp: f62e04925e | (Aleks-Daniel Jakimenko-Aleksejev)++ | tools/build/MOAR_REVISION
[release] Bump MoarVM revision to 2018.04
 14:24
nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2......2018.04
afb18b4bb7 | (Aleks-Daniel Jakimenko-Aleksejev)++ | VERSION
rakudo: cb6230f8f8 | (Aleks-Daniel Jakimenko-Aleksejev)++ | tools/build/NQP_REVISION
[release] Bump NQP revision to 2018.04
rakudo: version bump brought these changes: github.com/perl6/nqp/compare/2018.......2018.04
829607d88e | (Aleks-Daniel Jakimenko-Aleksejev)++ | VERSION
Zoffix yeeey 14:25
releasable6: status
releasable6 Zoffix, No! It wasn't me! It was the one-armed man! Backtrace: gist.github.com/541d9ca6e9fc37f3c4...59628ebb90
Zoffix releasable6: next
releasable6 Zoffix, No! It wasn't me! It was the one-armed man! Backtrace: gist.github.com/df05e2b47ab5b626fc...e37ade9d21
AlexDaniel hold on something went wrong… 14:26
Zoffix "Cannot resolve the tag for the previous release"
AlexDaniel this may be an issue in releasable6 itself
but the sakefile failed to upload a tar to rakudo.org, looking at it now… 14:27
OK should be good now 14:28
releasable6: status
releasable6 AlexDaniel, No! It wasn't me! It was the one-armed man! Backtrace: gist.github.com/2d1a2c17844759c526...11ee360dfb
AlexDaniel grr I think it simply needs to pull
Zoffix AlexDaniel: so that's it? Time for appropriate amount of fun and mergeage of all the branches? 14:29
AlexDaniel well let's give releasable6 a minute or two to catch up
maybe it's trying to tell us something
but from my side everything looks correct 14:30
Zoffix releasable6: next 14:32
releasable6 Zoffix, No! It wasn't me! It was the one-armed man! Backtrace: gist.github.com/d322d9499d1ddcee5c...a182324e6a
AlexDaniel releasable6: next 14:34
releasable6 AlexDaniel, Next release in ≈24 days and ≈4 hours. 0 blockers. 0 out of 0 commits logged
AlexDaniel alright
Zoffix celebrates with the appropriate amount of fun
AlexDaniel \o/ 14:35
basically releasable6 is using the same thing that other bots use, and that doesn't move forward until there's an actual build generated…
so it takes a few minutes
I should fix that
El_Che it's out-out? 14:37
AlexDaniel yeah 14:42
El_Che \o/
Zoffix \o/
AlexDaniel++
El_Che let's get the package rolling (lots of changes :) )
If curious if it builds ok on varios distros: travis-ci.org/nxadm/rakudo-pkg/builds/371099494 14:46
AlexDaniel++, samcv++
Zoffix "This is Rakudo version 2018.04 built on MoarVM version 2018.04-1-g577857a" 14:48
*wailing guitar solos*
AlexDaniel oh crap 14:51
Geth rakudo: zoffixznet++ created pull request #1773:
Post release 2018.04
rakudo/master: 4 commits pushed by (Zoffix Znet)++
Zoffix AlexDaniel: waht?
AlexDaniel Zoffix: it shouldn't be that way no?
Zoffix The 2018.04-1-g577857a thing? That's cause I build master-master-master
I imagine it reads 2018.04 on normal builds
AlexDaniel ah, ok 14:52
AlexDaniel cancels the heart attack
Zoffix :)
jnthn Yay, release :)
yoleaux 13:26Z <brrt> jnthn: if you maybe have any experience with ml64 or soemsuch assemlber for microsoft
jnthn That thought is so terrifying that he didn't finish the sentence? :P 14:53
AlexDaniel the script would never do that obviously, but there's a way to force it to use a non-release version so for a second I crapped my pants thinking that I did that
brrt what, assembly on win64?
i fear nothing :-P
jnthn :-)
Geth roast: zoffixznet++ created pull request #419:
Post release 2018.04
roast/master: 4 commits pushed by (Zoffix Znet)++ 14:54
jnthn Anyway, I...think I touched it once, years ago...
tony-o timotimo: i.imgur.com/tiLmSYV.png 14:58
stmuk windows assembler? like tasm :-) 15:02
Zoffix ok :) 15:05
El_Che fedora will undo it soonish
AlexDaniel Zoffix++ # post-release branch
tony-o timotimo: i.imgur.com/tiLmSYV.png # not sure if that got lost, this is using zefpan and uxmal to generate module dependency charts 15:07
Zoffix AlexDaniel: to answer your question, I assume it's 'cause it got it listed twice; once for build-depends and other for depends: github.com/tadzik/Shell-Command/bl...son#L7-L14 15:08
Same with JSON::Marshall: github.com/jonathanstowe/JSON-Mars...on#L19-L30 15:09
tony-o: if there's a way to color-code arrows to show with depends they're from that'd be cool :)
AlexDaniel yes there is 15:10
in fac tyou can do different kind of arrows
like dashed and stuff
timotimo tony-o: how does it have two dependencies on File::Find from Shell::Command? maybe one is test-depends? 15:17
AlexDaniel can we include something like this on modules.perl6.org ?
I think the graph is very helpful 15:18
Geth rakudo: d53fe90acc | (Jonathan Worthington)++ | src/vm/moar/Perl6/Ops.nqp
Use new tryfindmeth op for the sink code-gen

It results in slightly smaller code, saving around 10KB on the compiled output of CORE.setting. More significantly, it saves a duplicate hash lookup in the unoptimized path for each sink (previously, we had one for `can` and one for `findmeth` on the same key). While it was true that spesh could deal with this in many cases, this gives us better ... (6 more lines)		 15:28
rakudo: c40fc5d711 | (Jonathan Worthington)++ | src/vm/moar/Perl6/Ops.nqp
Merge branch 'better-sink-code-gen'
Zoffix w00t 15:31
jnthn Plus for those not on #moarvm, I merged a NFA optimization there that measurements suggest get ~1% of CPU instructions off Perl 6 source parsing. 15:33
So, uh, enjoy the unperceptible compile time performance, I guess... :P
*improvement
(It's actually 13% off the time spent processing NFAs)
lizmat jnthn: I guess we should bump NQP/MoarVM ? 15:34
jnthn There's no hard dependency between what I pushed here and the stuff I pushed in MoarVM :)
But yeah, bump needed for people not building master/master/master to get the win :)
timotimo maybe wait a tiny bit with bumps 15:35
i expect other people will want to merge some branches now that the release happened?
lizmat ok, cool, I'll wait a bit
El_Che got fails on 386 15:37
travis-ci.org/nxadm/rakudo-pkg/jobs/371111042 15:38
(that's ubuntu 16.04)
this is 17.10 (latest): travis-ci.org/nxadm/rakudo-pkg/jobs/371111043
t/hll/06-sprintf.t ..................... Failed 2/286 subtests 15:39
Failed tests: 111-112
AlexDaniel not cool but ok let's fix that 15:50
El_Che: by the way, can you ticket these warnings?
oooooooh 15:52
so these are somewhat related to floating point precision right?
Zoffix [Coke]: are you around? Does TPF take taxes on grants? I'm filing my taxes and they're telling me I'm supposed to receive sone form 15:58
*some form
that says some stuff about taxes
FWIW, the US company that sells my books sends me form 1042-S with what they paid me and how much tax they took. But tax person said they've no clue if TPF would have the same form. 16:10
Had to pay $380 to canuk gov on my grant's income :p 16:11
timotimo :o 16:12
Zoffix ZofBot: taxation is theft!
Geth star/master: 5 commits pushed by (Steve Mynott)++ 16:22
AlexDaniel stmuk: fwiw there are some i386 failures 16:23
stmuk AlexDaniel: yes I noticed in the scrollback but thanks for the heads up in case I hadn't! 16:24
Zoffix AlexDaniel: what sort of failures? I got a 32-bit box handy 16:26
AlexDaniel Zoffix: just run nqp tests
Zoffix AlexDaniel: the sprintf stuff?
AlexDaniel yeah
Zoffix ok
Zoffix builds.
gonna take a while :}
AlexDaniel: on a hunch, there won't be a point release... 16:27
And it'll turn out to be a wrong test 16:28
AlexDaniel it could be so yeah, but I'm not that sure
after all the test works on some platforms 16:29
which one is it by the way?
Zoffix Do 32-bit boxes have 64-bit floats?
AlexDaniel: based on test numbers, it's these two: github.com/perl6/nqp/blob/master/t...#L164-L165 16:30
I'm still building to find out what the failing result is
it better not be a point a release... I don't wanna be the blamee for two point releases in a row :P 16:31
AlexDaniel if the test is wrong, do we need a point release fixing the test? I'm not sure… but I have no problem cutting it
ilmari Zoffix: double is 64 bits on my debian i386 VM
ilmari@sid-i386:~$ uname -m; perl -V:doublesize
i686
doublesize='8';
Zoffix Thanks
ilmari 'long double' is 12 bytes
El_Che A dot release would make my life easier 16:35
Zoffix # Actual value: <3.141e-20> 16:36
# Expected value: <3.142e-20>
# Actual value: <-3.141e-20>
# Expected value: <-3.142e-20>
m: say -3.141e-20
camelia -3.141e-20
lizmat ? 16:38
AlexDaniel lizmat: it's about the i386 failure 16:39
lizmat yeah, but it appears to also happen on Camelia
and that's not an i386 ?
AlexDaniel El_Che: speaking of which, is there any way you can make pre-release tests? What can we do on our side to make it easier? 16:40
e: say -3.141e-20
evalable6 -3.141e-20
lizmat fwiw, this could just be an indication that that i386 has some floating point bug (they were known for it at the time)
ilmari the x86 FPU has 80-bit registers, so sequencial FP ops might execute with higher precision, and only be rounded down at the end 16:42
Zoffix m: printf("%7.3e", 3.1415e-20)
camelia 3.142e-20
Zoffix nqp: printf("%7.3e", 3.1415e-20)
camelia Cannot invoke this object (REPR: Null; VMNull)
at <tmp>:1 (<ephemeral file>:<mainline>)
from gen/moar/stage2/NQPHLL.nqp:1585 (/home/camelia/rakudo-m-inst-2/share/nqp/lib/NQPHLL.moarvm:eval)
from gen/moar/stage2/NQPHLL.nqp:1822 (/home/camelia/rakudo…
Zoffix nqp: nqp::printf("%7.3e", 3.1415e-20) 16:43
camelia No registered operation handler for 'printf'
at gen/moar/stage2/QAST.nqp:1635 (/home/camelia/rakudo-m-inst-2/share/nqp/lib/QAST.moarvm:compile_op)
from gen/moar/stage2/QAST.nqp:5986 (/home/camelia/rakudo-m-inst-2/share/nqp/lib/QAST.moarvm:compile_node)…
Zoffix nqp: say(nqp::sprintf("%7.3e", 3.1415e-20))
camelia Directive e not applicable for type VMNull
at gen/moar/stage2/NQPHLL.nqp:2739 (/home/camelia/rakudo-m-inst-2/share/nqp/lib/NQPHLL.moarvm:panic)
from gen/moar/stage2/NQPHLL.nqp:2746 (/home/camelia/rakudo-m-inst-2/share/nqp/lib/NQPHLL.moarvm:bad-type-for…
Zoffix oh screw you
ilmari -ffloat-store or -fexcess-precision=… can be used to control that
Zoffix nqp: say(nqp::sprintf("%7.3e", [3.1415e-20]))
camelia 3.142e-20
El_Che AlexDaniel: pre-releases, sure.
you want travis tests or packages? 16:44
(I guess the former?)
AlexDaniel El_Che: personally I just want to see test status on all platform that you support
Zoffix I get right results from C, so I'd say it's our bug
El_Che cool, I am considering adding ARM, but I need a RP first :) 16:45
AlexDaniel note: release Bother El_Che to test his packages before the release 16:46
notable6 AlexDaniel, Noted!
Zoffix "Stage parse : 216.928" -_-
stmuk well moar builds on a PI3 .. not sure I've the patience to progress further! 16:48
Zoffix :)
Kaiepi zoffix: rakudobot supports inline::perl5 now and won't be failing because of git branch issues anymore 16:50
Zoffix Kaiepi++ awesome
AlexDaniel BeastieBot: all
BeastieBot [freebsd] Running complete Rakudo build and tests... (this will take a while)...
Kaiepi oh it's less noisy now 16:51
El_Che diner now. Course of action is wait for a .release?
AlexDaniel El_Che: yes
El_Che th
x
bbl
Zoffix $ ./perl6 -e 'use nqp; say(sprintf("%7.3e\n", 3.1415e-20))'
3.142e-20
AlexDaniel well unless you want to make 2018.04 and 2018.04.1 packages :)
Zoffix $ ./nqp-m -e 'say(nqp::sprintf("%7.3e\n", [3.1415e-20]))'
3.141e-20
Kaiepi it won't talk until all the commands are done instead of after each one
Zoffix So looks like the bug is in how nqp parses its nums.
AlexDaniel Kaiepi: ok good 16:52
BeastieBot [freebsd] Complete build and tests failed. See the output at fpaste.scsys.co.uk/577247 16:53
Kaiepi shit put make clean on the wrong line 16:54
Zoffix Damn. I did fix the bug where re-numed-str in Grammar/Actions had precision issues in Rakudo, but it didn't occur to me to ask "wtf do we have two different num-ifiers; in rakudo and nqp" :( 16:55
(like $n days ago)
Kaiepi BeastieBot, build 16:56
BeastieBot [freebsd] Building Rakudo...
[freebsd] Build failed. See the output at fpaste.scsys.co.uk/577248
Kaiepi yeah i shouldn't be putting make clean there
or rather i should be checking if make was already ran 16:58
PufferBot, all 16:59
PufferBot [openbsd] Running complete Rakudo build and tests... (this will take a while)...
Kaiepi PufferBot, all 17:03
PufferBot [openbsd] Running complete Rakudo build and tests... (this will take a while)...
Kaiepi looks like the submodule code i added still needs some work 17:09
for now i'll just configure rakudo and run the tests from the bot once ready 17:11
robertle with the changes to the rakudo.org site, is there still a way to get a list of source packages? in the past rakudo.org/downloads/nqp e.g. did list all nqp source packages... 17:13
Zoffix Kaiepi: FWIW last year we had a bot that could not only build rakudo but cut a release too. Maybe you'll find some of its guts useful for something: github.com/zoffixznet/na 17:14
robertle: what do you mean source packages?
El_Che robertle: rakudo.perl6.org/downloads/rakudo/ 17:15
robertle: rakudo.perl6.org/downloads/nqp
robertle: moarvm.org/releases
Zoffix rakudo.org/downloads/nqp lists those: i.imgur.com/3MpFPIh.png 17:16
robertle perfect, thanks
AlexDaniel robertle: if you're working on debian packages, then perhaps you want to wait a bit for the point release 17:19
robertle: because we discovered an issue on i386
robertle ah, that's what I was doing!
bug in what? moar? 17:20
Zoffix likely in nqp
robertle k
Zoffix but it's not yet 100% known
El_Che robertle: are you part of the debian team packaging rakudo? 17:21
robertle dude, s390x and mips were weird eough, but i386! really is anyone even using that? ;)
El_Che: yup
El_Che robertle: lots of 32-bit cloud images out there
and VMs 17:22
AlexDaniel robertle: we're certainly not pretesting anything on that, yes…
Zoffix likely in moarvm 17:28
AlexDaniel ooo fancy
Kaiepi zoffix it doesn't look very useful for rakudobot since they handlle builds completely differently 17:30
but supporting nqp stuff like it does would be pretty cool
PufferBot, test
PufferBot [openbsd] Running tests...
Kaiepi bah, idk what screwed with the commands since they were working fine a couple days ago 17:33
stmuk I'm seeing a couple of roast failures on FreeBSD 11/clang t/spec/S11-modules/require.t and t/spec/integration/advent2011-day23.t 17:35
Zoffix t/spec/S11-modules/require.t is a frequent flopper 17:37
stmuk yeah I just saw it pass in isolation 17:39
Kaiepi oh lol i forgot to update the attribute names in a superclass 17:42
stmuk github.com/perl6/roast/issues/420 17:43
. o O ( 420 dude!)
Zoffix :) 17:44
ok, maybe not moarvm.... Having trouble finding where num parsing in NQP is at
All I see is `method dec_number($/) { make +$/; }` so then I dumped ASTs for say(+"3213.322e0") and it showed calling numify and I traced that to MVM_coerce_smart_numify in MoarVM, but if I stick debug prints, the numified number doesn't appear anywhere. 17:47
AlexDaniel stmuk: so is that one consistent? 17:49
El_Che giving rakudo a go on solaris: 17:54
Perl v5.10.1 required--this is only v5.8.4, stopped at ./Configure.pl line 4
:)
Zoffix jesus
El_Che Solaris 10, not 11 17:55
but even so
AlexDaniel :S
stmuk AlexDaniel: yes
nine El_Che: is it still the 90s where you live?
stmuk El_Che: I noticed go/sparc64 isn't being maintained past 1.8
:)
El_Che nine: it was a bug part of my life 17:56
:)
big
nine El_Che: same here :)
Zoffix ah, k, found the location 17:57
El_Che ./Configure.pl --prefix=/home/claudio/rakudo/`uname -m` --backend=moar --gen-moar && /opt/csw/bin/gmake && /opt/csw/bin/gmake test && /opt/csw/bin/gmake install 17:58
/usr/bin/env: No such file or directory 17:59
so, we hardcode the env location it seems :)
AlexDaniel stmuk: fwiw typically reports for failing tests should be filed against rakudo and not roast (after all it's rakudo that's failing the test)
unless there's a test that's obviously wrong
El_Che we hardcode everything, it seems: if (not -X '/usr/bin/gmake') { ... 18:01
stmuk AlexDaniel: ok 18:02
nine I've got one reason for those unhandled "No such symbol" failures in require 18:03
stmuk I was just happy it wasn't RT :)
Zoffix nqp: say(5e-324)
camelia 0
Zoffix AlexDaniel: so are there are any known *rakudo* problems stemming from the nqp test failures? Also, do nqp failures affect anything (e.g. packagers?) 18:04
AlexDaniel Zoffix: we don't know yet if there are any problems, it's probably unlikely but still possible 18:06
Zoffix: and nqp test failures are annoying for packagers
well, I mean people working on packages
Zoffix right 18:07
Well, it's gonna take awhile to fix it. I'd guess. :)
AlexDaniel it's ok 18:08
Zoffix nqp: say(.1e0+.2e0)
camelia 0.30000000000000004
AlexDaniel I'm also interested to know what's exactly wrong with github.com/perl6/roast/issues/420 18:09
because maybe that's something we can fix in the point release also
assuming this failure didn't exist previously (I don't know) 18:10
Kaiepi: how much space do you have available for the bot?
Kaiepi: like, potentially?
Kaiepi: and can you reproduce github.com/perl6/roast/issues/420 by any chance?
Kaiepi i could fit rakudo, nqp, moar, roast in there 18:11
that test passes on openbsd 18:12
lemme check on my vps 18:13
AlexDaniel my definition of “fit” is slightly different :)
do you have at least 15G? :) 18:14
Kaiepi sure
AlexDaniel Kaiepi: I'm just thinking, why not run a subset of *ables on freebsd/openbsd
so that we can test different versions, bisect, etc. 18:15
maybe that's not useful enough, I don't know
but then bisectable itself wasn't considered useful before it was born :)
Kaiepi the more important question is do i have enough ram 18:17
AlexDaniel how much do you have?
Kaiepi 8GB, i use 4-6 at a time when i'm not testing stuff in vms
my vps has 2GB
stmuk AlexDaniel: I've a theory about the failing test which suggests its not important
AlexDaniel bisectable6: uptime 18:18
bisectable6 AlexDaniel, 2 days, 19 hours, 10 minutes, and 37 seconds, 326.929688MiB maxrss. This is Rakudo version 2018.03-263-gd4a6b92f3 built on MoarVM version 2018.03-126-g876aa90ee implementing Perl 6.c.
AlexDaniel Kaiepi: maybe we should really focus on memory footprint… >300 MiB is indeed too much for a simple bot 18:19
stmuk AlexDaniel: should be able to confirm that or not soon
nine Ok, I have a patch that seems to make the Failures in S11-modules/require.t go away. But I'm not sure it's the right thing to do.
AlexDaniel stmuk: cool! Thanks!
nine: fwiw it's not going to affect the point release so you can go ahead and commit it
Kaiepi rakudobot uses 150MB of memory 18:20
nine Turns out, the Failures may actually be correct: the file loaded in "require S11-modules::NoModule <&bar>;" doesn't contain a module with that name. But require always generates an indirect lookup of the module name, even when no one is interested in require's return value. 18:21
So I've simply wrapped it in a QAST::Want to only generate that lookup when someone's....looking.
But in the case of the other two error messages, the modules actually should exist. Yet my patch makes the Failures go away, too. 18:23
So the patch may hide a further issue
AlexDaniel mmm 18:24
Geth rakudo: nxadm++ created pull request #1774:
Don't hardcode gmake locations		 18:26
El_Che AlexDaniel: could you provide e.g. rakudo.perl6.org/downloads/rakudo/...ase.tar.gz file as canary? The repo would download every day if there, otherwise you can always trigger it 18:31
stmuk grrr I need a Threadripper
El_Che does that sound reasonable?
AlexDaniel El_Che: yes
El_Che: maybe not on that url but I'll let you know once I have something 18:32
El_Che sure 18:33
Zoffix .tell jnthn we need an "executive" decision: **nqp**'s numerics are parsed with `MVM_coerce_s_n`. It has precision issues, such as parsing 3.145e-20 into 3.14149999999999969e-20 instead of the closest representable num 3.14150000000000029e-20 and it also can't handle denormals (5e-324 parses as 0). This makes 2 sprintf nqp tests fail on 32bit boxes and it's currently proposed to fix that and make a point
yoleaux Zoffix: I'll pass your message to jnthn.
Zoffix release (they didn't fail before likely due to limited precision num stringification in the past).... Looking at fixing it, I'm assuming we'll get hit with the same 14x perf loss we did with fixing the same problem in HLL nums. So I'm wondering: is it even worth it? Rakudo's nums aren't affected so it's only NQP num parsing that's affected. Is anything else using `MVM_coerce_s_n`? I'm kinda thinking of
fuding those failing tests a bit and leaving this under the rug until any real problems actually show up due to this. What do you think?
.tell jnthn rest of that message: irclog.perlgeek.de/perl6-dev/2018-...i_16094868 (it's fairly long)
yoleaux Zoffix: I'll pass your message to jnthn.
El_Che also: do you want a prelease file for moarvm and nqp or will rakudo download the correct one with --gen-moar?
Geth rakudo: 666eb3abe7 | (Stefan Seifert)++ | src/Perl6/Actions.nqp
Fix spurious unhandled Failures caused by require

We compiled a require unconditionally to the actual module loading code plus an indirect lookup of the loaded module. If the loaded comp unit does not actually contain a module with that name the lookup would fail. However we don't sink require's return value so the Failure would escape. Fix by only compiling the indirect lookup when someone's actually interested in the result. Seems to fix GH #1332		 18:34
synopsebot RAKUDO#1332 [open]: github.com/rakudo/rakudo/issues/1332 [⚠ blocker ⚠] Spurious Failures in successful `require`
Zoffix jnthn: to clarify: it's **nqp**'s spritnf fails because of how the num argument is given to it. It doesn't affect rakudo's sprintf's behaviour
s/how .+? \. <(/ is parsed/; 18:35
AlexDaniel El_Che: I think it should just use --gen-moar to pull the right one, yes
El_Che cool
preparing a travis file
Zoffix s/fuding/fudging/; 18:36
El_Che repo should be this one: github.com/nxadm/rakudo-pkg-canary
stmuk I've updated the 420 issue .. its just a simple path or cwd issue with running roast under star and not OS related 18:41
I can fix in a few hours if noone beats me to it 18:42
AlexDaniel okay 18:44
Zoffix camelia: help 18:45
camelia Zoffix: Usage: <(nqp-jvm|debug-cat|rakudo-moar|star-m|prof-m|rakudo-jvm|nqp-moarvm|nqp-js|p5-to-p6|sm|nqp-mvm|j|nqp-m|star|perl6|r-jvm|rakudo|rj|p6|m|r-m|r|master|nqp|p56|nqp-q|rm|r-j)(?^::\s(?!OUTPUT)) $perl6_program>
Zoffix nqp-jvm: test
camelia Error occurred during initialization of VM
Unable to allocate 2648KB card tables for parallel garbage collection for the requested 1355776KB heap.
Error: Could not create the Java Virtual Machine.
Error: A fatal exception has occurred. Program will exit.
Zoffix uuh... ok then 18:46
stmuk pl6anet.org/drop/rakudo-star-2018.04-pre0.tar.gz 18:49
RC0 to follow after the NQP fix
Zoffix or maybe I'm talking shit... strtod("3.141500", NULL) parses stuff just fine 18:50
or use the Gay atoa 18:51
:) 18:52
As in David M. Gay's dtoa.c
oh wait, that's going the other way. 18:53
jnthn Zoffix: So far as NQP goes, honestly, nums are really unimportant. :) 18:55
Zoffix: Compilers are all ints and pointers, and NQP's primary purpose is writing those and related things.
I don't think taking the performance hit to get things correct would actually affect anything negatively.
And if we can unify the code-path that NQP and Rakudo use, that's a win 18:56
Zoffix Cool. Then I'll fix it to give correct answers.
jnthn Zoffix++ 18:57
I've kinda pondered making NQP do int math by default unless you're actually dealing with stuff declared as `num`
timotimo yeah, we had fun times when things were going with num all the time 18:59
nqp's still not terribly good at that 19:00
Kaiepi ok so the reason rakudobot wasn't working earlier was because i handled abstracting the async logic badly 19:07
from a couple days ago
locking mutexes inside promises are how true experts handle threading 19:10
timotimo if you're using a "start"ed code block just as a lightweight worker thread, that would be fine 19:12
Kaiepi that's what i use them for, though i'm joking about using locks from inside the promise 19:15
PufferBot, all 19:18
PufferBot [openbsd] Running complete Rakudo build and tests... (this will take a while)...
Kaiepi ok nothing failed, even though it disconnected in the middle of the stress test 19:40
stmuk AlexDaniel: The test file README.md is absent from 6.c-errata (which is used by Rakudo Star) and should be added. 19:44
AlexDaniel stmuk: ok but there's “README”? 19:49
travis-ci Rakudo build passed. Zoffix Znet 'Merge pull request #1766 from MasterDuke17/condense_IO_Path.new_multis 19:50
travis-ci.org/rakudo/rakudo/builds/371208963 github.com/rakudo/rakudo/compare/6...baa29011a0
stmuk AlexDaniel: if you grep for README.md in 6.c-errata its used by a number of tests but the file itself is abscent on that branch 19:51
AlexDaniel ohhh…
IMO it would be better to fix those tests to not use README file whatsoever 19:55
otherwise it's too fragile
stmuk I'm not sure of the best fix
I imagine its basically that errata and master are desynced
AlexDaniel stmuk: just create a temp file, write some data into it and then read from it 19:58
the test itself is from this post: perl6advent.wordpress.com/2011/12/...ic-perl-6/
stmuk or just use LICENCE
lizmat m: dd chdir "perl5" # works as expected 19:59
camelia IO::Path.new("/home/camelia/perl5", :SPEC(IO::Spec::Unix), :CWD("/home/camelia"))
lizmat m: dd "perl5".IO.chdir # does not, Zoffix, opinions ?
camelia Cannot resolve caller chdir(IO::Path: ); none of these signatures match:
(IO::Path:D $: $path, :$test!, *%_)
(IO::Path:D $: IO $path, |c is raw)
(IO::Path:D $: $path is copy, :$d = Bool::True, :$r, :$w, :$x, *%_)
in block <unit> …
lizmat feels to me there's a .chdir candidate missing in IO::Path ?
stmuk AlexDaniel: whats the normal workflow with the test branches? 20:00
Zoffix lizmat: we explicitly removed it, didn't we? 20:01
AlexDaniel stmuk: commit to master, if there's a good reason for that change to be in 6.c-errata then someone will have to cherry-pick it
stmuk: if that's what you're asking
Zoffix e: dd chdir "perl5".IO 20:02
evalable6 Failure.new(exception => X::IO::Chdir.new(path => "/home/bisectable/git/whateverable/perl5", os-error => "does not exist"), backtrace => Backtrace.new)
Zoffix e: dd chdir ".".IO
evalable6 IO::Path.new("/home/bisectable/git/whateverable", :SPEC(IO::Spec::Unix), :CWD("/home/bisectable/git/whateverable"))
Zoffix e: dd ".".IO.chdir
evalable6 (exit code 1) Cannot resolve caller chdir(IO::Path: ); none of these signatures match:
…
Zoffix, Full output: gist.github.com/27edd966e6cb48c584...99cf5e1473
stmuk so I guess cherry pick README.md to 6.c-errata?
[Coke] Zoffix: All the payments & tax implications are handled by the treasurer. I will send the two of you an email to start the thrad. 20:03
*thread
AlexDaniel stmuk: well, we don't want to have that README.md in 6.c-errata actually, they're deliberately different 20:04
they can both be .md though
Zoffix lizmat: ah, now I see. We used to have .chdir that takes a path to chdir to: github.com/rakudo/rakudo/blob/mast...#L404-L457 . And that one is now deprecated. And just a "foo".IO.chdir without args. Yeah, I guess it makes sense as an alternative to the &chdir interface—except for the whole business of it clashing with the deprecated version of .chdir
[Coke]: thanks.
AlexDaniel so one option is to markdownify 6.c-errata README of course, but I really think that this test should be changed
it's just not nice to assume that some semirandom file exists in the directory you're in… 20:05
stmuk or change every reference to README.md in master to LICENSE and cherry pick to errata?
lizmat Zoffix: ah, so a bare .chdir() is deprecated in Perl 6
?
AlexDaniel yes, that's possible
Zoffix AlexDaniel: stmuk just create a different dummy file for the test? `use lib $?FILE.IO.parent(2).add: 'packages'; use Test::Util; my $test-file := temp-file :content("blah\nblah\nblah\n")`
lizmat: I think it never existed. It's the .chdir($path-to-chdir-to) is deprecated. 20:06
AlexDaniel Zoffix: I think you meant make-temp-file
Zoffix yeah
Or even just use $?FILE
m: say $?FILE 20:07
camelia <tmp>
Zoffix m: say $?FILE.^name
camelia Str
AlexDaniel m: say $?FILE.chop
camelia /tmp/evalbot-file-dAYzMfFFa
AlexDaniel *magic*
stmuk I suppose given the different branches contain different files we should discourage use of such hard coded filenames anyway
AlexDaniel yes
stmuk $?FILE sounds easy
AlexDaniel stmuk: I assume a PR is coming? 20:29
stmuk yes I can do that 20:30
AlexDaniel stmuk++
Geth roast: stmuk++ created pull request #421:
replace use of README.md (which doesn't exist on 6.c-errata) by $?FILE		 21:10
samcv i merged the MoarVM PR that randomizes our hash seed. i'll bump nqp and rakudo and we can run toaster and see if anything fails 21:12
stmuk note packages/RT125245.pm doesn't actually test anything
synopsebot RT#125245 [resolved]: rt.perl.org/Ticket/Display.html?id=125245 [BUG] Pre-compiled MoarVM-code segfaults on type-checking
jnthn stmuk: It's used from another test file to make sure the subset type doesn't explode on type-check in pre-comp 21:20
stmuk jnthn: ok I see S10-packages/precompilation.t but I am surprised it doesn't fail on errata since the file its based on doesn't exist on that branch 21:26
samcv hmm is Geth taking a nap? 21:29
lizmat: anyway rakudo and nqp are bumped
lizmat samcv++
stmuk similarly I wouldn't expect integration/advent2011-day24.t to pass on errata 21:30
anyway I don't know what's going on in either case but seems a code smell to me 21:34
El_Che there seem to be some netwerk issues at the docker hub 21:42
AlexDaniel: I have a +- feature complete canary, waiting on the Docker hub :)
AlexDaniel El_Che: alright, let's try it maybe: rakudo.perl6.org/downloads/prerelease/ 22:07
these files right there are identical to the release
right now
AlexDaniel kicks off releasable to generate new ones from head 22:08
El_Che AlexDaniel: I made it location/version/name independent: github.com/nxadm/rakudo-pkg-canary...vis.yml#L4 22:09
AlexDaniel errr that looks hardcoded? 22:10
AlexDaniel doesn't understand
dogbert17 ZOFFLOP: t/spec/S07-hyperrace/basics.t 22:11
El_Che well, you can put "rakudo.perl6.org/downloads/prerele...se.tar.gz" in that env value and leave it there, and push the build by touching something else
or if you want to test something else, you can change the url
that's also a push
whatever you do, every 24h a build is triggered from master 22:12
(or you can use a branch with a different url, etc, etc)
AlexDaniel El_Che: sorry, so what should I do exactly now? 22:18
I thought I'd be pushing tars to that url and travis will be triggered daily
El_Che (it doesn't work yet, buildig imgs and removing some errors)
AlexDaniel: yes, that will happen 22:19
AlexDaniel El_Che: oh ok
El_Che but you can change the URL in case you want to test something else
I put the 03 version there so I can test if it works :)
AlexDaniel anyway, I added `upload-canary` target to the release sakefile (WHICH I SHOULD PUBLISH DAMMIT), so that's gonna be very easy from now on 22:20
El_Che if you're releasing you can update a bogus file to trigger the build
AlexDaniel El_Che: can't I just hit restart in travis? 22:21
but that's only going to work when I get push access to that repo right?
El_Che AlexDaniel: I haven't figured out how to give you access
Geth rakudo: 6c1983a97c | (Samantha McVey)++ | tools/build/NQP_REVISION
Bump MoarVM/nqp

MoarVM Changes: 2018.04..2018.04-16-g24a91a460 24a91a460 use official API for thread lock count ffdd62d48 debugserver: allow unrecognized fields in messages b87b03e23 throw out superfluous debugspam 7d83f32bf Make a few of the MVM_getrandom comments more clear ... (9 more lines)
¦ rakudo: version bump brought these changes: github.com/perl6/nqp/compare/2018....gbc0b5e100
El_Che maybe once added to collaborators or to an org
samcv looks like geth is done napping
El_Che not clear yet
Zoffix samcv: looks like some glitch. Re-delivering the hook still gave 500, but restarting the bot made it work again
Geth nqp: bc0b5e100d | (Samantha McVey)++ | tools/build/MOAR_REVISION
Bump MoarVM

Changes: 2018.04..2018.04-16-g24a91a460 24a91a460 use official API for thread lock count ffdd62d48 debugserver: allow unrecognized fields in messages b87b03e23 throw out superfluous debugspam 7d83f32bf Make a few of the MVM_getrandom comments more clear ... (9 more lines)		 22:24
nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...g24a91a460
81373e88f2 | (Samantha McVey)++ | 2 files

Update the previous test which was incorrect and was failing after the change. Add new tests for Tangut and Nushi codepoint names as well.
Changes: 2018.04-16-g24a91a460..2018.04-18-ge7d79d869 e7d79d869 Fix issues with previous commit overzealously removing brackets 5728e1f39 Fix CJK Unified Ideograph and other derived Unicode names
El_Che AlexDaniel: do repos on the perl6 or rakudo org have shared access to travis? 22:29
AlexDaniel yes 22:30
El_Che then maybe moving the canary to an org may be the way to go then
AlexDaniel I think givin Admin access should also work maybe
oh, I guess there's no such thing 22:31
El_Che I only see a Collaborators setting (give push access)
and transfer ownership
AlexDaniel right… that's for non-org repos
move it to perl6 org then
El_Che ok, I will when it works properly 22:32
AlexDaniel ♥ 22:33
El_Che however, I need to test that. The containers are built from the link with my github user on github.
I'll figure it out
AlexDaniel samcv: is there a test failing somewhere? 22:35
t/nqp/106-unicodenames.t ............... Failed 1/31 subtests
samcv yeah i haven't bumped moarvm yet
i just made a fix for that unicode bug 22:36
oh oops yeah actually that is my mistake 22:40
but yeah i have not yet bumped moarvm yet so, will have that fixed before i bump it with unicode fix 22:41
j: 0x135CD.uniname.say 23:13
camelia <unassigned>
samcv j: 0x17000.uniname.say 23:14
camelia <unassigned>
samcv j: 0x1B170.uniname.say
camelia <unassigned>
lizmat hmm... after the last update of rakudo, it looks like mi6 (App::mi6) is broken :-( 23:26
gist.github.com/lizmat/f7ce01f608c...ce2b4286ae 23:27
lizmat goes to bed
timotimo wow, that looks interalsy 23:28
samcv timotimo: maybe related to the NFA? i'm getting asan errors on that 23:29
since jnthn's recent merge
AlexDaniel but I did not see that in toaster? Hmm… 23:32
it was failing on 2018.03 :) 23:33
or at least it did fail in that particular case… 23:34
and then it was green on 2018.03-260-gfa979c922
well, if we're talking HEAD, then it's different :) 23:35
samcv AlexDaniel: it was merged after release i thought? 23:42
AlexDaniel yea
rakudo: 94b4783dfc | (Samantha McVey)++ | 2 files
Bump MoarVM/NQP and update Unicode name tests for changes

MoarVM Changes: 2018.04-16-g24a91a460..2018.04-18-ge7d79d869 e7d79d869 Fix issues with previous commit overzealously removing brackets 5728e1f39 Fix CJK Unified Ideograph and other derived Unicode names		 23:57

Powered by Cro and the Raku® Programming Language.

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

Thank you!