travis-ci Rakudo build passed. Elizabeth Mattijsen 'Make Str.rindex about 30% faster 00:03
travis-ci.org/rakudo/rakudo/builds/145277854 github.com/rakudo/rakudo/compare/f...fe3c04ed47
gfldex I used to work for a former subsidiary for Lufthansa and we had one gentleman who was still working for Lufthansa proper as a manager. He used to carry the name Süß but after having to travel a lot to northern america he changed his name to Suess. At times he was even on lists because of his uncomputable name. 00:10
s/was/wasn't/ 00:11
m: say "in Massen".fc ~~ "in Maßen".fc 00:24
camelia rakudo-moar 980f49: OUTPUT«True␤»
gfldex :-/
that means that "a lot" is the same as "in moderation" 00:25
Zoffix $ perl -wlE 'say fc "in Massen" eq fc "in Maßen" ? "Yeah" : "Nope"' 00:26
Nope
m: "ß".fc.say
camelia rakudo-moar 980f49: OUTPUT«ss␤»
Zoffix m: "ß".fc.uninames.say
camelia rakudo-moar 980f49: OUTPUT«(LATIN SMALL LETTER S LATIN SMALL LETTER S)␤»
Zoffix Oh, nm, I forgot use utf8 in Perl 5. It also says the two are equal. 00:27
gfldex that means it's wrong in Perl 5 in the same way 00:28
that also means that for german texts you have to scan for ß and special case it 00:29
Zoffix m: say "in Massen".fc ~~ "in Maẞen".fc 00:30
camelia rakudo-moar 980f49: OUTPUT«True␤»
gfldex m: say "ẞ".uninames.say 00:36
camelia rakudo-moar 980f49: OUTPUT«(LATIN CAPITAL LETTER SHARP S)␤True␤»
travis-ci Rakudo build passed. Zoffix Znet '.polymod with a lazy list must not lose mods if the list runs out
travis-ci.org/rakudo/rakudo/builds/145282087 github.com/rakudo/rakudo/compare/f...0f496fa8be
Zoffix shush
gfldex m: say "ẞ".uninames.say # called "Versaleszetts" in german 00:37
camelia rakudo-moar 980f49: OUTPUT«(LATIN CAPITAL LETTER SHARP S)␤True␤»
gfldex m: say "Haßloch".fc 00:38
camelia rakudo-moar 980f49: OUTPUT«hassloch␤»
gfldex context: de.wikipedia.org/wiki/Ha%C3%9Floch
i wouldn't want to live in hatehole 00:39
Zoffix Jesus swift's docs are shit 00:42
Can't find a list of methods on a String
(╯°□°）╯︵ ┻━┻ 00:45
Wanted to see what it gives for "in Massen".fc ~~ "in Maẞen".fc but I can't even find a .foldcaseString method
gfldex Zoffix: maybe developer.mozilla.org/en-US/docs/W...eLowerCase 00:47
Zoffix Well, there's stringByFoldingWithOptions but I've no idea how to use it. 01:08
And I can't even find one working example of swift code.... Either that or glot.io/new/swift is busted. 01:09
Zoffix chucks Swift into "annoying languages" bin
.uppercaseString ~_~ who thought of such a verbose name
So basically "in Massen".fc ~~ "in Maẞen".fc => True makes sense. Since it erases case, there's no little s or big S, there's no case so both of them are the same 01:20
m: dd 'ß'.fc, 'ẞ'.fc, 'SS'.fc, 'ss'.fc 01:21
camelia rakudo-moar 980f49: OUTPUT«"ss"␤"ss"␤"ss"␤"ss"␤»
Zoffix m: say so 'assen' ~~ 'aẞen' 01:22
camelia rakudo-moar 980f49: OUTPUT«False␤»
Zoffix m: say so 'assen' ~~ 'aẞen'.fc
camelia rakudo-moar 980f49: OUTPUT«True␤»
Zoffix m: say 2.Rat.polymod: 0 02:19
camelia rakudo-moar 980f49: OUTPUT«Attempt to divide 2 by zero using polymod␤ in any at /home/camelia/rakudo-m-inst-2/share/perl6/runtime/CORE.setting.moarvm line 1␤ in block <unit> at <tmp> line 1␤␤»
dalek kudo/nom: 7976fe9 | (Zoffix Znet)++ | src/core/ (2 files):
.polymod with zero divisor must not reference guts in error

Fixes RT#128646		 02:23
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=128646
dalek ast: c61b572 | (Zoffix Znet)++ | S32-num/polymod.t:
.polymod with zero divisor does not reference guts in error

RT#128646
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=128646
lizmat bisectable: good=2015.12 bad=HEAD my int $i = 10; say Buf.new($i +& 0xFF) 10:22
bisectable lizmat: No build for 'bad' revision
lizmat bisectable: good=2015.12 bad=b06a99ebdd79393ce109cc214a6019ada5a0218a my int $i = 10; say Buf.new($i +& 0xFF) 10:23
bisectable lizmat: bisect log: gist.github.com/c8804ae8cbbe7200f7...cfd72517d0
lizmat: (2016-07-11) github.com/rakudo/rakudo/commit/de5d9e7
lizmat ok, so what's weird about this is that the error message actually states:
"expected uint8 but got Int"
which means that this: 10:25
m: use nqp; say nqp::istype(uint8,Int)
camelia rakudo-moar 7976fe: OUTPUT«1␤»
lizmat returned 0 instead of 1
psch m: for &infix:<+&>.candidates -> $cand { try say Buf.new($cand(my int $ = 10, 255)).WHAT if \(my int $, 0) ~~ $cand.signature } 10:38
camelia rakudo-moar 7976fe: OUTPUT«(Buf)␤(Buf)␤(Buf)␤»
psch m: for &infix:<+&>.candidates -> $cand { try say Buf.new($cand(my int $ = 10, 255)) if \(my int $, 0) ~~ $cand.signature }
camelia rakudo-moar 7976fe: OUTPUT«Buf:0x<0a>␤Buf:0x<0a>␤Buf:0x<0a>␤»
psch oh, that just sinks the 255 'cause the my gobbles it i think 10:39
m: for &infix:<+&>.candidates -> $cand { try say Buf.new($cand((my int $ = 10), 255)) if \(my int $, 0) ~~ $cand.signature }
camelia rakudo-moar 7976fe: OUTPUT«Buf:0x<0a>␤Buf:0x<0a>␤Buf:0x<0a>␤»
psch hm, maybe not. i don't get it :S
i guess it could use a different candidate for infix:<+&> for some reason..? 10:40
nine lizmat: $i +& 0xFF returns an Int, even if both arguments are ints
lizmat yes, which *should* make nqp::istype($a,Int) return a 1, instead of a 0 10:41
so, this is something *really* weird 10:42
dalek ar: 69c56ad | (Steve Mynott)++ | docs/announce/2016.07.md:
roll in some of the rakudo changelog from 2016.07
 14:23
timotimo we still have "none-multi's" in there 14:24
dalek ar: 9c3b904 | (Steve Mynott)++ | docs/announce/2016.07.md:
tweak formating
 14:27
ar: 52be35f | (Steve Mynott)++ | modules/ (10 files):
bump submodule versions
 14:31
kudo/nom: c889250 | (Zoffix Znet)++ | docs/ChangeLog:
Fix typo
 14:42
Zoffix timotimo, I saw you mentioning it in #perl6, which made no sense: irclog.perlgeek.de/perl6/2016-07-15#i_12847397 14:43
I didn't even realize you were talking about the Rakudo's ChangeLog
dalek ar: a2e2f2c | (Steve Mynott)++ | docs/announce/2016.07.md:
changelog for submodules
 14:56
Zoffix Does anyone know the story of why diag() prints to STDERR? It does so in Perl 5 too, but I'm reading testanything.org/tap-specification.html and it has no mention of STDERR and in fact, states "A harness must only read TAP output from standard output and not from standard error." 14:59
dalek kudo/nom: 20a648a | niner++ | src/core/CompUnit/PrecompilationRepository.pm:
Don't die generating a debug message if source file is unavailable
 15:00
kudo/nom: d1604b5 | niner++ | tools/build/install-core-dist.pl:
Fix make DESTDIR not finding CompUnit::Repository::Staging
nine Zoffix: maybe diag() is not considered TAP? 15:05
Zoffix: it's purpose is to communicate something to the user after all and not to the harness 15:06
Zoffix Ah
mst exactly what I just said when he cross-spammed the question to #perl
Zoffix Sorry
mst in this case, I think getting the same answer twice is no bad thing
note also that #perl-qa is the best channel for this sort of question, followed by #toolchain 15:07
dalek ar: e9effe0 | (Steve Mynott)++ | / (3 files):
bump to 2016.07
 15:08
nine If packagers notice an issue with the Staging repo, please advise them to include commit d1604b5c8e 15:10
dalek ar: 90934e5 | (Steve Mynott)++ | tools/star/Makefile:
bump rakudo, nqp and moarvm versions to 2016.07
 15:14
[TuxCM] This is Rakudo version 2016.07-11-gd1604b5 built on MoarVM version 2016.07 15:30
test 15.184
test-t 8.047
csv-parser 16.043
8.NUL!
FWIW CSV::Parser also went down from 35 in May to 16 today 15:32
Well done!
Zoffix :o 15:33
dalek ar: b1629bf | (Steve Mynott)++ | .gitmodules:
use same lpw-simple upstream as panda with recent test fix
 15:40
ar: e93c762 | (Steve Mynott)++ | modules/perl6-lwp-simple:
correct commit sha for new upstream lwp-simple
 15:46
kudo/nom: 9d0055c | (Zoffix Znet)++ | / (3 files):
Do not lose diag()'s output at start of test file or during TODO tests

diag() is internally used for many types of messages, including TODO test outputs. It uses todo and test run counts to decide whether to place output into STDOUT or STDERR. This creates a problem when diag() is explicitly used by a module's user at the start of the test files or in the middle of a set of TODO runs.
To avoid that, we use an internal _diag() internally and export a diag() that will ask _diag() to always place user's messages into STDERR.
Fixes RT#128098 nine++ mst++ 480d8c0 | (Zoffix Znet)++ | / (3 files): Revert "Do not lose diag()'s output at start of test file or during TODO tests"
This reverts commit 9d0055cfd27460817e2fe2b86b9fa5884a93e592.		 16:11
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=128098
Zoffix github.com/rakudo/rakudo/commit/9d0055c
>_< I'm dumb 16:26
kudo/nom: 68b66c2 | (Zoffix Znet)++ | / (3 files):
Unrevert "Do not lose diag()'s output at start of test file or during TODO tests""

This unreverts commit 9d0055cfd27460817e2fe2b86b9fa5884a93e592 and fixes a bug in its code (incorrect precedence of conditional when assigning to $is_todo in _diag())		 16:55
Zoffix Well, this was embarrassingly messy. :/ 16:56
stmuk pl6anet.org/drop/rakudo-star-2016.07-RC0.tar.gz 17:01
gfldex m: (join('', "un" xx 10) ~ "revert").tc.say 17:06
camelia rakudo-moar 480d8c: OUTPUT«Ununununununununununrevert␤»
gfldex Zoffix: ^^^ Perl 6 can help you with that! :->
ugexe lizmat: gist.github.com/ugexe/7ce08d3ad4cf...7ba9713471 This points at the commit `Add "push-all" to sequential-map "normal"` 17:36
psch r: sub f { loop { print $++; last } };f;f;f;f; 18:25
camelia rakudo-jvm cd19db: OUTPUT«0123»
..rakudo-moar 68b66c: OUTPUT«0000»
psch that's #126528
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=126528
psch r: sub f { loop { state $x; print $x++; last } };f;f;f;f; 18:28
camelia rakudo-jvm cd19db: OUTPUT«0123»
..rakudo-moar 68b66c: OUTPUT«0000»
psch which looks to me like either we do state vars wrong on jvm, or the loop block doesn't get its own call frame on jvm 18:41
bartolin . o O ( or both ) # SCNR 18:44
psch yes, or that :) 18:47
Zoffix Isn't the JVM version the correct one? 18:48
psch Zoffix: well, if it is our Perl 6 level implementation of HYPER is wrong
Zoffix r: my $x; sub f { loop { print $x++; last } };f;f;f;f; 18:49
camelia rakudo-moar 68b66c, rakudo-jvm cd19db: OUTPUT«0123»
psch r: sub f { state $x; loop { state $y; say $x++, $y++ } };f;f;f;f
oh duh 18:50
didn't last >_>
camelia rakudo-moar 68b66c: OUTPUT«(timeout)00␤11␤22␤33␤44␤55␤66␤77␤88␤99␤1010␤1111␤1212␤1313␤1414␤1515␤1616␤1717␤1818␤1919␤2020␤2121␤2222␤2323␤2424␤2525␤2626␤2727␤2828␤2929␤3030␤3131␤3232␤3333␤3434␤3535␤3636␤3737␤3…»
..rakudo-jvm cd19db: OUTPUT«00␤11␤22␤33␤44␤55␤66␤77␤88␤99␤1010␤1111␤1212␤1313␤1414␤1515␤1616␤1717␤1818␤1919␤2020␤2121␤2222␤2323␤2424␤2525␤2626␤2727␤2828␤2929␤3030␤3131␤3232␤3333␤3434␤3535␤3636␤3737␤3838␤3939…»
psch now *that* is interesting
'cause moar changes its tune when there's a state in the sub block 18:51
also, i hope the jvm just didn't tell us that it timed out and isn't still running :S
Zoffix m: state $x; say print $x
camelia rakudo-moar 68b66c: OUTPUT«Use of uninitialized value $x of type Any in string context␤Any of .^name, .perl, .gist, or .say can stringify undefined things, if needed. in block <unit> at <tmp> line 1␤True␤»
Zoffix m: state $x; say print $x++
camelia rakudo-moar 68b66c: OUTPUT«0True␤»
Zoffix The '0000' is wrong. state $x keeps $x around, so those ++ should keep increasing the same variable and growing its values. So JVM's version is the correct one. 18:52
m: state $x; say $x; say $x++
camelia rakudo-moar 68b66c: OUTPUT«(Any)␤0␤»
Zoffix This kinda iffy IMO, 'cause the value ++ returns is different than the original 18:53
psch Zoffix: right, which means moar erroneously binds to the loop block instead of the sub block
Zoffix m: my $x = class {}; say $x; say $x++ 18:54
camelia rakudo-moar 68b66c: OUTPUT«(<anon|63429520>)␤0␤»
psch well, i suppose that means we can retag the ticket to [moarvm] and then rewrite &HYPER 18:55
i suppose the rewrite could limit itself to using a lexical instead of a state var in the loop in &HYPER 18:57
Zoffix .tell stmuk not sure if it affects R* release, but FYI the 2016.07 release seems to have issues with custom DESTDIR that I think are fixed in HEAD: rt.perl.org/Ticket/Display.html?id=128652 19:00
yoleaux2 Zoffix: I'll pass your message to stmuk.
Zoffix m: my $x = Str but role { method gist { 'speshul' } }; say $x++ 19:02
camelia rakudo-moar 68b66c: OUTPUT«0␤»
Zoffix m: my $x = 0 but role { method gist { 'speshul' } }; say $x++
camelia rakudo-moar 68b66c: OUTPUT«speshul␤»
Zoffix ¯\_(ツ)_/¯ probably OK.
psch r: sub f { state $x; loop { state $y; say $x++, $y++; last } };f;f;f;f 19:04
camelia rakudo-jvm cd19db: OUTPUT«00␤11␤22␤33␤»
..rakudo-moar 68b66c: OUTPUT«00␤10␤20␤30␤»
psch there, i was wrong
not lasting obviously changes the behavior
because, well, we loop contiuously so none of the state vars get reset 19:05
i really don't know if a loop block tracks its state var seperately from the next outer
m: sub f { state $x; say $x++, (state $y)++ for ^1 }; f;f;f;f; 19:06
camelia rakudo-moar 68b66c: OUTPUT«00␤11␤22␤33␤»
psch r: sub f { state $x; say $x++, (state $y)++ for ^1 }; f;f;f;f;
camelia rakudo-moar 68b66c, rakudo-jvm cd19db: OUTPUT«00␤11␤22␤33␤»
Zoffix m: sub f { state $x; for ^4 { state $y; dd [$x++, $y++]; last if $_ > 2} };f;f;f;f 19:07
camelia rakudo-moar 68b66c: OUTPUT«[0, 0]␤[1, 1]␤[2, 2]␤[3, 3]␤[4, 0]␤[5, 1]␤[6, 2]␤[7, 3]␤[8, 0]␤[9, 1]␤[10, 2]␤[11, 3]␤[12, 0]␤[13, 1]␤[14, 2]␤[15, 3]␤»
psch that is clearly correct, statement mod if doesn't block scope, so yeah
m: sub f { state $x; g($x++) }; sub g($x) { state $y; say $x, $y++ };f;f;f;f;
camelia rakudo-moar 68b66c: OUTPUT«00␤11␤22␤33␤»
psch ...are loop block that different to new subs? 19:08
Zoffix hm?
psch Zoffix: well, my last example prints the same that the loop example does on jvm
Zoffix I'm not sure what you mean by 'clearly correct'. The output in my eval is buggy. The numbers should be increasing the same. `last` seems to re-init state 19:09
psch Zoffix: i wasn't talking about your code
Zoffix Ah. Sorry
psch Zoffix: i was talking about mine, the statement level if
Zoffix m: sub f { state $x; for ^4 { state $y; dd [$x++, $y++]; next if $_ > 2} };f;f;f;f 19:10
camelia rakudo-moar 68b66c: OUTPUT«[0, 0]␤[1, 1]␤[2, 2]␤[3, 3]␤[4, 0]␤[5, 1]␤[6, 2]␤[7, 3]␤[8, 0]␤[9, 1]␤[10, 2]␤[11, 3]␤[12, 0]␤[13, 1]␤[14, 2]␤[15, 3]␤»
psch i actually kind of want to assume moar does right
mostly ad authoritam, though :P
which is annoying, 'cause it means another icky bug on nqp-j 19:11
Zoffix I don't think we need to make any assumptions. Initialization of state happens just once. That isn't the case on MoarVM. 19:14
Here's Perl 5's version too: $ perl -E 'sub f { state $x; for (1..4) { state $y; print $x++ . " " . $y++ . " | "; last if $_ > 2} };f;f;f;f'
0 0 | 1 1 | 2 2 | 3 3 | 4 4 | 5 5 | 6 6 | 7 7 | 8 8 | 9 9 | 10 10 | 11 11 |
psch right, i was mostly complaining about exactly that assumption :) 19:15
*of mine
ugexe m: my $s := (^5).rotor(3, :partial).map: -> $c { slip(@$c) }; my $buf = Buf.new($s || 0); say $buf.perl # lizmat: this is the golf 19:20
camelia rakudo-moar 68b66c: OUTPUT«Buf.new(0,3,4)␤»
ugexe its always index 1 and 2
lizmat ugexe: it's really weird 19:21
ugexe seems related to rotor somehow
lizmat I'm looking at !push-list in src/core/Buf
and it's getting a BOOTInt (no .WHAT associated with that)
which makes the nqp::istype fail, which in turn makes it croak 19:22
now, an int +& 0xFF should create an Int, so no idea *how* it becomes a native
ugexe m: my $s := (^5).rotor(3, :partial).map: -> $c { slip(@$c) }; my $buf = Buf.new($s); say $buf.perl 19:24
camelia rakudo-moar 68b66c: OUTPUT«Buf.new(0,1,2,3,4)␤»
psch lizmat: maybe inlining? 19:25
i mean, it sounds like it could be from not-guarded-enough inlining
lizmat yeah, feels like that, because it's also a Heisenbug: adding debug text to infix:<+&> fixes the problem it seems :-(
ugexe m: my $s := (^5).rotor(3, :partial).map: -> $c { slip(@$c) }; say $s.elems; my $buf = Buf.new($s || 0); say $buf.perl 19:26
camelia rakudo-moar 68b66c: OUTPUT«5␤Buf.new(0,1,2,3,4)␤»
dalek ast: 2fb0068 | usev6++ | S32-list/combinations.t:
Skip test on JVM that started to die recently
 19:27
lizmat ugexe: I just fixed the problem by adding an nqp::say to the +& candidate that is *not* being called :-( 19:32
timotimo oh lord :( 19:33
ugexe m: my $s := (^5).rotor(3).map: -> $c { slip(@$c) }; say $s.elems; my $buf = Buf.new($s || 0); say $buf.perl 19:34
camelia rakudo-moar 68b66c: OUTPUT«3␤Buf.new(0,1,2)␤»
ugexe m: my $s := (^5).rotor(3).map: -> $c { slip(@$c) }; my $buf = Buf.new($s || 0); say $buf.perl # oops
camelia rakudo-moar 68b66c: OUTPUT«Buf.new(0)␤»
lizmat ugexe: I'm focusing on:
m: my int $i = 10; say Buf.new($i +& 0xFF)
camelia rakudo-moar 68b66c: OUTPUT«Type check failed in initializing element #0 to Buf; expected uint8 but got Int (10)␤ in any at gen/moar/m-Metamodel.nqp line 1736␤ in block <unit> at <tmp> line 1␤␤Actually thrown at:␤ in any at gen/moar/m-Metamodel.nqp line 3055␤ in any…»
lizmat j: my int $i = 10; say Buf.new($i +& 0xFF) 19:36
camelia rakudo-jvm cd19db: OUTPUT«Type check failed in initializing element #0 to Buf; expected uint8 but got Int (10)␤ in any at gen/jvm/Metamodel.nqp line 1725␤ in block <unit> at <tmp> line 1␤␤Actually thrown at:␤ in any at gen/jvm/Metamodel.nqp line 3044␤ in any at ge…»
lizmat hmmm... so it is actually consistent across backends!
I'm starting to feel this is something related to the method caching algorithm 19:46
somehow confusing (Int:D \a, Int:D \b) with (int $a, int $b) 19:47
ugexe lizmat: when I first ran into this I manually iterated through commits and got the method caching commit from around the same time
lizmat if I remove the infix:<+&>( int $a, int $b ) candidate altogether, it's ok :-( 19:48
dalek ast: a1c52ba | usev6++ | S04-phasers/in-loop.t:
Add test for RT #126001
 19:51
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=126001
ugexe my int $i = 10; say Buf.new($i +& 0xFF) works with --optimize=0 or 1 19:57
timotimo that's a good tip. so maybe it's compile-time-deciding what candidate to use for +& there? 19:58
lizmat the profile for the above code is bizarre 20:00
ugexe++ 20:01
ugexe fwiw gist.github.com/ugexe/baa168a64189...812724f76d (after-after.pm6) would "always fails first time after `rm -rf lib/.precomp` with a segfault but always passes all subsequent runs" 20:02
lizmat ok, it's definitely some kind of mixup between candidates 20:17
if I add an nqp::null to the int/int candidate, all is fine
if I add a "nqp::say("foo") to the Int:D/Int:D candidate, it does *not* get executed, and I get the error 20:18
dalek ast: 47b646c | (Zoffix Znet)++ | S24-testing/line-numbers.t:
Use `like` instead of ok() with smartmatch
lizmat so it looks like it is calling the int/int candidate with incorrect parameters
timotimo i wonder if the thing you're adding is preventing the static optimizer from setting the inline info on the sub?
lizmat could well be 20:19
I mean, an nqp::null in sink context... what could it do? :-) 20:20
timotimo hm, we should be able to grab the node attribute out of the qast when we do the "add inline info" and build a little "report" about all the things that do get inline info
lizmat timotimo: how?
another way to fix the issue, is to add an .Int to the result of the nqp::bitand_i in the int/int candidate 20:21
so at least it gets the right type of return value (for this case)
timotimo well, we should be able to just print out "code object named 'foobar' on line 1234 got inline info" or "failed to make inline info for reason $foo" and parse that with a tiny script 20:22
lizmat m: say Buf.new(12354352413543265165462154654261546573625765389836927693761897681768271768726187623876876812763876128366524165465241 +& 0xFF) 20:23
camelia rakudo-moar 68b66c: OUTPUT«Buf:0x<99>␤»
lizmat m: say Buf.new(1235435241354326516546215465426154657362576538983692769376189768176827176872618762387687302909283098210983209820139803126812763876128366524165465241 +& 0xFF)
camelia rakudo-moar 68b66c: OUTPUT«Buf:0x<99>␤» 20:24
lizmat always 0x99, seems odd ?
timotimo: ok, where do I add this debug info ?
timotimo i think the method is in Actions somewhere? 20:25
literally "add_inline_info"
actually, i think it has "try" in its name
lizmat add_inlining_info_if_possible
dalek ast: 49ef2b1 | usev6++ | S11-modules/require.t:
Change back to dir that contains executable

When this testfile was run before 'make install' it blowed up on rakudo-j because executable './perl6-j' was not found in 't/packages/packages'. Probably a precompilation process was started in that dir.		 20:26
timotimo that's it
it's not straightforward to get the line number out of a node attribute, i believe
you have to go through the compiler object i think? 20:27
there's other places that should show you how it's done
lizmat timotimo: so the plan is to actually prevent it from being inlined ? 20:31
timotimo er, maybe?
lizmat if so, that feels wrong to me, as the inlining is just exposing the bug
timotimo i'm not so sure
i thought that report thing would be a debug help
lizmat we have an Int:D/Int:D and an int/int candidate
it calls the int/int candidate with the Int:D/Int:D parameters 20:32
*that* is the issue, as far as I can see
which smells like a method cache issue
to me
timotimo but don't we generate code in that case to make it still work right?
i dunno :(
i think we used to have - or maybe still do have - trouble picking the native candidates properly in the static optimizer? i'm not sure any more what that was about :(
dalek ast: 35eaee5 | (Zoffix Znet)++ | S04-statements/for.t:
[TODO FUDGE] ("{$_}") inside for loop has correct value

RT#128054		 20:35
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=128054
timotimo my suggestions may be absolutely useless :| 20:40
lizmat fwiw, RT #128655 21:13
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=128655
dalek kudo/nom: 242baf2 | lizmat++ | src/core/Int.pm:
Temporary fix for RT #128655

By removing the int/int candidates for now, we can at least ensure correct functionality.		 21:15
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=128655
ast: e0330b7 | lizmat++ | S03-operators/buf.t:
Add tests for RT #128655
 21:28
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=128655
raiph psch: Zoffix: Aiui: state vars are init'd per "closure clone". `sub ...` creates a closure clone each time the declaration is called. usually once in a program. `loop ...` creates and loops a closure clone each time it's called. Or something like that. I think moar's output is correct. 21:29
lizmat hope jnthn has some insights tomorrow
good night, #perl6-dev! 22:04
timotimo gnite liz
Zoffix raiph, and Perl 5 is wrong? And state is pretty much useless? I disagree. 22:05
In fact, your reasoning is flawed in that `state` works fine in the loop, it's the last/next that break it, not the loop. 22:06

Powered by Cro and the Raku® Programming Language.

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

Thank you!