dalek ast: 046e0bf | (Zoffix Znet)++ | S17-supply/basic.t:
.tap block with incorrect signature must fail

RT#126379		 00:10
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=126379
hoelzro awwaiid: I found the regression: c8c8de6 00:34
dalek ast: 7211997 | (Zoffix Znet)++ | S03-operators/arith.t:
extremely large exponents must throw numeric overflow

RT#125811		 00:35
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=125811
dalek ast: 909b4c5 | (Zoffix Znet)++ | S03-operators/arith.t:
Remove trailing whitespace
AlexDaniel Zoffix: there's a chance that tests for #125811 will have to be fudged due to problems on OS X. See this: github.com/perl6/roast/commit/2ff9...7080d779f8 00:46
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=125811
AlexDaniel Zoffix: if OS X issues are gone, then other tests have to be unfudged :)
Zoffix Damn. I can't believe there's no OS X VMs. 02:03
There's no way I'm spending $2.5K on a laptop just to test Perl 6 :P 02:04
Zoffix hopes $work will buy one as $boss wants an Apple app
geekosaur apple will only license OS X VMs on Apple machines running OS X as host. this generally means extra expense and annoyance for hosters, since Apple has nothing designed for data center use 02:08
Zoffix I'd pay a few bucks to test my code on a VM. I'm surprised no one's even trying to monetize this :/ 02:09
Stupid ass Apple
geekosaur enh. they've convinced desktop users to pay 5x as much for half the desktop machine. they don't care about data centers, and certainly not about third party software not through the app store. 02:12
Zoffix The prices are even more ridiculous in Canadian dollars :P
geekosaur arguably the correct answer is to ignore apple and tell anyone looking for OS X versions of general Unix software to run Linux in a VM, because Apple doesn't *want* there to be OS X versions of such things 02:13
MasterDuke geekosaur: i quite dislike Apple, but i'd have to say i only agree with the second two of your three statements
i.e., not the pay 5x 02:14
that definitely used to be the case, but the various Intel ultrabooks are just about the same price for just about the same specs 02:15
maybe not so true for desktops though 02:16
geekosaur I admit I was thinking more of the desktop than the notebook market, which I don't follow closely
but then. that's why I said desktop users 02:17
MasterDuke and converse for me, i would never buy an Apple desktop (so don't follow them), but could be persuaded to buy a notebook (just for the hardware, linux would get installed right away)
geekosaur (Apple has an enterprise sales department. It has trouble holding onto both customers and employees, mostly because both end up wasting their time arguing with the rest of Apple and not getting what they want/need.) 02:27
dalek ast: 4710138 | (Zoffix Znet)++ | S02-one-pass-parsing/misc.t:
Spurious backslash at end of file error

The error must not inform about a missing semicolon and should be the
  "Confused" error with a pointer to location.
RT#123887		 03:44
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=123887
ast: 1bf2a1b | (Zoffix Znet)++ | S17-promise/basic.t:
Multiple promises execute asynchronously

RT#122803		 04:00
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=122803
Zoffix Feedback needed (there's also a couple of emails on perl6-compiler): touching '/' during tests is a bad idea IMO: rt.perl.org/Ticket/Display.html?id=126976 04:07
In a related note, how can we keep the ticket-relevant conversation on tickets and not spreading it partially over the mailing list with nothing propagated to the ticket?
And thirdly, I've reply+resolved a whole bunch of tickets and those replies did not show up on perl6-compiler, however, a couple of emails did show up. Is it because resolvings do not get mailed or because we have some email issues there somewhere? 04:08
Zoffix waves good night
geekosaur rmdir would be a potentially bad thing to attempt. mkdir would only do something unexpected if the system is already corrupted beyond recovery 04:13
Zoffix Are you Brandon? 04:18
Aye
geekosaur, what is it meant to do on Windows tho?
It's actually succeeding, unless I run it in C:\. I've no idea what it's creating.
geekosaur it's not creating anything. this is a backward compatibility wart in Windows 04:19
Zoffix Ah
geekosaur anything that behaves differently when run in \ than in a subdir is touching backward compatibility code 04:20
(specifically it is mimicking behavior of FAT filesystems; in this case the original behavior was arguably a bug, but because so many programs rely on such odd behavior it is retained) 04:22
...actually iirc it was not a bug in FAT, it was a workaround to allow buggy programs to work back in DOS 2 04:24
Microsoft may never find and eradicate all the ancient cruft >.> 04:25
hoelzro awwaiid: sorry, I should've clarified - the regression is that 'ok' is not among the completions when Test.pm is loaded 04:26
dalek ast: 56b8416 | (Zoffix Znet)++ | S32-io/mkdir_rmdir.t:
Strip trailing whitespace
 04:42
ast: 423ee64 | (Zoffix Znet)++ | S32-io/mkdir_rmdir.t:
'/'.IO.mkdir must not segfault

RT#126976
  geekosaur++ for explaining Windows side of things for this test
  (irclog.perlgeek.de/perl6-dev/2016-0..._12784679)
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=126976
Zoffix I'm still unclear on fudging... gist.github.com/zoffixznet/d01742c...1dd76f8cc0 there are 4 skips and it's telling me I ran 4-fewer tests than planned... It isn't seeing the skips or something? 04:44
ĀÆ\_(惄)_/ĀÆ
Zoffix falls to bed
geekosaur oh, I misremembered... it was not compatibility for programs. It's because the root directory on FAT was synthetic; its size and behavior was defined before subdirectories existed. So any directory operations on it short-circuited out because it didn't actually exist. 05:56
This was mostly retained on NTFS because the same operations still didn't make any sense, but for different reasons (notably, you can't refer to a disk except by the low level operations used by e.g. FORMAT if you can't refer to its root directory). 05:58
[Tux] lizmat, *adding* Perl6.pm on YAPC profile? Just add <space> Perl6.pm or <comma><space> Perl6.pm to Monger group(no s here)? 06:53
and nothing about the new speed record in the blog. You're doing such a nice job: it is worth noting 06:57
lizmat [Tux] yeah, that's it 07:05
afaik :-) 07:06
[Tux] with or without the comma? 07:07
<c> Amsterdam.pm , perl6.pm </c> works ok 07:17
lizmat yeah, it's doing a LIKE in the db, I think
Files=1109, Tests=52376, 225 wallclock secs (13.51 usr 3.81 sys + 1410.38 cusr 116.85 csys = 1544.55 CPU) # on a cold machine 07:19
lowest I've seen in a *looong* time
(in wallclock and CPU) 07:20
[Tux] This is Rakudo version 2016.06-137-g234aa4f built on MoarVM version 2016.06-9-g8fc21d5 07:26
test 15.612
test-t 9.072 07:27
csv-parser 16.615
(just to top it) :) \o/
dalek kudo/nom: 9a3a571 | lizmat++ | src/core/Rakudo/Internals.pm:
Streamline ShapedArrayCommon.pairs|antipairs
 07:28
kudo/nom: 8cb0fac | lizmat++ | src/core/Match.pm:
No need to use named params for Match|SET-SELF
lizmat [Tux]: yeah!
nine m: use nqp; say nqp::atkey(CompUnit::Loader.load-source(Buf.new(q<package Qux { BEGIN EVAL q<CompUnit::DependencySpecification.new(:short-name(q{Bar}))> }; class Foo {};>.encode)).unit, q<$?PACKAGE>); 07:50
camelia rakudo-moar 234aa4: OUTPUTĀ«(Qux)ā¤Ā»
nine This ^^^ is a golfed version of rt.perl.org/Ticket/Display.html?id=128457
The output should be (GLOBAL) instead of (Qux)
This demonstrates the BEGIN time EVAL's lexicals leaking into the outer compilation unit. 07:51
Oh, this an be even shortened to:
use nqp; say nqp::atkey(CompUnit::Loader.load-source(Buf.new(q<package Qux { BEGIN EVAL q<> }; class Foo {};>.encode)).unit, q<$?PACKAGE>);
m: use nqp; say nqp::atkey(CompUnit::Loader.load-source(Buf.new(q<package Qux { BEGIN EVAL q<> }; class Foo {};>.encode)).unit, q<$?PACKAGE>);
camelia rakudo-moar 8cb0fa: OUTPUTĀ«(Qux)ā¤Ā»
dalek kudo/nom: 4188109 | niner++ | src/core/CompUnit/Loader.pm:
Relax type requirements on CompUnit::Loader's methods

We don't need a mutable Buf. A Blob will do just fine. Fixes CompUnit::Loader.load-source("1 + 1".encode)		 08:00
psch nine: that looks like indiscriminately merging all mast_frames could be behind it 08:01
nine psch: may very well be. Will have a closer look in the evening. Unless someone manages to fix this before that :) 08:02
psch has no idea how one would discriminate between the mast_frames vOv
nine m: use nqp; say nqp::atkey(CompUnit::Loader.load-source(q<package Qux { BEGIN EVAL q<> }; say $?PACKAGE>.encode).unit, q<$?PACKAGE>); 08:04
camelia rakudo-moar 8cb0fa: OUTPUTĀ«Type check failed in binding $bytes; expected Buf but got utf8 (utf8.new(112,97,99,10...)ā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
nine m: use nqp; say nqp::atkey(CompUnit::Loader.load-source(Buf.new(q<package Qux { BEGIN EVAL q<> }; say $?PACKAGE>.encode)).unit, q<$?PACKAGE>);
camelia rakudo-moar 8cb0fa: OUTPUTĀ«(GLOBAL)ā¤(Qux)ā¤Ā»
dalek p: d138ca2 | (Pawel Murias)++ | src/vm/js/ (10 files):
[js] Change the calling convention to pass the invocant as a javascript positional argument.
 09:05
kudo/nom: 1fbb8b4 | peschwa++ | src/core/Exception.pm:
Fix a few CX creation cases and don't give a null to p6box_s
 10:33
kudo/nom: 774e2de | lizmat++ | / (4 files):
Make [x]:foo:bar and {x}:foo:bar 20% faster

  - don't create Pair for call of SLICE_ONE_(LIST|HASH)
  - simplify adding of named to named hash
  - use HANDLED in stead of DELETEKEY, which doesn't need the hash passed		 10:45
lizmat bisect: 2 ** 99999999999999999999999999999999999 11:47
bisectable lizmat: on both starting points the exit code is 1 and the output is identical as well
dalek kudo/nom: 8a7db27 | lizmat++ | / (4 files):
Make [x]:foo:bar and {x}:foo:bar 20% faster

  - don't create Pair for call of SLICE_MORE_(LIST|HASH)
  - simplify adding of named to named hash
  - use HANDLED in stead of DELETEKEY, which doesn't need the hash passed		 11:49
lizmat oops, s/x/@x/ 11:50
thanks to jnthn's named param method cache work, this now makes sense 11:51
BrokenRobot lizmat: why did you bisect that? Are you on OSX? 12:06
I added a test for that last night and someone told me the bug it tests for isn't fixed on OSX
BrokenRobot doesn't have an OSX to test 12:08
nine How can this be OS dependent anyway?
BrokenRobot No idea. It's clamed on OSX it returns 1 instead of throwing. 12:09
irclog.perlgeek.de/perl6-dev/2016-0...i_12784268
nine Btw. I'm for either skipping or un-todoing those tests. It's OS X where this is broken but right now gives spec test failures everywhere else. 12:11
BrokenRobot huh?
nine (passing todos are counted as failures by make spectest)
BrokenRobot Ah, right, the power tests. 12:12
tbrowder [Coke]: ref RT: My plea to RT support worked well, and my user ID of 'tbrowder' has been restored; however, although my ticket shows me as the owner in the general listing, RT doesn't show it in my "10 highest priority tickets I own." Can you fix that or do I need another RT support request? 12:38
lizmat m: 2 ** 99999999999999999999999999999999999 12:49
camelia rakudo-moar 8a7db2: OUTPUTĀ«WARNINGS for <tmp>:ā¤Useless use of "**" in expression "2 ** 99999999999999999999999999999999999" in sink context (line 1)ā¤Numeric overflowā¤ in block <unit> at <tmp> line 1ā¤ā¤Actually thrown at:ā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
lizmat $ 6 'say 2 ** 99999999999999999999999999999999999'
1
so indeed, on OSX this is 1 12:50
BrokenRobot Thanks, I'll fudge them
lizmat fwiw, only 1 test fails on OS X
t/spec/S03-operators/arith.t (Wstat: 256 Tests: 158 Failed: 1) 12:51
Failed test: 86
BrokenRobot: ^^
BrokenRobot Yeah. That's the one I added last night.
s/them/it/; :)
dalek ast: 9e1e6f7 | (Zoffix Znet)++ | S (2 files):
Fudge huge exponent tests as 'skip'

These tests fail on OSX but pass on Linux. Fudging them as 'todo' makes spectest runs produce failures due to passing TODOs on Linux.		 13:18
nine \o/ 13:21
hoelzro awwaiid: I think the problem is ctxsave - I don't know if it sets things up correctly if it's called in Perl 6 land 13:23
awwaiid yeah -- but I don't see it working before I moved it into Perl 6 land either 13:25
What I'm hoping for is to do like my $ctx = CALLER::; or create a new one or similar right when the REPL is invoked, but it should never be nil/Mu when it hits that spot 13:26
jnthn m: gist.github.com/jnthn/36792fd1bbb7...0502a1b668 13:30
camelia rakudo-moar 8a7db2: OUTPUTĀ«1..3ā¤Constraint type check failed for parameter '@checks'ā¤ in sub pkgname-pkgver-pkgrel at <tmp> line 15ā¤ in sub pkgname-pkgver-pkgrel at <tmp> line 17ā¤ in sub pkgname-pkgver-pkgrel at <tmp> line 12ā¤ in block <unit> at <tmp> line 26ā¤ā¤# Loā€¦Ā»
lizmat commute to amsterdam.pm meeting& 13:31
dalek kudo/nom: d4ac15c | jnthn++ | src/Perl6/Metamodel/BOOTSTRAP.nqp:
Fix constrained nameds multi-dispatch regression.

We could put candidates in the cache that were not cacheable, due to having named parameters with where constraints or unpacks. This fixes it by re-ordering the checks.		 13:45
ast: 9be78ea | jnthn++ | S06-multi/positional-vs-named.t:
Test to cover RT #128522.
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=128522
jnthn Hopefully that deals with the regression. :) 13:49
awwaiid you can give m: a gist? fancy 14:45
BrokenRobot m: my $x = 42; $x does $_ for role {}; say $x.WHAT 14:51
camelia rakudo-moar d4ac15: OUTPUTĀ«WARNINGS for <tmp>:ā¤Useless use of $x in sink context (use Nil instead to suppress this warning) (line 1)ā¤Cannot mix in non-composable type <anon|82787552> into object of type Intā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
BrokenRobot .oo( use Nil?? )
m: my $x = 42; $x does role {} for ^1; say $x.WHAT; # not useless use
camelia rakudo-moar d4ac15: OUTPUTĀ«WARNINGS for <tmp>:ā¤Useless use of $x in sink context (use Nil instead to suppress this warning) (line 1)ā¤(Int+{<anon|73490736>})ā¤Ā»
BrokenRobot m: my $x = 42; for ^1 { $x does role {}; Nil } say $x.WHAT; 14:54
camelia rakudo-moar d4ac15: OUTPUTĀ«===SORRY!=== Error while compiling <tmp>ā¤Strange text after block (missing semicolon or comma?)ā¤at <tmp>:1ā¤------> $x = 42; for ^1 { $x does role {}; Nil }ā say $x.WHAT;ā¤Ā»
BrokenRobot m: my $x = 42; for ^1 { $x does role {}; Nil }; say $x.WHAT;
camelia rakudo-moar d4ac15: OUTPUTĀ«(Int+{<anon|69666800>})ā¤Ā»
BrokenRobot m: say chr 9999999999999999999999 15:24
camelia rakudo-moar d4ac15: OUTPUTĀ«Cannot unbox 74 bit wide bigint into native integerā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
BrokenRobot m: say chr 9999999999
camelia rakudo-moar d4ac15: OUTPUTĀ«Error encoding UTF-8 string: could not encode codepoint 1410065407ā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
BrokenRobot m: say chr 2**63 15:26
camelia rakudo-moar d4ac15: OUTPUTĀ«ā€ā¤Ā»
BrokenRobot m: say chr 2**63-2
camelia rakudo-moar d4ac15: OUTPUTĀ«chr codepoint cannot be negativeā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
BrokenRobot m: say chr ((2**63)-2)
camelia rakudo-moar d4ac15: OUTPUTĀ«chr codepoint cannot be negativeā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
dalek ast: 4afc0db | (Zoffix Znet)++ | S32-io/spurt.t:
For non-existent file after spurting, .e says it exists

RT#126006		 16:07
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=126006
BrokenRobot m: my %h = foo => 'bar'; { temp %h<foos> = 'meow'; }; say %h 16:15
camelia rakudo-moar d4ac15: OUTPUTĀ«{foo => bar, foos => (Any)}ā¤Ā»
BrokenRobot That's wrong, isn't? Or is the behaviuour for use of `temp` on keys undefined? 16:16
Oh, there's already fudged test for it. nm
nine bisect: use nqp; die if nqp::atkey(CompUnit::Loader.load-source(q<package Qux { BEGIN say "eval!"; BEGIN EVAL q<>; BEGIN say "eval done!" }; class Foo {};>.encode).unit, q<$?PACKAGE>).^name eq "Qux"; 17:01
bisectable nine: exit code is 1 on both starting points, bisecting by using the output
nine: (2016-02-18) github.com/rakudo/rakudo/commit/9983c2c
nine m: use nqp; die if nqp::atkey(CompUnit::Loader.load-source(q<package Qux { BEGIN say "eval!"; BEGIN EVAL q<>; BEGIN say "eval done!" }; class Foo {};>.encode).unit, q<$?PACKAGE>).^name eq "Qux";
camelia rakudo-moar d4ac15: OUTPUTĀ«eval!ā¤eval done!ā¤Diedā¤ in block <unit> at <tmp> line 1ā¤ā¤Ā»
nine bisect: use nqp; die if nqp::atkey(CompUnit::Loader.load-source(q<package Qux { BEGIN EVAL q<>; };>.encode).unit, q<$?PACKAGE>).^name eq "Qux"; 17:02
bisectable nine: exit code is 1 on both starting points, bisecting by using the output
nine: (2016-02-18) github.com/rakudo/rakudo/commit/9983c2c
nine Indeed! We're actually dealing with a different BEGIN time EVAL issue. This has nothing to do with precompilation and the example doesn't use a nested World either. 17:03
psch: ^^^ 17:04
psch ...why is bisectable pointing at that commit 17:05
BrokenRobot It bisected by output and I'm guessing message format changed 17:06
psch oh, yeah, that makes sense
well, that also dies on r-j 17:10
which is... something, at least
nine Oh, YAPC::EU CFP is still open? 18:30
gfldex BrokenRobot: please review github.com/perl6/Pod-To-HTML/commi...a0aa250ca3 18:37
i don't know how the updated module is deployed. That may need powers I don't possess. 18:38
nine BrokenRobot: I have the first possible fix :) 18:43
psch: ^^^
It is...surprising I dare say: gist.github.com/niner/dc13b9e26acb...07fc0210af 18:44
timotimo ooooooh 18:46
that's a good idea
psch ohh 18:48
nine: can you run that against r-j as well?
nine psch: I'm quite sure it's the same there. Despite its looks, this is a quite high level issue actually. 18:49
psch nine: i'm mostly just asking for --backends=moar,jvm and running your test against perl6-j too :) 18:51
nine will do :)
psch thanks
nine Gives my computer something to do while I'm preparing dinner ;)
tbrowder ref adding new nqp file: src/Perl6/DebugPod.npq; purpose is to save debugging code (at least for my use); I have added the file to all the build tools i can find that reference src/Perl6/Pod.nqp but still have a build problem: 19:05
error: make: *** No rule to make target 'src/Perl6/DebugPod.nqp', needed by 'blib/Perl6/DebugPod.moarvm'. Stop 19:06
psch tbrowder: can you gist your diff? 19:07
tbrowder any idea where to look for a fix? thanks
psch: yes, wait a bit...
timotimo tbrowder: you potentially added it a bit too much?
tbrowder timotimo: probably, you can see in a minute or three... 19:08
psch huh, right, making DebugPod.nqp shouldn't be a thing i think? 19:09
as in, DebugPod.moarvm should be a target that is depended on, but DebugPod.nqp just shows up in that target
...if i don't misrember my make knowledge
timotimo it might have a piece of code that's usually reserved for per-backend changes or something 19:10
or gen-cat
dalek kudo/nom: c8e127a | bartolin++ | tools/build/Makefile-JVM.in:
Fix Makefile-JVM.in

duplicates change to tools/build/moar_core_sources (commit ac19874266)
kudo/nom: b8bc1b1 | niner++ | tools/build/Makefile-JVM.in:
Merge pull request #809 from usev6/patch-2

Fix Makefile-JVM.in
tbrowder adding new file DebugPod.nqp www.irccloud.com/pastebin/nFAnRlIb/diff.txt 19:12
timotimo ah, there's an error here 19:14
+$(PERL6_P_JAR): src/Perl6/Pod.nqp $(PERL6_DP_JAR)
er, no.
psch yeah, i'm not seeing it either right now :|
timotimo i don't see a problem 19:15
there's a flag to make that makes it output an absurd amount of information on what it thinks
it's kind of hard to chew through it, but it could help
i don't remember the flag :|
+ $(PERL6_O_JAR) $(PERL6_P_JAR) $(PERL6_C_JAR) $(PERL6_M_JAR) $(PERL6_B_JAR) $(PERL6_P_JAR) 19:18
there's PERL6_P_JAR in there twice
instead of having DP once
psch oh 19:20
hm, no, make isn't *that* anal about tabs i think 19:21
timotimo tbrowder: ^
psch i was thinking the spaces in l:87 of the gist maybe, but yeah, probably not
BrokenRobot gfldex: that commit is a bit weird. How come <a> is by itself and isn't actually around the index term? 19:22
tbrowder that doubling may be it, what line is it on?
BrokenRobot gfldex: and personally, I'd probably do it like this. One return: return qq[<span id="index-entry-@indices.join('-')">$text\</span>];
timotimo 26 of the gist, tbrowder 19:23
but ... that's only in the JVM makefile
BrokenRobot gfldex: the wrapping of $text is more important though. Because when it's wrapped, we can do .index-entry:target { background: yellow } or something like that to highlight the refeferenced word.
timotimo and your output was from moarvm?
tbrowder I see it (off to investigate...)
gfldex BrokenRobot: good point 19:24
BrokenRobot gfldex: as for deployment. Last time moritz update it it. Others have sudo too (like timotimo )
s/update it/updated/; # can't spell today >_<
timotimo er, what were we talking about?
BrokenRobot timotimo: world domination! Are you in? :)
timotimo too lazy 19:25
psch +1
i mean like, world domination also means taking care of the world to some degree, right
that sounds awefully exhausting
BrokenRobot nine: wow.. That's a very small fix :D
tbrowder re diffs: good catch, but that doubling was in the JVM file (which I'm actually not using) 19:28
BrokenRobot [id^=index-entry]:target # I meant... 19:30
bartolin it looks like module loading is completely broken on r-j at the moment: gist.github.com/usev6/ab0d3f9c6bba...b8be9c74ca 19:44
dalek rakudo/nom: 027dd39 | niner++ | src/core/control.pm: 19:46
rakudo/nom: Fix BEGIN time EVAL screwing up MAIN_CTX
rakudo/nom:
rakudo/nom: When compiling a source file, we store the compilation unit's main context
rakudo/nom: for later merging with the outer compilation unit's globals. This storing
rakudo/nom: is done by the compiler calling the ctxsave method of an object stored in
rakudo/nom: a contextual - $*CTX_SAVE - when it finished the compilation unit. The
rakudo/nom: ctxsave method ensures that it's called only once.
rakudo/nom:
rakudo/nom: Now EVALs are their own compilation unit. When running an EVAL at BEGIN time
rakudo/nom: of a currently compiling source file, the compiler finds a $*CTX_SAVE and
rakudo/nom: uses it to store the EVAL's main context. The source file's context won't be
rakudo/nom: stored anymore after that.
rakudo/nom:
psch huh 19:47
bartolin: that looks like it could be from my BEGIN EVAL workaround 19:48
bartolin: assuming \spec is what we read from .deserialize
bartolin psch: the line of code was added/changed with ac198742 19:49
src/core/CompUnit/RepositoryRegistry.pm:125
psch: but you're way more familiar with that code :-) 19:50
psch ...i'm really not familiar with that code :)
nine That should just split a string on '#', modify the last element of the list and join it back together 19:51
psch right, but something non-stringy ends up in $parts at that spot 19:52
fwiw, i can load Test and 'ok 1' fine on 2016.06-127-g74a0799
r-j that is
m: say $*PERL.compiler
camelia rakudo-moar 027dd3: OUTPUTĀ«rakudo (2016.06.147.g.027.dd.39)ā¤Ā»
psch ...wow, why is that 20 commits old
well, that's a start for a bisect 19:53
i'm currently on ~3s lag to hack, so i'd rather not do that myself... :S
bartolin if I had to guess, I'd say it is this line (from commit ac198742) 19:55
nqp::bindpos($parts, $path, $SPEC.canonpath(nqp::atpos($parts, $path)));
psch it could be hidding in some Exception and actually trigger in 1fbb8b40 19:56
but that's kinda unlikely
nine The backtrace points at the following line but could be off by one
tbrowder ref build: is it possible that any new file needs to be mentioned in the guts of stage0
psch tbrowder: not in rakudo, no
tbrowder: there's an nqp make target to regen bootstrap files, but rakudo doesn't care about that 19:57
bartolin aha, adding an nqp::unbox_s around the $SPEC.canonpath(...) seems to help 20:00
psch /o\
i really dread all the magic un-/boxing moar does 20:01
mostly cause it's terrible to reproduce on r-j
we ha{d,ve} a similar problem with getattr
where on moar getattr can fill in for getattr_{i,n,s}
i *think* i got that mostly right by now, but we probably just don't hit the broken cases yet 20:02
bartolin updated the gist
psch: you did put in a fix for that (getattr) on rakudo-j, didn't you? 20:03
psch bartolin: yeah, i did, but as i said, i'm not sure it really covers everything
bartolin oops, you just said that
psch bartolin: but the thing is, autoboxing is somewhat similar, in that moar can elide (sp?) containers in some cases, but jvm can't 20:04
bartolin I see
psch and it's kinda hard to make people follow the (maybe justifiably stricter, maybe overly strict) rules on the almost-left-behind backend, when the we-actually-use-this backend does things in way more user-friendly manor vOv 20:05
i'm actually not sure boxing falls into the specific thing i'm thinking of, though
but yeah, i'm ok with it cause in the end it's about the language, not one (inconsistently contributing) devs pet-backend :P 20:06
bartolin hugs psch and his pet-backend 20:07
psch ...i feel like i'm seeming horribly needy the last few days :P 20:08
geekosaur bah. there's a lot of communities where you're expected to be superhumanly non-needy or at least act as such. funny thing, many of the devs in those are more than a little antisocial... 20:11
nine psch: the unbox error doesn't make any sense. It's definitely happening on the nqp::join and it's joining two Strs 20:12
oh...should have backlogged 20:13
psch nine: you mean unboxing fixing it doesn't make any sense?
nine No, I wrote that before seeing that you guys have been faster :)
psch ah, alright
geekosaur: thanks for the reassurance, it's appreciated :) 20:14
tbrowder ref build and new file: mea culpa!! sorry for the linenoise; I tend to switch chars and am always putting npq instead on nqp, dummy yosh named the real file DebugPod.npq even though the build files say nqp--problem solved...thanks all. BTW, looking closely at the output of "make -d" did the trick. All is well with build tools. PROBLEM SOLVED 20:16
psch tbrowder++
timotimo ooooh 20:18
dalek ast: 6da250f | niner++ | S29-context/eval.t:
Test for BEGIN time EVAL screwing up outer comp unit's symbols

RT #128457		 20:19
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=128457
nine aaaand closed :) 20:21
psch nine++
vendethiel ugh, the learnxiny needs a good cleanup. I wish I knew how to split it up. Or how to write a book... 20:26
bartolin yeah, nine++ 20:34
lizmat decommute&
[Tux] nine++ that looks like real good work! 20:42
bartolin a spectest on r-m only had one failing test in S17-lowlevel/lock.t so I opened a PR (github.com/rakudo/rakudo/pull/810)
Zoffix Wooo! nine++
dalek kudo/nom: bd85716 | bartolin++ | src/core/CompUnit/RepositoryRegistry.pm:
Fix module loading on rakudo-j

Trying to load a module on rakudo-j failed with
  'This type cannot unbox to a native string'
Adding an 'nqp::unbox_s' fixes this and works for rakudo-j as well.		 20:51
kudo/nom: d5abea9 | niner++ | src/core/CompUnit/RepositoryRegistry.pm:
Merge pull request #810 from usev6/patch-2

Fix module loading on rakudo-j		 20:52
vendethiel nine++ # holy BEGIN stuff. that looks hard to debug
travis-ci Rakudo build errored. Stefan Seifert 'Fix BEGIN time EVAL screwing up MAIN_CTX 20:54
travis-ci.org/rakudo/rakudo/builds/142565689 github.com/rakudo/rakudo/compare/b...7dd390ea74
psch git didn't respond it looks like
*github 20:56
dalek kudo/nom: 41b685e | niner++ | src/core/CompUnit/ (2 files):
Use the CompUnit::Handle as $*CTXSAVE instead of CompUnit::Loader

Save a contextual and keeps more code dealing with module_ctx in CompUnit::Handle.		 21:08
nine Now that I finally fully understand this $*CTX_SAVE business, this was kind of obvious :)
You learn with every bugfix
gfldex BrokenRobot: please review github.com/perl6/Pod-To-HTML/commi...478f36407c 21:30
BrokenRobot: a complete list of name="foo" and where they come from gist.github.com/gfldex/ea31c8f951c...ca46c8f7bb
BrokenRobot: actually that list is not complete, I foolishly used --parallel=2 21:33
AlexDaniel bisect: good=2015.10 2 ** 99999999999999999999999999999999999 21:38
bisectable AlexDaniel: (2015-12-03) github.com/rakudo/rakudo/commit/6779524 21:39
AlexDaniel .tell lizmat fwiw, numeric overflow changes were before 2015.12, but you can still find it with bisectable: bisect: good=2015.10 2 ** 99999999999999999999999999999999999 21:41
yoleaux2 AlexDaniel: I'll pass your message to lizmat.
gfldex BrokenRobot: complete list gist.github.com/gfldex/b27f34fb0cc...3c94a1ff98 21:56
BrokenRobot: `X<|twigils, !>` translates to `Pod::FormattingCode.new(type => "X", meta => [["twigils", " !"],], config => {}, contents => [""])` what looks like a parser weakness to me 22:28
travis-ci Rakudo build failed. Stefan Seifert 'Use the CompUnit::Handle as $*CTXSAVE instead of CompUnit::Loader 22:46
travis-ci.org/rakudo/rakudo/builds/142589416 github.com/rakudo/rakudo/compare/d...b685ee102b
Zoffix Well, this is new: t/spec/S06-other/main-usage.rakudo.moar TODO passed: 29 23:52
ok 29 - failed constraint check doesn't leak internal exception out to the user# TODO RT \#127977 23:54
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=127977
Zoffix Looks fixed. Sometime from today, 'cause I was spectesting ~10 hours ago and it wasn't in there 23:58

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!