nine This looks kinda scary: 06:36
RabidGravy eugh 06:58
dalek kudo/nom: 5435f32 | bartolin++ | src/core/
Re-add check for X::Method::NotFound for JVM

With commit 222d16b0b9 the check for X::Method::NotFound was moved to a location in src/Perl6/Metamodel/BOOTSTRAP.nqp, that was only interpreted by MoarVM.
This lead to about 30 spectest failures on JVM, since Exceptions changed to X::AdHoc.
kudo/nom: c1b9878 | niner++ | src/core/
Merge pull request #889 from usev6/patch-3

fail rather than return a type object for null deref
Also add a better Bool coercion for Pointer
[TuxCM] This is Rakudo version 2016.09-73-gfc3160c built on MoarVM version 2016.09-1-gdebb859
csv-ip5xs 4.344
test 15.923
test-t 7.111
csv-parser 17.518
kudo/nom: c01fc3a | niner++ | lib/NativeCall/Types.pm6:
Merge pull request #888 from jonathanstowe/new-deref-null

Prevent a segfault on deref of null typed pointer
RabidGravy Ooh I was just going to say that should be fine to merge :) 07:02
psch bartolin: 5435f32 is a lot of a workaround, honestly 07:29
bartolin: what we really want to do is invoke a handler (stored in P6EX) from NQP when we cannot find the method
bartolin: compare nqp commit f84cc8a for a rough idea how that should happen 07:30
bartolin: that handler basically gets invoke in all spots that currently .dieInternal(tc, "Method .+ not found for invocant of type .+") 07:31
(because invocation is a lot different between moar and jvm, probably because of indy..? vOv) 07:32
RabidGravy .seen Zoffix 07:38
oh 07:39
bartolin psch: you're right, that workaround is ugly. I can try (later today or on the weekend) to find a cleaner solution. for now the alternative I saw was to fudge all those failing tests -- and I wasn't keen on doing that :-) 08:00
psch bartolin: well, it's not really ugly per se, and it's what we had been doing before, anyway 08:01
bartolin: it's just that, in general, we probably want to throw typed Exceptions directly from the backend via a hll handler when we can
bartolin: which, honestly, in this case i'm not completely sure we always can. i'm not sure we have all we need in e.g. resolve_noa, so it's possible there has to be a more creative/involved solution 08:03
bartolin yeah, I fully agree with 'throw typed Exceptions from the backend' .oO( if only I had more knowledge (or tuits to aquire that knowledge) ... ) 08:04
psch bartolin: but in the end, the clean solution is to replace all .dieInternal(tc, "Method not found") with an Ops.invokeDirect call for the handler in the HLLConfig, which probably is mostly pretty straight forward
the nqp commit i mentioned demonstrates the basic way to do it, although it probably can't just be cargo culted straight away and needs some thought on how to make it work in all the spots we dieInternal for MethodNotFound right now 08:05
bartolin psch: thanks for those pointers! (speaking about gaining knowledge :-)
lizmat Files=1142, Tests=53171, 228 wallclock secs (12.86 usr 3.69 sys + 1394.19 cusr 130.84 csys = 1541.58 CPU) 08:14
Zoffix m: sub foo (*@what) { @what[*-1] ~= 42; dd @what }; foo <foo bar ber> 10:09
camelia rakudo-moar c01fc3: OUTPUT«["foo", "bar", "ber42"]␤»
Zoffix m: sub foo (*@what) { @what[*-1] ~= 42; dd @what }; foo 'foo', 'bar', 'ber'
camelia rakudo-moar c01fc3: OUTPUT«Cannot assign to an immutable value␤ in sub foo at <tmp> line 1␤ in block <unit> at <tmp> line 1␤␤»
Zoffix bisect: sub foo (*@what) { @what[*-1] ~= 42; dd @what }; foo 'foo', 'bar', 'ber'
ah, right no bot.
lizmat Zoffix: I would argue the first case is wrong 10:11
Zoffix I think I recall this being a topic of discussion awhile back. 10:12
Zoffix is mostly bugged by inconsistency :(
lizmat m: sub foo (*@what) { @what[*-1] ~= 42; dd @what }; foo <foo> # even more inconsistent
camelia rakudo-moar c01fc3: OUTPUT«Cannot assign to an immutable value␤ in sub foo at <tmp> line 1␤ in block <unit> at <tmp> line 1␤␤»
lizmat m: use nqp; sub a(*@a) { say nqp::iscont(@a[0]) }; a <a> 10:15
camelia rakudo-moar c01fc3: OUTPUT«0␤»
lizmat m: use nqp; sub a(*@a) { say nqp::iscont(@a[0]) }; a <a b>
camelia rakudo-moar c01fc3: OUTPUT«1␤»
lizmat so apparently the slurpy binder is not deconting its values
psch j: use nqp; sub a(*@a) { say nqp::iscont(@a[0]) }; a <a b> 10:16
camelia rakudo-jvm cd19db: OUTPUT«1␤»
psch j: use nqp; sub a(*@a) { say nqp::iscont(@a[0]) }; a <a>
camelia rakudo-jvm cd19db: OUTPUT«0␤»
psch what strikes me as odd is that <a b> shouldn't have containers in the first place
jnthn *@a is meant to give you a mutable array
*@a is raw gives you a List of the original values 10:17
psch ah, so the single-elem and flat-list cases are wrong
well, not "flat-list"... the other example that died :S
lizmat jnthn: TIL
ok, so I guess there's something wrong in one of the "from-" methods in src/core/List 10:18
jnthn Yeah. The *@a vs **@a vs +@a causes a differnet method to be called 10:19
(from-slurpy-flat, from-slurpy, and from-slurpy-onearg) 10:20
And the "is raw" causes it to be called in the List type object rather than the Array one
psch geez i'm stressing myself out a lot here, apparently o.o 10:25
like, i got an informal job offer about two weeks ago
for some digital signage company with a very impressive client base
and the guy i talked to is somewhere in software management 10:26
and now i'm supposed to write a semi-formal and short application to their info@ address
and i'm like "but imposter syndrome" and "how do i do semi-formal" /o\
but a job would be kinda cool around now, soo i guess i gotta cope somehow 10:27
thanks for listening :S
arnsholt psch: Imposter syndrome is a real thing (and I sympathize) 10:30
Also, don't underestimate the blank page effect 10:31
timotimo i've heard if you put some bullshit onto the page first, it makes it easier? "editing is easier than adding" or something?
arnsholt Yeah, editing is so much easier
psch well, i'm struggling to hit semi-formal, honestly
arnsholt Yeah, that sounds pretty hard actually 10:32
Zoffix Write formal but remove "To whom it may concern" :P
psch also condensing my c/v without feeling like i'm cherry picking the best parts is kinda hard too
i mean, sure, on one hand i want the best parts in there 10:33
but on the other i also don't want to feel like i'm lying about how spotty my c/v actually is
Zoffix Just pick what's relevant to that particular job.
gfldex they want that you work for them, let them condense it
arnsholt That's kind of what a CV is supposed to achieve though, showcasing the most impressive parts of your experience
timotimo i thought a CV is there to shame you for taking extended periods of time off from work or training? 10:34
arnsholt Alternatively =) 10:35
timotimo i think when i created my twitter account i chose to get the business-related twitter newsletter to go with it 10:37
probably for shits and giggles ... it's kind of surreal to read these sometimes 10:38
" Research shows that over half of the users who follow small to medium sized businesses notice Twitter ads, and the majority of those who notice take action as a result.*" 10:39
that's strange, because the promoted tweets i see are almost exclusively big businesses
Zoffix heh. Whose research? The guy's who's trying to make you buy adspace? :) 10:40
timotimo "*Twitter + DB5 “Small Business Customer Insights Study,” U.S. 2014 " 10:43
masak .oO( is it "imposter" or "impostor"? I'm afraid if I ask I'll come across as ignorant ) 10:45
Zoffix Both are correct. 10:46
masak that explains why both of them look slightly wrong to me :P
Zoffix Though, with the former you can also be hung on the wall :P
.oO( I'm poster... )
masak I try to avoid that whenever possible 10:47
psch well, 225 words, 1731 bytes. the guy said to keep it short, so yeah
guess i'm gonna send this now :S
timotimo fingers crossed!
(not wires crossed!)
masak .oO( I majored in bomb defusing, with a minor in Never Cut The Red Wire ) 10:49
psch oh geez, a formatting error slipped through 10:50
welp, no edit function for email :S 10:51
it's gonna be fine
masak aye
psch back to thinking about how i can get the perl6 hllconfig in 10:52
or maybe i should shunt that error handling some where else in the first place, actually
right, so what happens seems to be that we compose classes with repr<CStruct> from NQP, not from Perl 6 10:53
which means i don't have the right hllConfig to reach the handler that throws a typed exception, and instead .dieInternal 10:54
Zoffix gifts psch
Zoffix bough an early-pre-release-copy and is disappointed with it so far. 10:55
arnsholt psch: tc.gc.getHLLConfigFor("perl6") should do it, I think? 10:59
psch arnsholt: yeah, probably, but then we die when we run a different HLL 11:01
arnsholt: as in, NQP probably shouldn't rely on only running perl6 forever
arnsholt Yeah, that'd suck
psch and apparently the current HLL *is* NQP, because that's what curFrame runs under
arnsholt I think tc.curFrame.codeRef.staticInfo.compUnit.hllName(); gets the current HLL 11:02
psch which is, i think, because we compose in nqp
arnsholt Yeah, it's an NQP-level API
I guess NQP will have to licence keys in the HLL config which can contain callbacks to throw typed exceptions
If the key doesn't exist, you can dieInternal like we do today 11:03
psch that's what i'm doing
it's also what we're doing for e.g. lexical_handler_not_found_error
arnsholt Kewl
psch but the latter happens from perl6, while what i'm doing happens from compose, which is nqp
arnsholt In that case, I'll stop telling you what you already know, now =) 11:04
psch well, the thing is we'd somehow have to either compose from perl6 level
or have another handler that hands composition errors to the hll
which also sounds kind of not-quite-right, because the composing hll is always nqp anyway 11:05
arnsholt Isn't the HLL a global?
Or was it the inverse I tripped over in Snake, perhaps 11:06
I can't remember =/
psch i don't think i understand the question 11:07
the hll can change during runtime, but there's always exactly one current hll..? 11:08
arnsholt I *think* so
jnthn: I guess you might know this one better than me? 11:09
jnthn The current HLL is a property of a compilation unit 11:11
So, if you're in NQP code, at that point the current HLL is NQP code
arnsholt Ah, okay 11:14
I was remembering the opposite of what actually happened
In that case, I'm not sure what we should do 11:15
psch yeah, it seems quite difficult to sort out 11:18
arnsholt jnthn: You have any ideas how to get that right, when NQP is in between the actual HLL and the runtime code? 11:23
timotimo enter the compilee's HLL and leave it later on :P
and use that to signal what HLL is "out there"
(don't do that, that's a very bad idea)
jnthn Well, if you know which HLL config you want, you can look it up by HLL name 11:25
psch i don't think we want to like CStruct hard to perl6, do we? 11:30
timotimo to be really fair, we don't have other serious consumers of nqp at the moment, so might as well make it a bit more difficult to do this in the future and just put a comment for future users of nqp-for-other-langs in there? 11:31
psch i can't think of a reasonable solution, honestly 12:03
i'm really against tying a NC REPR that hard to perl6, even as a workaround
so i guess it stays dieInternal and X::AdHoc until someone figures it out
[Coke] bartolin: want me to back out those few fudges I did? 12:33
RabidGravy there fixed the "not getting enough header" in LWP::Simple 12:42
hackedNODE RabidGravy++ 12:43
jnthn Where are we tending to file langauge design bugs/questions these days? 12:47
moritz I still do it as perl6/specs issues
jnthn .tell TimToady I've ran into a Unicode regex syntax ambiguity issue if you've a moment to look: 12:56
yoleaux2 jnthn: I'll pass your message to TimToady.
dalek p: bfa94f7 | jnthn++ | tools/build/MOAR_REVISION:
Bump to MoarVM with Unicode 9.
kudo/nom: 2cad3d2 | jnthn++ | tools/build/NQP_REVISION:
Get MoarVM with basic Unicode 9 support.
kudo/nom: f6524e6 | (Zoffix Znet)++ | src/core/
Fix .subst-mutate not returning all matches with :x

synopsebot6 Link:
dalek ast: 391ecba | (Zoffix Znet)++ | S05-substitution/subst.t:
Test .subst-mutate with :x returns all matches

synopsebot6 Link:
hackedNODE <3 coverage tests... Finding bugs, before our uses find them :D
our users too 13:05
travis-ci NQP build failed. Jonathan Worthington 'Bump to MoarVM with Unicode 9.'
hackedNODE "You asked me to build master, but 2016.09-3-g9b39aa5 is not new enough to satisfy version 2016.09-4-g224a261" 13:06
jnthn arse, I was on the wrong branch :/ 13:07
dalek p: 2c16625 | jnthn++ | tools/build/MOAR_REVISION:
Correct bogus Moar version bump.
gfldex jnthn: you may want to consider 13:10
nine I'm so used to that by now that I get thoroughly confused and disoriented when I'm at a shell that doesn't tell me git's status
jnthn Not sure I'd have paid enough attention to it to prevent this, tbh...
Added it anyways :) 13:12
ooh, having the current directory in a different color is worth that alone :) 13:13
hackedNODE That's neat. gfldex++ 13:14
nine jnthn: do you just have the branch name, or full status? I run with GIT_PS1_SHOWDIRTYSTATE=1 GIT_PS1_SHOWUPSTREAM="auto" GIT_PS1_SHOWCOLORHINTS=1 PROMPT_COMMAND='__git_ps1 "\u@\h:\w" "> "'
jnthn Just branch name 13:15
nine The GIT_PS1_SHOWDIRTYSTATE=1 also tells you if you have local changes or patches added to the index. 13:16
gfldex m: sub f( | ( :$a, :$b ) where { $a.defined ^^ $b.defined } ) {}; f a=>42 13:49
camelia rakudo-moar f6524e: OUTPUT«Cannot call method 'defined' on a null object␤ in sub f at <tmp> line 1␤ in block <unit> at <tmp> line 1␤␤»
gfldex jnthn: ^^^ that's what I actually wanted to do
m: sub f( :$a, :$b where { $a.defined ^^ $b.defined } ) {}; f a=>42 13:50
camelia ( no output )
gfldex m: sub f( :$a, :$b where { $a.defined ^^ $b.defined } ) {}; f a=>42, b=>43
camelia rakudo-moar f6524e: OUTPUT«Constraint type check failed for parameter '$b'␤ in sub f at <tmp> line 1␤ in block <unit> at <tmp> line 1␤␤»
jnthn So, sticking it in the subsig works?
Your `sub f( |c where { c } ) {}; f 42` one is a legit bug and I'm spectesting a fix for it locally 13:51
gfldex yes but the where-clause _has_ to be at the very last argument and the error message ain't nice 13:52
jnthn That's normal with where clauses though; consider the case where you didn't have the unpack there 13:55
($x, $y where $x < $y)
gfldex we don't really have a where clause for the whole sub. That's why I wanted to have it on the anon sub sig, to cheat that feature into Perl 6. 13:56
jnthn Yeah, but since you can write it on the last arg of the sub-sig to get that meaning, then we instead leave the where to guard the unpacking of the sub sig. 14:00
travis-ci Rakudo build failed. Jonathan Worthington 'Get MoarVM with basic Unicode 9 support.' 14:03
dalek kudo/nom: fe5423a | jnthn++ | tools/build/NQP_REVISION:
Bump to NQP with fixed MOAR_REVISION.
kudo/nom: 5f91031 | jnthn++ | src/Perl6/ (2 files):
Make declaration of sigilless params sooner.

For consistency with variable parameters. Fixes `-> |c where { c }` style constructs (previously, `c` was declared too late to be used in the `where`).
ast: 4fdfa4c | jnthn++ | S15-unicode-information/uniname.t:
Test for Unicode 9 character support.
ast: 2a11bdf | jnthn++ | S12-subset/subtypes.t:
Test to cover RT #129430.
synopsebot6 Link:
[Coke] I like how I can't read that character in the unicode 9 test anywhere yet. :| 14:09
(seriously, it means we're ahead of the curve.)
gfldex we need a font designer to volunteer for Perl 6 :-> 14:10
hackedNODE Why? 14:11
I can see that char.
jnthn: So... we now have Unicode 9 support or are you still working on it? (/me contemplates a tweet to brag about it) 14:14
jnthn hackedNODE: Not fully 14:15
hackedNODE: The graphame clustering algorithm has changed
hackedNODE ok. I'll wait :)
jnthn In such a way that's not trivial to handle
hackedNODE :(
jnthn To handle emoji, of all things :/
So that'll need some more work 14:16
hackedNODE cool
jnthn You used to be able to tell if there was a grapheme break by just looking at two codepoints
So you'd consider two of them and say "yes, there's a break between them" or "no, there ain't"
And now the rules have stuff like 14:17
(E_Base | EBG) Extend* × E_Modifier 14:18
^ (RI RI)* RI × RI fwiw :) 14:19
But underneath it still says "Grapheme cluster boundaries can be easily tested by looking at immediately adjacent characters." 14:20
[Coke] jnthn: do you have any comments on the recent ask about unicode not getting normalized? a user wanted to, for example, only change one portion of a document, not normalize everything in the document. 14:31 14:32
jnthn At some point you'll be able to do it with Uni 14:34
(e.g. you'll be able to open a file saying you want .get/.lines etc to give you Uni, not Str) 14:36
We'll need to flesh out the things you can do with a Uni too
Complaining you "can't roundtrip Unicode" is a bit silly though. The input and output may not be byte equivalent, but they're Unicode equivalent. 14:38
I figure saying what normalization form you want to write out will also cover a bunch of the use cases (e.g. "I need my input to be in NFD") 14:39
dalek p: e14ebce | (Pawel Murias)++ | t/qast/01-qast.t:
[js] Test a QAST::Var with a 'param' decl and children nodes.
p: c45a1c2 | (Pawel Murias)++ | / (4 files):
Move test passed on all backends out of t/jvm and t/moar/ to t/nqp.
p: 6633624 | (Pawel Murias)++ | src/vm/js/Compiler.nqp:
[js] Treat static similiar to contvars.
p: 779b39d | (Pawel Murias)++ | t/qast/01-qast.t:
Test that statics are rebindable.
kudo/nom: 2673ca8 | jnthn++ | src/core/
Remove untested, experimental, Promise coercions.

Coercions are a bad thing to have block. ~~ SomeEnumValue would hang as a result of the Numeric one, due to the numeric coercion performed by the ACCEPTS of the enum value, for example.
travis-ci Rakudo build failed. Jonathan Worthington 'Make declaration of sigilless params sooner. 17:01
dalek ast: 70aa3cf | (Zoffix Znet)++ | S17-promise/basic.t:
Smartmatching a Promise against Enum does not hang

synopsebot6 Link:
psch what *is* wrong with 13-union.t..? 17:02
timotimo everything, basically 17:04
psch but it's osx only..? 17:05
or isn't it and just shows up there
timotimo oh, i'd ahve to ehcek older buildes
um, what happened there
hackedNODE psch: yeah, it sems to be OSX only. It's only ever happens on OSX travis (as far as I can notice) and I had it running in a loop on Linux and it never failed 17:16
psch mhm, moar NC isn't really my cup of tea though :/ 17:31
especially considering i don't have anything running osx :)
btw, i did get a reply email to call the CEO of the company 17:33
he suggested i make an appointment to visit next monday, which is a national holiday 17:34
hackedNODE \o/
psch also, it was really, *really* weirdly informal
like, it starts with "Cool.... call me.... come over."
translated, of course
although that was literally the format of the... well, it's not actually a sentence i guess 17:36
also, does anyone wanna do the moarvm bits for throwing an adhoc exception on a zero size struct? :P
'cause i don't really wanna push tests for something that moar would fail
iirc we don't even fudge in t/, so that's out too 17:53
bartolin btw, i created a PR for NQP: 17:56
psch bartolin++
dalek nqp: 9da2705 | usev6++ | src/vm/jvm/runtime/org/perl6/nqp/runtime/ 17:57
nqp: Throw exception instead of calling dieInternal
nqp: After removing this check for a SaveStackException
nqp: the special case for JVM in Rakudo's src/core/
nqp: is no longer needed.
nqp: There are no new failing tests (neither with 'make test'
nqp: for NQP nor with 'make test; make install; make spectest'
nqp: for Rakudo).
nqp: Fixes two tickets for rakudo-j:
nqp: *
kudo/nom: 2a16055 | peschwa++ | tools/build/NQP_REVISION:
Bump NQP_REVISION for SaveStackException fix, bartolin++.
bartolin thanks :-)
psch you too :)
did you get a chance to poke at the MethodNotFound thing, ooc?
nine psch: if noone beats me to it and I don't forget, I'll have a look at the zero size struct thing tomorrow. 18:00
bartolin not yet (just returned from $work), now it's bedtime for the kids .. 18:01
nine psch: though it wouldn't exactly be unfair for you to break moar after so many people broke rakudo-j ;)
psch nine: well, the fix itself isn't that hard, it's just that i don't want travis complaining about broken builds for something this minor
nine: i could probably easily PR it myself and all, but that still just puts it into a queue somewhere 18:02
eh, actually the fix itself is quite a few lines 18:03
because we're currently using info_alloc to mean "what do we have to allocate", because we're special casing the 0 attrs situation
so what could be "num_attrs * sizeof(MVMint32)" turned into "info_alloc * sizeof(MVMint32)" because we cheat when we have no attrs 18:04
i mean, on a purely functional level it's just removing the special case for 0 attrs and checking for info_alloc == 0 and then throwing adhoc 18:05
but info_alloc becomes aesthetically annoying when we don't have a special case 18:06
so, yeah, practically it's just removing the ternary in that line and adding a == 0 check after the decls 18:07
but aesthetically that'd leave a huge mess :P
although maybe semantically it doesn't, i'm not sure on that
(also, i wouldn't even *break* r-m, i'd just introduce a test i *know* it fails, which isn't fudgable. i feel that's significantly different :P) 18:08
dalek ast: 1cff953 | usev6++ | S05-substitution/subst.t:
Fudge newly added test for JVM
hackedNODE :o 19:02
j: '' 19:03
camelia rakudo-jvm cd19db: OUTPUT«WARNINGS for <tmp>:␤Useless use of constant string "" in sink context (line 1)␤»
hackedNODE That's from July 16.... :/ 19:04
bartolin yes, would be nice to get a new r-j from time to time
hackedNODE bartolin: when was the previous passing spectest run? 19:05
(just thinking if the failures are due to the 2 tests I added today, rather than some other in that subtest that I added yesterday) 19:06
bartolin hackedNODE: it was about 20 hours ago. that was before your change from today (I checked that) 19:07
on rakudo commit fc3160c7e5
dalek ast: 09eaf48 | usev6++ | S06-advanced/wrap.t:
Test for RT #69312 now passes on JVM
synopsebot6 Link:
hackedNODE Ah. Then it's one or both of these: 19:09
m: say ($='').subst-mutate: /<[abc]>/, '', :5x
camelia rakudo-moar 2a1605: OUTPUT«Nil␤»
hackedNODE j: say ($='').subst-mutate: /<[abc]>/, '', :5x
camelia rakudo-jvm cd19db: OUTPUT«Nil␤»
hackedNODE j: say ($='').match: /<[abc]>/, :5x 19:10
camelia rakudo-jvm cd19db: OUTPUT«()␤»
nine bartolin: I thought rakudo-j would be updated automatically like rakudo-m?
hackedNODE shrugs
bartolin nine: dunno what triggers an update, sorry
hackedNODE m: say ($='').subst-mutate: /<[abc]>/, :x(1..5) 19:11
camelia rakudo-moar 2a1605: OUTPUT«Cannot resolve caller subst-mutate(Str: Regex, Range); none of these signatures match:␤ (Cool:D $self is rw: $matcher, $replacement, *%named)␤ (Str:D $self is rw: $matcher, $replacement, :ii(:$samecase), :ss(:$samespace), :mm(:$samemark), *%optio…»
bartolin hackedNODE: I tried to identify the failing subtest, but failed. I'll try again a bit later ...
hackedNODE ^ that error is confusing. I'm not calling it with (Str: Regex, Range), but with (Str: Regex, :x(Range))
m: class { multi method foo ($x, $y) {} }.foo: 2, :x 19:15
camelia rakudo-moar 2a1605: OUTPUT«Cannot resolve caller foo(<anon|61906960>: Int, :x); none of these signatures match:␤ (<anon|61906960> $: $x, $y, *%_)␤ in block <unit> at <tmp> line 1␤␤»
bartolin [Coke]: would be nice if you undo the three fudges from yesterday. in case you don't have tuits today, I can do it tomorrow 19:16
hackedNODE I can do it.. 19:17
dalek ast: 69d8dae | (Zoffix Znet)++ | S14-traits/routines.t:
Revert "Fudge RT #129375 for rakudo.jvm"

This reverts commit 7a20048943c927398e2643d09c522e8bb03eb892.
ast: 962850d | (Zoffix Znet)++ | S14-roles/composition.t:
Revert "fudge RT #64766 for rakudo.jvm"

This reverts commit 7e25c941716b47d4bfc3c99e5f868f1b6f0cdaa4.
synopsebot6 Link:
synopsebot6 Link:
bartolin hackedNODE++ 19:19
dalek ast: 18e3964 | (Zoffix Znet)++ | S32-temporal/Date.t:
[coverage] Cover all nudity in
hackedNODE I guess that's misleading... not all nudity, but methods that are reported as uncovered. I've not yet investigated how accurate the reports are for conditions inside of methods 19:45
dogbert17 o/ what's the difference between the .rakudo.moar files and the .t files? 20:41
geekosaur the former are fudged to not run tests known to fail on rakudo-moar 20:45
(to reduce the noise when looking for regressions) 20:46
arnsholt The fudging being disabling tests know to crash the test script hard
geekosaur there's a perl5 script that reads annotations in the .t files to generate the implementation specific test scripts 20:47
dogbert17 so if I have a test marked 'todo' it won't be run? 20:49
nine bartolin: building new camelia-j 20:50
bartolin: local changes prevented the git pull
bartolin \o/ nine++ 20:51
dogbert17 the following file suddenly got a rakudo.moar version:
geekosaur yes, think that happened yesterday and it would be the #?rakudo todo line that triggered it 20:56
dogbert17 my idea was that when the bug in question was fixed the test would 'fail' and then be unfudged 21:00
geekosaur well, the tests are the language spec, so you don't want to encode a known broken feature as a test for its brokenness 21:01
but I think the fudged tests are still run, and an unexpected pass is noted in the summary 21:03
dogbert17 geekosaur: thanks for the explanation
geekosaur unless, as arnsholt said, it causes a hard crash in wjhich case it is skipped completely because the test harness can't deal
dogbert17 so if a bug is found it's probably better to add a test after it's been fixed 21:04
nine j: say $*PERL.compiler 21:13
camelia rakudo-jvm cd19db: OUTPUT«rakudo (2016.06.276.gcd.19.db.2)␤»
nine God damn it: java.lang.RuntimeException: Missing or wrong version of dependency 'gen/jvm/stage2/QRegex.nqp' 21:14
Let's see if git clean -xfd and removing the inst directory helps 21:20
j: say $*PERL.compiler 21:35
camelia rakudo-jvm 2a1605: OUTPUT«rakudo (2016.09.83.g.2.a.16055)␤»
nine Finally :)
jnthn 'night, #perl6-dev 21:53
MasterDuke are filesystem operations supposed to be concurrency safe? e.g., if i have a bunch of start blocks making changes to the filesystem, but don't ever write to the same file, should i be guaranteed that they will all complete successfully (assuming no outside interference of course)? 22:03
gfldex MasterDuke: given they all operate with different file names/paths, they should all be successfull. Do you got specific code that bugs you? 22:11
dalek kudo/nom: c4055e4 | usev6++ | src/core/
Remove special case for JVM

After NQP commit 9da2705b1b this workaround for JVM is not necessary anymore. Fixes RT #122732
kudo/nom: ed0ced2 | lizmat++ | src/core/
Merge pull request #890 from usev6/rt122732

Remove special case for JVM
synopsebot6 Link:
gfldex what class/role is CALL-ME actually coming from? (that's another ENODOC) 23:29
geekosaur Callable 23:30
gfldex in Rakudo it's not defined in that role
or declared for that matter 23:31
i will doc it there anyway tho
geekosaur hm, InvocationProtocol.nqp 23:32
and I thought that method was the point of Callable, but. sigh.
... ok that's weird. doesn't document any way to specify *what* to call. (which presumably would be circumfix:<<( )>> canonically... unless it's CALL-ME, which you say it isn't.) 23:34
gfldex i'm not saying it shouldn't be in Callable. Likely leaky internals (or not leaky enough). There are plenty of roastlings tho, hence the ENODOC. 23:37
$ENODOC-- :) 23:40
geekosaur yes, this needs to be clarified. I'm not the one to do it, since how it seemed to work is apparently not how it actualy works :/ 23:50
actually I think it really is supposed to have CALL-ME but that is not enforced in the role, it's just assumed by things relying on it 23:52
since otherwise it does nothing whatsoever to say how to call a Callable