notviki m: say now - 1 01:42
camelia rakudo-moar a96292: OUTPUT«Instant:1482198165.690788␤»
notviki m: say now-1
camelia rakudo-moar a96292: OUTPUT«===SORRY!=== Error while compiling <tmp>␤Undeclared routine:␤ now used at line 1. Did you mean 'not', 'HOW'?␤␤»
notviki m: say rand - 1
camelia rakudo-moar a96292: OUTPUT«-0.950830785771641␤»
notviki m: say rand-1
camelia rakudo-moar a96292: OUTPUT«-0.833198465193581␤»
japhb .tell ugexe I'm switching over from panda to zef in my "rebuild the world" scripts, and I'm running across a testing problem. When a test crashes with a compile time error (===SORRY!===) such as a missing include, zef's testing framework reports 'No subtests run' and then 'All tests successful.' and later 'Testing [OK] for <dist>'. An example of this is the tests for META6. 02:24
yoleaux2 japhb: I'll pass your message to ugexe.
ugexe japhb: does that happen if you run with --/tap-harness? I'm guessing its a bug with TAP::Harness6 - if that option works that would confirm 03:33
yoleaux2 02:24Z <japhb> ugexe: I'm switching over from panda to zef in my "rebuild the world" scripts, and I'm running across a testing problem. When a test crashes with a compile time error (===SORRY!===) such as a missing include, zef's testing framework reports 'No subtests run' and then 'All tests successful.' and later 'Testing [OK] for <dist>'. An example of this is the tests for META6.
japhb .tell ugexe Good catch; --/tap-harness fixes the problem (which is to say that it correctly fails the tests on a compile-time error). 04:22
yoleaux2 japhb: I'll pass your message to ugexe.
japhb .tell ugexe As a side note, the backtrace of zef itself after "Aborting due to test failure: META6:ver('0.0.11'):auth('github:jonathanstowe') (use --force to override)" just seems confusing, not terribly helpful to the end user (maybe only show it with a debug option? 04:24
yoleaux2 japhb: I'll pass your message to ugexe.
ugexe reverted zef back to using perl5 prove as the default tester 04:45
yoleaux2 04:22Z <japhb> ugexe: Good catch; --/tap-harness fixes the problem (which is to say that it correctly fails the tests on a compile-time error).
04:24Z <japhb> ugexe: As a side note, the backtrace of zef itself after "Aborting due to test failure: META6:ver('0.0.11'):auth('github:jonathanstowe') (use --force to override)" just seems confusing, not terribly helpful to the end user (maybe only show it with a debug option?
[Tux] This is Rakudo version 2016.12-37-ga96292882 built on MoarVM version 2016.12 07:17
csv-ip5xs 3.152
test 13.425
test-t 6.145
csv-parser 14.028
samcv [Tux], help 07:53
[Tux] how?
samcv oh you aren't a bot?
[Tux] nope
samcv oh ok
well not hard to see why, you always say the same thing at like intervals of whatever time :P 07:54
samcv waves
lizmat Files=1155, Tests=53962, 197 wallclock secs (11.21 usr 4.74 sys + 1187.30 cusr 125.45 csys = 1328.70 CPU) 08:45
and good morning, #perl6-dev! 08:46
DrForr waves. 08:50
lizmat news.ycombinator.com/item?id=13216924 # labster's advent post on HN 08:58
m: my @a <== 0..Inf; dd @a # this feels incorrect to me, am I right ? 09:22
camelia rakudo-moar a96292: OUTPUT«Array @a = []␤»
lizmat Zoffix: re rakudo.org/2016/12/17/lexical-module-loading/ , added it to the list of things nine did last week 09:26
$ 6 'my @a <== 0..Inf' 09:27
Cannot push a lazy list onto a Array
is more what I expected. jnthn notviki moritz ^^^ opinions ?
moritz sounds OK 09:28
timotimo lizmat: but the advent post is by tadzik :) 09:29
lizmat aaaaahhhhh
oops
why doesn't it say "tadzik" as the author then :-)
labster was the one who posted it on HN 09:30
timotimo "ttjjss" is also the name of his own blog 09:31
lizmat: but awwaid didn't post the thing to HN either ... it was labster :) 09:34
i suggest a nice cup of tea :) 09:35
lizmat eh... where did I say it was awwaiid ?
tadzik yeah, my wordpress username is different 09:41
lizmat tadzik++ # elves :-) 09:45
afk&
timotimo um, you didn'. 09:46
i suggest a nice cup of tea for myself, then
jnthn chuckles at [Tux] passing the inverse Turing test in the backlog :) 09:52
tadzik :D 09:57
notviki samcv: don't believe him! That's what all bots want you to think! 11:46
samcv !!!
notviki lizmat: I know nothing about feeds
mst FEED ME SEYMOUR
samcv the bots are taking over 11:47
lizmat m: my @a <== 0..Inf; dd @a # jnthn masak this feels incorrect 11:48
camelia rakudo-moar a96292: OUTPUT«Array @a = []␤»
dalek kudo/nom: 99e33fc | lizmat++ | src/core/Array.pm:
Another batch of fail -> throw conversions

To prevent a Failure from becoming a one element list.
This breaks one spectest:
   my @a <== 0..Inf
This used to "work" (resulting in an empty @a), it now throws:
   Cannot push a lazy list onto a Array
I think the new behaviour is correct and the test is actually faulty.		 11:52
awwaiid what hmm? 12:25
oh, I just replied 12:26
dalek kudo/nom: ece2c69 | lizmat++ | src/core/ (2 files):
Abstract xx * Seq logic into R:I:RepeatValueIterator

Part of an array_slice refactor I'm working on		 12:37
p: 0c9f07f | (Pawel Murias)++ | src/vm/js/ (3 files):
[js] Implement nqp::ctxcallerskipthunk and nqp::ctxouterskipthunk.
 13:13
p: c43f720 | (Pawel Murias)++ | t/qast/01-qast.t:
Test nqp::ctxouterskipthunk and nqp::ctxcallerskipthunk.
kudo/nom: 5d084a7 | lizmat++ | src/core/ (2 files):
R:I:RollerIterator should return an Iterator, not a Seq

Also use a more general name of source variable, as this should work on any object that has a .roll method.		 13:19
kudo/nom: af1501a | lizmat++ | src/core/ (3 files):
Remove a few is-lazy methods: the default in Mu§ is False
 13:37
notviki betlu/L 13:45
lizmat znif\O 13:46
notviki I guess now/time are keywords and rand does have a sub variant 14:00
m: say time-1
camelia rakudo-moar af1501: OUTPUT«===SORRY!=== Error while compiling <tmp>␤Undeclared routine:␤ time used at line 1␤␤»
notviki m: say time -1
camelia rakudo-moar af1501: OUTPUT«1482242448␤»
notviki m: say rand-1
camelia rakudo-moar af1501: OUTPUT«-0.671731379135938␤»
notviki The, ahem, "kok" token doesn't let keywords be followed by '-' 14:01
dalek kudo/nom: ba7be97 | (Jimmy Zhuo)++ | src/vm/moar/ops/perl6_ops.c:
Add missing MVMROOT in finddispatcher.
 14:10
kudo/nom: af6c681 | (Jimmy Zhuo)++ | src/vm/moar/ops/perl6_ops.c:
use MVMROOT correctly
kudo/nom: 8998646 | jnthn++ | src/vm/moar/ops/perl6_ops.c:
Merge pull request #956 from zhuomingliang/patch-9

Add missing MVMROOT in finddispatcher.
kudo/nom: bd4e1f4 | lizmat++ | src/core/List.pm:
Make sure List.roll always returns a Seq

and throws Failures rather than potentially masquerading them as a single element list.		 14:42
notviki still feels like exploding failures like that severely limits usefulness of Failure mechanism... :/ 14:57
m: given Failure.new { .elems.say for .Slip, .kv, .keys, .values, .pairs, .antipairs, .pick, .roll, .iterator, .classify(%), .categorize(%), .rotor(1), .reverse, .combinations, .permutations, .nodemap({;}), .duckmap({;}), .deepmap({;}), .tree }; 15:02
camelia rakudo-moar bd4e1f: OUTPUT«1␤2␤1␤1␤1␤1␤1␤1␤1␤1␤1␤1␤1␤2␤1␤1␤1␤1␤1␤»
notviki Are we gonna make all of these throw too? Feels very wrong. 15:03
lizmat well, that was my interpretation of TimToady's stance on this
notviki Yeah, mine too.
And it feels wrong :) 15:04
lizmat there's always rule 2
notviki :)
lizmat notviki: fwiw, the changes don't cause any regressions (except on suspect tests :-) 15:09
notviki m: dd Failure.new.duckmap({;}) 15:10
camelia rakudo-moar bd4e1f: OUTPUT«(Nil,)␤»
notviki Actually the above "all of these" example is not good, some of those return failures and it's just the .elems method that hides them 15:12
dalek kudo/nom: 0745ae0 | lizmat++ | src/core/ (2 files):
Rename R:I:RepeatValueIterator to UnendingvalueIterator

It describes what it does better. With "Repeat" one would expect a fixed number of repetitions		 15:15
notviki m: use MONKEY; augment class Failure { method elems { self } }; given Failure.new { .elems.say for .keys, .pairs, .antipairs, .pick(5), .roll(5), .iterator, .classify(%), .categorize(%), .rotor(1), .reverse, .combinations, .permutations, .nodemap({;}), .duckmap({;}), .deepmap({;}), } 15:18
camelia rakudo-moar bd4e1f: OUTPUT«1␤1␤1␤1␤5␤1␤1␤1␤1␤1␤2␤1␤1␤1␤1␤»
notviki Time will tell, I guess. There's always 6.d :) 15:19
lizmat hmmm.... interesting idea
I mean, why don't we let Failure.elems do the throwing ?
notviki Well, it could just propagate it, but all of the other methods above would need to be added to Failure too, since they return something other than Failure 15:20
So there's a lot of method adding involved, but we'll maintain the Failure mechanism even for listy things
m: Failure.^mro.say 15:21
camelia rakudo-moar bd4e1f: OUTPUT«((Failure) Nil (Cool) (Any) (Mu))␤»
notviki m: Failure.^methods(:all).elems.say 15:22
camelia rakudo-moar bd4e1f: OUTPUT«274␤»
notviki well, that's the upper bound. Many of them likely work good as is
or generate them automatically during compilatio? method $method-name(|) { self } 15:23
lizmat notviki: that feels getting dirtier and dirtier 15:25
notviki hehe
Well, but we made List.roll throw now, so it no longer propagates failures, but the problem is still half-fixed since Failure.new.roll(5) still hides the failure 15:26
lizmat notviki: it does ?
]ah
ah 15:27
notviki m: Failure.new.roll(5).elems.say
camelia rakudo-moar bd4e1f: OUTPUT«5␤»
lizmat ok, sorry, it does...
m: (^Inf).roll(5)
camelia rakudo-moar bd4e1f: OUTPUT«Cannot .roll from a lazy list␤ in block <unit> at <tmp> line 1␤␤»
lizmat that actually throws now
m: my $a = (^Inf).roll(5) 15:28
camelia rakudo-moar bd4e1f: OUTPUT«Cannot .roll from a lazy list␤ in block <unit> at <tmp> line 1␤␤»
lizmat star-m: my $a = (^Inf).roll(5)
camelia ( no output )
notviki I mean that if you call .roll(5) on something that you think is a list, but is actually a Failure, it doesn't throw.
lizmat yeah, gotcha
dalek p: 2302a03 | (Pawel Murias)++ | src/vm/js/nqp-runtime/ (2 files):
[js] Implement serialization and deserialization for the NFA repr.
 15:54
p: c624649 | (Pawel Murias)++ | t/serialization/01-basic.t:
Test serializing and deserializing a NFA repr.
ugexe nine: is there a 1 liner / single file way to demonstrate lexical module loading working? 16:36
nine m: { use NativeCall; } my $p = Pointer.new; 16:37
camelia rakudo-moar 0745ae: OUTPUT«===SORRY!=== Error while compiling <tmp>␤Strange text after block (missing semicolon or comma?)␤at <tmp>:1␤------> { use NativeCall; }⏏ my $p = Pointer.new;␤ expecting any of:␤ infix␤ infix stopper␤ stat…»
nine Yep, if it explodes, everything is fine.
notviki m: { use NativeCall; }; my $p = Pointer.new;
camelia rakudo-moar 0745ae: OUTPUT«===SORRY!=== Error while compiling <tmp>␤Undeclared name:␤ Pointer used at line 1␤␤»
ugexe er, i mean I want to show 2 different versions of something being used
nine oh, thanks notviki 16:38
ugexe not sure if thats easily doable without having 2 physical distributions
nine Err...we'd need something with 2 different versions installed
ugexe would it theoretically be possible if 2 different Distribution were dynamically generated and loaded? 16:39
nine Of course. That would just be a very long 1 liner :) But a cool demo nevertheless 16:50
japhb .tell ugexe Current zef seems to be doing the right thing around compile-time test failures, thank you. However, now I'm noticing that zef doesn't seem to be pulling the latest from github for github.com/jonathanstowe/META6 , which specifically includes a fix for the lexical import error I'm still seeing when I do a from-scratch rebuild and reach the `zef install` stage for that module. 17:04
yoleaux2 japhb: I'll pass your message to ugexe.
ugexe japhb: that distribution was updated but the version was not bumped, so the cached copy doesn't know its stale 17:18
yoleaux2 17:04Z <japhb> ugexe: Current zef seems to be doing the right thing around compile-time test failures, thank you. However, now I'm noticing that zef doesn't seem to be pulling the latest from github for github.com/jonathanstowe/META6 , which specifically includes a fix for the lexical import error I'm still seeing when I do a from-scratch rebuild and reach the `zef install` stage for that module.
ugexe `zef --/cached install META6` should work in that scenario 17:19
the cache is stored at ~/.zef, so it persists between zef reinstallations 17:20
you can also store a copy of the resources/config.json in ~/.zef/config.json to modify the default behavior, such as removing the cache all together. But technically this is the correct behavior given the version number has not changed 17:24
japhb ugexe: Ah, OK, gotcha. 17:30
That option name sounds funny in my head: "dash dash slash cached" 17:31
ugexe you can change the option name if you want as well heh github.com/ugexe/zef/blob/master/r...g.json#L13 17:32
japhb ugexe: :-D 17:34
lucasb lizmat: hi, you don't have a 32 bit machine, per chance, do you? :) 17:42
I commented in rt.perl.org/Public/Bug/Display.htm...xn-1440487
^^ with a snippet that segfaults on my 32 bit machine...
I hope someone can confirm it still segfaults on their machine 17:43
for 64 bit users, I guess raising the "f(839)" value to something bigger should help trigger the segfault (I forgot to mention this in the comment)
notviki lucasb: yup 17:44
"-833 -834 -835 -836 -837 -838 Segmentation fault"
lucasb notviki: thanks!
noticed the ranges? starts in 839 and segfaults in -838. funny
notviki Doesn't segfault if I turn off the optimizer 17:45
with --optimize=off 17:46
And adding MVM_SPESH_DISABLE=1 causes it to segfault after -837 17:47
lucasb hm, interesting. I expected to be something related to the optimizer... 17:50
it's those kinds of bugs that only triggers after some threshold of iteration or allocation, idk :) 17:51
dalek p: d0b1381 | jnthn++ | tools/build/MOAR_REVISION:
Get MoarVM with deep recursion fix.
 17:56
ast: 275f7a2 | jnthn++ | integration/deep-recursion-initing-native-array.t:
Test to cover RT #130294.

Gets its own test file, because it's rather fussy about actually blowing up prior to the fix. (The first version of this test that took the code directly from the RT was worthless, since adding
  `use Test` hid the crash!)		 17:57
synopsebot6 Link: rt.perl.org/rt3//Public/Bug/Displa...?id=130294
notviki jnthn: would adding a comment inside that file indicating that it should be kept alone be acceptable (as far as repoing the crash goes)? 17:58
jnthn notviki: Yeah, that could be a good idea
notviki 'cause I can see the potential for it being merged into something else one day, since we currently are considering possibility to reorganize the roast some time in the future
jnthn I don't think that'd do enough to hide it
I can check ;)
Yeah :)
lucasb: heh, what timing...looks like the RT you're looking at is the one I spent a couple of hours hunting down this afternoon :P 17:59
notviki \o/
jnthn Just giving it a spectest run before doing a version bump, just in case. 18:00
Hm, I get a failurein t/spec/S03-feeds/basic.rakudo.moar 18:12
Anyone know about that? 18:13
notviki Yup
jnthn: that's from lizmat's change: irclog.perlgeek.de/perl6-dev/2016-...i_13771524
I don't see the spectest being updated since that comment
lizmat before updating the spectest, I would like to know if we are in agreement on the test being faulty 18:14
afk again& 18:15
notviki jnthn: yeah, there was a question on that: irclog.perlgeek.de/perl6-dev/2016-...i_13771060 18:16
jnthn notviki: bah, sto the answer is "we can have a short comment without hiding the SEGV" :P
dalek ast: 6c23a08 | jnthn++ | integration/deep-recursion-initing-native-array.t:
Comment on why a distinct test file.
notviki \o/
dalek kudo/nom: 58c79e2 | jnthn++ | tools/build/NQP_REVISION:
Get latest MoarVM.

This fixes a bug that affected a handful of programs that do deep recursion, causing GC panics or SEGVs.		 18:18
kudo/nom: e8573e7 | jnthn++ | t/spectest.data:
Run deep-recursion-initing-native-array.t.
lucasb hehe, jnthn++ :) 18:33
notviki confirmed lucasb's test file no longer segfaults with the above fix 18:34
jnthn Nice :) 18:36
lizmat jnthn: could it be that using a grammar could trigger the deep recursion issue as well 18:44
?
jnthn lizmat: Potentially 18:45
It'd need a large grammar
Note that we didn't encounter this parsing Perl 6 programs. :)
lizmat true
jnthn So "yes" but "not immediately likely"
lizmat so far, I haven't been able to crash HARNESS_TYPE=6 make spectest anymore :-)
5 runs so far 18:46
ugexe the bug i'm seeing with prove6 is when a test does `use Not::Found;` it exits 0. `use-ok('Not::Found')` works as expected 19:06
lizmat 10 runs and clean 19:07
:-( 12th run crashed 19:16
so, it feels stabler than it was before
Unhandled exception in code scheduled on thread 15 19:17
No such method 'made' for invocant of type 'TAP::Runner::State'
dalek kudo/nom: b8df3a6 | jnthn++ | src/core/Backtrace.pm:
Fix detection of runtime errors in threads.

I can only imagine this must have worked at some point in the past, but it sure didn't now. More odd is that this only showed up some of them time. Anyway, fixed now, and simple programs doing things like `await start die 'foo'` will now not report errors as if they were compile-time.		 19:20
japhb jnthn: So that bug fix basically translates to "THREAD-ENTRY is a ForeignCode, not a Sub" ? 19:22
jnthn japhb: It's more like, nqp::backtrace gives back the raw code refs
Which then get wrapped in ForeignCode 19:23
japhb Ah, gotcha
Yeah, now I get it.
jnthn Which is a tad wasteful, but if we're rendering a backtrace then I figure it's all over at that point anyway.
otoh I guess we could just be logging it somewhere :)
But that doesn't need this detection either, and the .gist already came out fine. 19:24
japhb That depends if by "wasteful" you mean "takes a few milliseconds" or "takes a couple seconds"
jnthn The former :)
Which in this context means "can be safely ignored"
japhb quite 19:25
jnthn And probably far less that milliseconds too
Righty, I'm off to cook/eat/relax
Will hunt more bugs tomorrow, including trying to turn the OO::Monitors crash into a test case and see why my attempted fix wasn't sufficient. 19:26
bbl
japhb jnthn: \o/ 19:33
.ask nine With lexical import, how do you handle the case of A.pm6 uses B.pm6, but at runtime B needs to refer to an our variable in A? 19:35
yoleaux2 japhb: I'll pass your message to nine.
ugexe perl6 -e 'use Test; my $fake_test_dir = "/tmp/t".IO andthen *.mkdir; my $fake_test_file = $fake_test_dir.child("should-fail.t") andthen *.spurt(q|use Test; use Not::Found; ok 0;|); nok( run($_, $fake_test_file), "Failing test should fail: $_" ) for <prove prove6>;' # prove6 reporting passes for failing tests example 19:37
lizmat jnthn: while benchmarking, I ran into a weird thing wrt to Num.floor: while all other entries in the profile are abut the same, the variance on Num.floor is about 2.5x slower
(or faster, whichever way you look at it). 19:38
there *are* random numbers involved (100K of them)
so this could also indicate a problem with rand() ? 19:39
hnmm... now I even got a factor of 4x slower 19:41
dalek ast: bcec811 | (Zoffix Znet)++ | integration/error-reporting.t:
Test broken promise exceptions' backtraces know they're runtime

Bug find: irclog.perlgeek.de/perl6/2016-12-20#i_13774176 Rakudo fix: github.com/rakudo/rakudo/commit/b8df3a6131		 19:43
notviki ugexe: where's this prove6?
lizmat tap-harness6/bin/prove6 19:44
aka github.com/Leont/tap-harness6.git
notviki Thanks. 19:45
ugexe: well, it's here: github.com/Leont/tap-harness6/blob...prove6#L48 19:49
It just checks the failed number and doesn't consider exit code of the program itself, it seems
ugexe note that the bug itself occurs in rakudo/lib/TAP.pm6 (not the prove6 script itself)
because it happens to zef's interface as well: github.com/ugexe/zef/blob/master/l...ce/TAP.pm6 19:50
oh ha its checkign failed == 0 19:51
notviki ugexe: seems the bug is in prove6, it's checking number of failed tests only, even tho TAP::Run has exit-code attribute too 19:52
I mean .exit-status method: github.com/rakudo/rakudo/blob/nom/...P.pm6#L710
lizmat afk for some Rogue One 19:53
ugexe yeah, i just noticed i used failed == 0 in zef as well whoops 19:54
gfldex m: class { method ^name ($) { 'oi‽' } } 20:10
camelia ( no output )
notviki .ask jnthn should stuff like class { method ^name ($) { "elems" } }; be spectested or is it a Rakudo-only detail? Like overriding the metamodel methods like that?
yoleaux2 notviki: I'll pass your message to jnthn.
gfldex îs this ^^^ valid Perl 6?
nine Serialization Error: missing static code ref for closure 'BUILD' (/usr/share/perl6/sources/C57EBB9F7A3922A4DA48EE8FCF34A4DC55942942:4) 20:19
yoleaux2 19:35Z <japhb> nine: With lexical import, how do you handle the case of A.pm6 uses B.pm6, but at runtime B needs to refer to an our variable in A?
nine Doesn't sound all that nice :/
japhb: wait...what?! 20:20
japhb nine: See github.com/ab5tract/Terminal-Print...et.pm6#L46 for a concrete failing example. Terminal::Print uses Terminal::Print::Grid to implement the screen grid, and Terminal::Print::Widget uses Terminal::Print::Grid to implement its own backing grid. A uses C, B uses C, now B wants to refer to a singleton in A. 20:24
(concrete, and mildly more complex) 20:25
nine japhb: I thought Perl 6 really gave you enough tools, so you wouldn't need global variables anymore... 20:28
japhb: why isn't this grid thing passed to Terminal::Print::Widget's method? 20:30
japhb The check on that line is checking whether the target grid is the singleton's grid, in which case print to screen by default, otherwise, just composite onto the parent's grid -- it's a DWIM. So the API *does* send the target grid explicitly, but this DWIM functionality is obviously not something you would want to trigger by being explicit, that's the opposite of DWIM. :-) 20:33
nine Terminal::Print::Grid could simply have a flag for whether it's the screen grid 20:35
japhb Callers of .composite, such as animation controllers, won't necessarily know that any more easily.
nine japhb: why? Just default it to False and only set it to True in github.com/ab5tract/Terminal-Print...nt.pm6#L86 20:37
japhb nine: Because Terminal::Print can have several grids at once, and at any given time only one is set as current-grid. 20:38
And sure, I suppose T::P::Grid could add a flag indicating it is currently the screen singleton's current-grid, that seems like the wrong place to put that data, especially as it just screams thread UN-safety to me. 20:39
nine And global variables have been known to be especially thread safe?
japhb Note that the code I originally linked does a *method call* on that global.
To get the current-grid. 20:40
That method call can be made perfectly thread safe, and the global is a singleton that never changes after creation.
I just can't refer to the global singleton at all, which seems wrong.
gfldex if you provide your own sub EXPORT you can expose anything (I may not have understood what your problem is) 20:41
japhb gfldex: Since A uses C and B uses C, I don't know that B can use A without horrible confusion. Perhaps I am wrong in this ....
nine Put it into D 20:42
used by both A and C
gfldex you can do "forward declarations for modules"
japhb gfldex: You mean like: `package Terminal::Print { ... }` ? 20:43
gfldex yes
japhb Interesting, I didn't know that worked with package decls.
japhb tries that.
gfldex and if you start with trianglish dependencies you will end up in a world of pain
japhb: err sorry, I ment you can't (I think) 20:44
nine You can but not between files
gfldex :D
nine Only for packages (classes, modules) in the same file
japhb Yeah, first variation didn't work. 20:45
nine: If I put the singleton in D, and both A and B use both C and D, will A and B see the same singleton? (Meaning, will I accidentally create a doubleton?) 20:47
nine As long as it's initialized at runtime, you should be fine. 20:48
japhb Gah, just realized it's even worse: A uses B and C, B uses C.
OK, so this is bringing up a general question for me. In the new lexical imports world, 'use' forms a DAG of reachable imported declarations, with the main program at the root. OK, fine, I get that. But where's the escape hatch if I don't want a DAG, but want a DG instead? 21:11
arnsholt Circular uses have never been possible in Perl 6 21:13
I can never remember *quite* why, but it's something to do with one-pass parsing, and allowing circular uses just messes everything up, IIRC 21:14
japhb arnsholt: But in Rakudo they have.
Not circular 'use', circularity in the reachable declaration space.
arnsholt What do you mean by "reachable declaration space"? 21:15
japhb In other words, Rakudo's previous installation of packages into the global namespace allowed any package to reach any other package's "our" declarations just by using going through the global namespace. 21:16
Even if neither package used the other, just as long as both were used somewhere, at some point.
arnsholt Ah, right
That use-case
japhb I'm not saying I want that back by default -- I just want an escape hatch.
arnsholt use MONKEY-CLAIRVOYANCE perhaps =) 21:17
jnthn Putting stuff in PROCESS is probably one such escape hatch 21:18
japhb Something like, I dunno, CUR.package('Other::Package') returning a WHO
jnthn: Oh, that's a decent point.
Actually, I guess that wouldn't be CUR but rather CU or somesuch, but you get my general point. 21:19
japhb finds temporary relief with PROCESS::<> 21:25
It ain't ideal, but it means I don't have to do surgery on ab6tract's design 21:26
RabidGravy I took a quick look at that 21:27
japhb jnthn++ # Thank you for the suggestion, it worked: github.com/ab5tract/Terminal-Print...436b3268ac 21:32

Powered by Cro and the Raku® Programming Language.

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

Thank you!