01:23 [Coke] left
Geth rakudo: ugexe++ created pull request #6167:
RakuAST: fix subset-typed parameter binding under v6.e
02:23
03:20 [Coke] joined 03:30 guifa joined 03:56 guifa left 04:06 guifa joined 04:11 guifa left
gfldex m: sub higher-order(&c:(Int:D), *@a){ @a».&c }; sub lower-order(Str(Int:D) $s){ $s.Int }; say higher-order(&lower-order, <1 2 3>); 04:44
evalable6 (exit code 1) Signature constraint check failed in binding to parameter '&c'; expected :(Int:D $) but got :(Str(Int:D) $s)
in sub higher-order at /tmp/EudlE9jCxC line 1
in block <unit> at /tmp/EudlE9jCxC line 1
gfldex I'm not sure how I feel about that.
m: say Int ~~ Any; say Int:D ~~ Str(Int:D); 04:55
evalable6 True
False
gfldex This feel wrong. In both cases the "narrow" type should match the "wider" type. 04:56
Shall I rakudobug?
Geth rakudo/main: d4d146a978 | (Nick Logan)++ (committed using GitHub Web editor) | src/Raku/ast/statementprefixes.rakumod
Revert "RakuAST: make sure non-blocks are sunk in start" (#6161)

This breaks two spectests with RakuAST. Resolves #6142
This reverts commit d76960baa4641f65a34f057ec9870670bb3c3d04.
07:04
[Coke] lizmat: gist.github.com/coke/72c5436fbaa23...f954bae46a 13:15
there were some gaps in the blin run. as always, take those bisects with a grain of salt.
I did see a bunch of "this folder is locked" which might have impacted the bisects. 13:20
Geth rakudo/main: 92e003e28a | (Nick Logan)++ | src/Raku/ast/type.rakumod
RakuAST: fix subset-typed parameter binding under v6.e

PRODUCE-META-OBJECT on RakuAST::Type::Simple was wrapping the resolved type in IMPL-MAYBE-NOMINALIZE, added in bb4e88a28 ("RakuAST/v6.e - Fix subset type defaults in containers"). Under v6.e that helper calls HOW.nominalize on the type, and for a subset HOW.nominalize returns the refinee. So when a subset was used as a parameter type, the ... (24 more lines)
13:56
rakudo/main: 0106c6b80f | (Nick Logan)++ (committed using GitHub Web editor) | src/Raku/ast/type.rakumod
Merge pull request #6167 from ugexe/ugexe/fix-rakuast-subset-param-binding

RakuAST: fix subset-typed parameter binding under v6.e
lizmat gfldex: in the case of coercive types, the final type is leading 15:32
m: say Int:D ~~ Str(Int:D); # this is correct because:
evalable6 False
lizmat m: say Int:D ~~ Str 15:33
evalable6 False
lizmat [Coke]: looks like the failures are all my work :-) 15:38
[Coke] yay? 15:46
hopefully caught early enough to fix intead of revert 15:47
lizmat probably
AccountableBagHash fixed 15:57
Geth rakudo/main: f5773837bd | (Elizabeth Mattijsen)++ | src/core.c/Rakudo/QuantHash.rakumod
Fix two cases of ungarded nqp::atkeys

Spotted by failures in Phsyics::Measure, made visible by an early blin run Coke++
17:08
lizmat *unguarded grrr
[Coke]: the Physics::Constants failure was actually caused by Physics::Measure failing 17:13
that has now been fixed apart from 1 test that *may* have to do with 6.e sprintf work 17:14
librasteve ^^
the Clifford failure appe :-(ears to be caused by the GQR tickling a dispatch issu 17:16
cannot reproduce errors on CSS:TagSet: had problem installing CSS::Font::Resources as a dependency, but installed fine by itself 17:27
and then CSS::TagSet also tests clean
meh: the Astro::Utils problem is caused by the new .assuming 17:31
Serialization Error: missing static code ref for closure 'assumed.great-circle-waypoint'
ugexe ^^
looks like a module precompiled with an .assuming cannot be used by another module being precompiled :-( 17:37
Have it golfed, will make an issue 17:40
ugexe github.com/rakudo/rakudo/issues/5819 17:42
it is preexisting
lizmat hmm.. I thought that was fixed recently with one of your commits? 17:43
ah... 17:44
yeah, *that* case is fixed
the error occurs now only if that module is being used by another module
github.com/rakudo/rakudo/issues/6169 17:51
[Coke] I can do a limited run with just those modules when you tell me, then I'll redo a full one close to release. 17:53
lizmat If you could run a blin on just Physics::Measure to verify whether the remaining issues is indeed caused by the new 6.e sprintf logic, that would be brill 17:55
lizmat goes afk for a few hours
[Coke] running physics::measure test now. (most of time is spent on setup. :) 18:16
Physics::Measure still an always fail 18:46
not ok 53 - infix:<in>\n[Physics::Measure] # Failed test 'infix:<in>'\n[Physics::Measure] # at t/05-cvt.rakutest line 195\n[Physics::Measure] # expected: '3810mm ±114.3'\n[Physics::Measure] # got: '3810mm ±0.1143e3' 18:48
Didn't @librasteve just change something about rendering to string with an exponent?
gfldex m: say Int:D ~~ Int:D; 20:25
evalable6 False
20:26 vrurg joined
gfldex m: say Int ~~ Str(Int); say Int ~~ Int(Str); 20:27
evalable6 True
True
gfldex lizmat: the problem is not with coercion, it's the type-smiley
20:29 vrurg_ left
gfldex m: sub higher-order(&c:(Int), *@a){ @a».&c }; sub lower-order(Int(Str) $s){ $s.Int }; say higher-order(&lower-order, <1 2 3>); 20:29
evalable6 [1 2 3]
20:30 vrurg_ joined 20:33 vrurg left
gfldex That will do to conclude fixing the coersion protocol ENODOC. 20:33
20:42 ShimmerFairy left 20:43 ShimmerFairy joined
disbot2 <librasteve> please drop Physics::Measure from the tests for now - I am doing a big refactor atm - and generally sprintf should be cosmetic (ie output formatting only) so I can just patch that later 20:50
Geth rakudo: ugexe++ created pull request #6170:
Fix .assuming causing serialization error during precomp
23:32