00:31 lizmat_ joined 00:33 lizmat left 01:55 librasteve_ left 09:15 lizmat joined, Geth left 09:16 Geth joined
Geth rakudo/main: c96286acb7 | (Nick Logan)++ (committed using GitHub Web editor) | 2 files
RakuAST: let a lexical shadow the now and time terms

  `now` and `time` are built-in listop terms, but a lexical of the same
name should shadow them, so `my \time = calendar-from-posix ...; time.year` reads the binding rather than calling the built-in `&time`. The grammar matched the built-in term tokens unconditionally, so the bareword always became the built-in and a method call on it hit an Int. ... (6 more lines)
10:16
rakudo/main: b1c2d21920 | (Nick Logan)++ (committed using GitHub Web editor) | 2 files
RakuAST: preserve enclosing declarations across a heredoc splice

The `cheat-heredoc` token leaves the current lexical scope so the heredoc body can splice in during the following `<.ws>`, then brings it back. Previously it did so through `enter-scope`, which builds a fresh scope holding none of the declarations registered into the original. Any phaser body run at parse time past that point could then resolve no enclosing ... (6 more lines)
10:18
rakudo/main: 546962231b | (Nick Logan)++ (committed using GitHub Web editor) | 4 files
RakuAST: make every variable declaration a declarator doc target

Previously only attributes were made the declarand when a variable is declared, so a trailing #= doc after a my, state, our or HAS declaration had nothing to attach to. The doc was left orphaned and the next declarand reported it with "Missing declarand for trailing declarator doc", a warning the legacy frontend does not emit. S26 specifies ... (24 more lines)
10:20
rakudo/main: 12123d5763 | (Nick Logan)++ (committed using GitHub Web editor) | 4 files
RakuAST: box native values produced by begin-time evaluation

Previously a begin-time evaluated expression of a native type, such as the initializer in HarfBuzz's
   constant HB_SET_VALUE_INVALID = (my uint32 $ = -1);
... (21 more lines)
10:22
rakudo/main: fc8fbb8f80 | (Nick Logan)++ (committed using GitHub Web editor) | 2 files
RakuAST: produce a once's state declaration when reached first

A `once` in a sub in a role body reaches code generation before the enclosing scope has produced the once's implicit state declaration, so its
  `$!state-decl` was still a bare type object and reading `.sentinel-value`
died "Cannot look up attributes in a RakuAST::VarDeclaration::Implicit::State type object". A `once` in a sub in a class or module, or in a method in a ... (5 more lines)
10:25
rakudo/main: 14b87d5f73 | (Nick Logan)++ (committed using GitHub Web editor) | 2 files
RakuAST: let the NFA see into a <before ...> regex argument

Previously an LTM alternation branch that begins with a <before ...> assertion had its declarative prefix cut off at the zero-width assertion, so it could lose to a shorter branch it should beat. The legacy frontend annotates the regex-argument thunk with its regex QAST as orig_qast, which QRegex::NFA uses to inline the lookahead's content ... (9 more lines)
10:26
rakudo/main: 992c4e24c5 | (Nick Logan)++ (committed using GitHub Web editor) | 2 files
RakuAST: give each signature-less block its own signature meta-object

Previously every signature-less block shared a single constant Signature meta-object for its implicit $_ parameter. That shared Signature never had its $!code attribute bound, so Signature.ACCEPTS and .cando on such a block died with "p6invokeunder first argument has to be a concrete MVMCode" when Binder.is_bindable tried to invoke under the code the signature belongs to. ... (12 more lines)
10:29
10:52 lizmat_ joined
lizmat_ RakuIRCLogger 10:52
10:55 lizmat left, lizmat_ left, lizmat joined 11:21 guifa_ joined, finanalyst left 14:02 nine left, nine joined 14:39 japhb left 14:42 japhb joined 16:26 vrurg_ joined 16:29 vrurg left
Geth rakudo/main: f59a02a585 | (Nick Logan)++ (committed using GitHub Web editor) | 2 files
RakuAST: give a no-initializer list declaration its container list as value

A `my (...)` list declaration with no initializer compiled its value to a bare null. In most rvalue positions that was harmless, but binding it on the right of a signature bind, as in `my (\a, \b) := my ($x, $y)`, calls .Capture on that null and died at compile time with "No such method 'Capture' for invocant of type 'VMNull'". Legacy yields the list of the declared containers ... (10 more lines)
17:04
18:47 finanalyst joined 20:03 librasteve_ joined 20:34 vrurg_ left 20:51 vrurg joined 22:05 finanalyst left
Geth rakudo: ugexe++ created pull request #6411:
RakuAST: adopt an imported stub package's WHO when declaring the package
22:41
22:42 librasteve_ left
japhb Starting to come back to life after multi-week sick, and first step was to get up to date with a full-stack Raku build. I haven't checked all the backlogs, is HEAD failing to install JSON::Class:ver<0.0.6>:auth<zef:vrurg>:api<1.0.5> known? 22:45
Looks like the test failure is in t/060-config.rakutest, Subtest: Global defaults, line 57, 'setting a global config parameter affects serialization'. 22:47
"===SORRY!=== Error while compiling /tmp/test-async-code-112001-000003\nMerging GLOBAL symbols failed: duplicate definition of symbol EXPERIMENTAL-MACROS\nat /tmp/test-async-code-112001-000003:1\n" 22:48
23:00 sjn left 23:01 sjn joined
ugexe first i heard of it, i'll take a look 23:14
japhb Thanks ugexe ! 23:31
ugexe seems like its failing on 2026.06 as well 23:33
japhb Bah. 23:37
ugexe oh 23:42
it is probably github.com/rakudo/rakudo/commit/e93a9dfac
so i suppose the argument could be made JSON::Class is at fault here since it is using use v6.e.PREVIEW 23:43
japhb I wonder if it is using other v6.e-isms (and thus trapped because there would be no way to specify a release that covers everything it needs) 23:48