Geth rakudo: ugexe++ created pull request #6458:
RakuAST: port six legacy optimizer rewrites to the optimize pass
02:23
ugexe i think the last two optimizations are lexical to local lowering + escape analysis, and compile time multi dispatch selection + routine inlining. they're also probably the most complicated ones 02:29
the rakuast copy returns optimization actually fixes a bug that exists in legacy 02:31
04:13 disbot left 06:34 hurufu joined 09:40 finanalyst joined 10:34 finanalyst left 11:23 apogee_ntv joined 12:19 sp1983 joined 12:23 sp1983 left
Geth rakudo/lizmat-24: b649afd8fb | (Elizabeth Mattijsen)++ | 2 files
Better fix for directives issue in Format objects

Commit 2e49395508cc592b9 added an additional cache for directives. This in itself caused potential race condition issues, because the two caches can not be updated simultaneously with a lock.
This commit changes this back to a single cache that *can* be ... (6 more lines)
12:30
rakudo: lizmat++ created pull request #6460:
Better fix for directives issue in Format objects
12:31
[Coke] I need to do a Blin run this weekend for the upcoming release. will do a second one for ast specifically. 12:34
please let me know if we have any PRs that we're eyeing for merging before release in a week. 12:35
12:37 sp1983 joined
Geth rakudo/main: dea6a8c7db | (Nick Logan)++ (committed using GitHub Web editor) | src/Perl6/ModuleLoader.nqp
Remove dead code from Perl6::ModuleLoader

resolve_repossession_conflicts has been dead since 7262b98d3e moved repossession-conflict handling to CompUnit::RepositoryRegistry. Both frontends and the legacy World wire that class's resolver; nothing calls this leftover copy. It also carried a stray-semicolon bug and an
  "XXX typecheck??". Drop it.
register_absolute_path_func has no callers, so $absolute_path_func is never set and absolute_path is already an identity function. Drop the unused hook and reduce absolute_path to a passthrough.
12:38
12:38 sp1983 left
rakudo/main: 4b66775281 | (Nick Logan)++ | 3 files
Compare revision-gated candidate signatures structurally

Previously the dispatch filter for revision gated protos grouped candidates by their rendered signature, so superseding a candidate only worked when the replacement copied its parameter names exactly. With different names both candidates were admitted, and every call from a revision that admits both died with an ambiguous dispatch ... (13 more lines)
12:39
12:39 sp1983 joined
rakudo/main: 7d0ce4d4df | (Nick Logan)++ | 2 files
RakuAST: check duplicate multi candidates within a revision gate

Previously a revision gated proto turned off the equivalent-signature redeclaration worry entirely, because equivalent candidates gated at different revisions are how a routine evolves per language version. That also silenced the worry for candidates that clash within the same revision, which can never both be dispatched to.
The check now compares only candidates competing at the same effective revision, with an ungated candidate defaulting to the proto's revision as it does at dispatch time.
rakudo/main: 8a09e50e5f | (Elizabeth Mattijsen)++ (committed using GitHub Web editor) | 4 files
Merge pull request #6448 from ugexe/ugexe/revision-gated-signature-equivalence

Compare revision-gated candidate signatures structurally
rakudo/main: 55bd399a03 | (Nick Logan)++ (committed using GitHub Web editor) | 4 files
NativeCall: enforce definiteness constraints on native signatures

Previously a definiteness constraint on a native sub was broken in two ways. A constrained type could not even be mapped: `--> Foo:D` (or
  `--> ::?CLASS:D` under RakuAST, which keeps the `:D` where the legacy
frontend resolves it to the bare `Foo`) died during setup, first with
  "Unknown type Foo:D used in native call" from type_code_for, and past
... (37 more lines)
lizmat [Coke]: that's about what *I* was confident about merging (apart from my latest PR that I will merge as soon as it turns green in CI) 12:42
12:44 sp1983 left
[Coke] v. cool. I have stuff tomorrow morning, won't get to blin until at least early afternoon Eastern US time. 12:49
releasable6: next 12:58
releasable6 [Coke], Next release in ≈7 days and ≈6 hours. There are no known blockers. 132 out of 208 commits logged
[Coke], Details: gist.github.com/4592f5064036f9dd5d...322d95b0b7
[Coke] ... Why does it give me one number in another channel and a different one here??
currently at 199/208 12:59
afk
13:10 sp1983 joined 13:28 sp1983 left 13:46 sp1983 joined 13:50 sp1983 left 13:56 sp1983 joined, sp1983 left, sp1983 joined 14:01 sp1983 left 14:03 sp1983 joined 14:07 sp1983 left
[Coke] releasable6: next 14:08
releasable6 [Coke], Next release in ≈7 days and ≈4 hours. There are no known blockers. 199 out of 208 commits logged
[Coke], Details: gist.github.com/b4e2581e8577f1695f...2ee714b3e5
ugexe i'll be merging the current optimization pr later today 14:25
[Coke] ugexe lizmat++ 15:44
ugexe++ # oops
Anyone remember pickling from Modula-3? (I'm so old). Do we have a preferred way to save data between runs? My default go to these days is JSON if it works, but we don't have anything more compact/raku-native/efficient, right? 15:45
(add portable to that list) 15:59
Geth rakudo/main: 7 commits pushed by (Nick Logan)++ 16:26
japhb [Coke]: CBOR (via CBOR::Simple) 16:59
coleman +1 cbor cbor.io/ 17:22
Geth rakudo/main: 8b21bfd105 | (Elizabeth Mattijsen)++ (committed using GitHub Web editor) | 2 files
Better fix for directives issue in Format objects

Commit 2e49395508cc592b9 added an additional cache for directives. This in itself caused potential race condition issues, because the two caches can not be updated simultaneously with a lock.
This commit changes this back to a single cache that *can* be ... (6 more lines)
17:26
[Coke] japhb++ ah, of course. Danke. 19:03
japhb :-) 19:06
[Coke] lizmat, ugexe - is that all the commits for now? If so, I'll kick off normal Blin right now. 19:07
ugexe yeah 19:08
[Coke] c: 8b21bfd105 e.say 19:18
committable6 [Coke], ¦8b21bfd: «2.718281828459045␤»
[Coke] started. 19:23
21:14 apogee_ntv left, apogee_ntv joined
Geth rakudo: ugexe++ created pull request #6461:
RakuAST: lower non-escaping lexicals to locals and flatten frame-independent bodies
21:58
22:24 hurufu left