| [Coke] | Do we have any docs on JITDUMP we can link to in the release notes? | 01:24 | |
| japhb: github.com/MoarVM/MoarVM/blob/main...b-todo.txt ?? | 01:25 | ||
| japhb | [Coke]: Woah. I'd completely forgotten that. Would have been around when I joined Google and promptly disappeared for a few years. | 01:29 | |
| The stuff in there *was* real, but I don't know how much still is. | 01:30 | ||
| I remember the first bit was that I was really annoyed how warnings-unclean the MoarVM build was. | |||
| (Because I was pretty sure that was hiding actual subtle bugs.) | 01:31 | ||
| As for the speed, I was at the time thinking that MoarVM could be sped up a fair amount if its baseline interpreter was faster, before even spesh got involved. | 01:32 | ||
| I had been researching stuff like the LuaJIT interpreter engine when jnthn asked me kindly to not make my own VM, but to improve his instead. | |||
| (LuaJIT's baseline interpreter is wicked fast, in order to give the JIT more breathing room.) | 01:33 | ||
| [Coke]: Err, does that answer your questions? | 01:34 | ||
| [Coke] | my implied request: clean that up or delete it. :) | 01:59 | |
| (doesn't have to be today. Be nice if it was cleaned up by 2026.05) | 02:00 | ||
| japhb | Just nuke it. I'm up to my neck in $day-job, and the little time I can spare is going into terminal stuff | 02:52 | |
| Geth | rakudo/main: 6baf373546 | (Nick Logan)++ | 2 files Fix "Missing infix inside R" when using RZ/RX infix form `infix:sym<Z>` and `infix:sym<X>` carried a negative lookahead `<!before <.sym> <.infixish>>` added in 2010 (commit 1120182ed) so that `Z-` style metaops would still parse. Back then `infixish` tried `<infix>` before `<infix_prefix_meta_operator>`, so without the lookahead `infix:sym<Z>` would greedily consume the `Z` and leave the ... (21 more lines) |
02:55 | |
| rakudo/main: 764a3382b8 | (Nick Logan)++ (committed using GitHub Web editor) | 2 files Merge pull request #6132 from ugexe/ugexe/issue-6128 Fix "Missing infix inside R" when using RZ/RX infix form |
|||
| rakudo/main: 9620e240e1 | (Nick Logan)++ | src/Raku/ast/term.rakumod RakuAST: drop dead compile-time-value read in package-lookup parse Fix spec/S12-enums/misc.t test 28 ("enum raku method should be round-trippable") and a broader EVAL regression. The minimized case: my enum Foo <up>; EVAL 'Foo::<up>' ... (16 more lines) |
02:56 | ||
| rakudo/main: 995fdc5817 | (Nick Logan)++ (committed using GitHub Web editor) | src/Raku/ast/term.rakumod Merge pull request #6143 from ugexe/ugexe/fix-spec-s12-enums-misc.t RakuAST: drop dead compile-time-value read in package-lookup parse |
|||
|
02:58
apogee_ntv left
|
|||
| Geth | rakudo/main: 2408d05969 | (Nick Logan)++ | src/Raku/ast/variable-declaration.rakumod RakuAST: store explicit container base type as RakuAST::Type AST Replace RakuAST::ContainerCreator's $!explicit-container-base-type / $!conflicting-base-type meta-object slots with $!explicit-container-base-type-ast / $!conflicting-base-type-ast holding the RakuAST::Type node. Meta-objects are derived on read via IMPL-EXPLICIT-CONTAINER-BASE-TYPE / IMPL-CONFLICTING-BASE-TYPE (the node's cached .meta-object). Routes internal readers through the accessors and adds IMPL-EXPLICIT-CONTAINER-BASE-TYPE-AST so callers that want the resolved lookup AST (not just the type object) can get it. |
02:58 | |
| rakudo/main: 3fc33fda68 | (Nick Logan)++ | src/Raku/ast/variable-declaration.rakumod RakuAST: emit lexical lookup for generic 'is T' attribute container initializer Fix 'Method T.new not found' when a parametric role's attribute is declared with a generic container base type, e.g. my role R[::T] { has @.a is T; ... (17 more lines) |
|||
| rakudo/main: 836f8fcbd9 | (Nick Logan)++ (committed using GitHub Web editor) | src/Raku/ast/variable-declaration.rakumod Merge pull request #6139 from ugexe/ugexe/fix-spec-s02-types-is-type.t RakuAST: fix 'is T' on generic role attributes |
|||
| rakudo/main: 2cf777b21b | (Nick Logan)++ | src/Raku/ast/variable-declaration.rakumod RakuAST: nominalize sigilless variable invocant for .= initializers A sigilless variable declared with a definite-constrained type and a '.=' initializer, e.g. `my Int:D \b .= new: 42`, died with: You cannot create an instance of this type (Int:D) ... (16 more lines) |
|||
| rakudo/main: 3983159b54 | (Nick Logan)++ (committed using GitHub Web editor) | src/Raku/ast/variable-declaration.rakumod Merge pull request #6144 from ugexe/ugexe/fix-rakuast-sigilless-definite-callassign RakuAST: nominalize sigilless variable invocant for .= initializers |
|||
| rakudo/main: c4dcb61c16 | (Nick Logan)++ | 2 files RakuAST: emit runtime construction for generic Definedness/Coercion types Fix `T:D`, `T:U`, `T()`, `T:D()` on type-capture bases inside parametric role method bodies, e.g. my role R[::T] { method t-nominalizables { T:D, T:U, T(), T:D() } ... (39 more lines) |
|||
| rakudo/main: f9a1a517af | (Nick Logan)++ | src/Raku/ast/type.rakumod RakuAST: tighten generic detection in Type::Simple via IMPL::Archetypes.is-generic Two `nqp::can($value.HOW, 'archetypes') && $value.HOW.archetypes.generic` spots in RakuAST::Type::Simple called archetypes bare. Per the gotcha documented on RakuAST::IMPL::Archetypes.is-generic (and observable in Metamodel/DefiniteHOW.nqp / Metamodel/CoercionHOW.nqp), the bare form returns the non-generic prototype archetype and silently misreports ... (7 more lines) |
|||
| rakudo/main: bdb8ee9e9f | (Nick Logan)++ (committed using GitHub Web editor) | 2 files Merge pull request #6148 from ugexe/ugexe/fix-rakuast-definite-coercion-generic RakuAST: fix generic Definedness/Coercion in parametric role bodies |
|||
|
03:00
apogee_ntv joined
|
|||
| japhb | ugexe++ # Clearly on a roll! | 03:28 | |
| Geth | nqp/main: 9d4f8b61df | (Elizabeth Mattijsen)++ | tools/templates/MOAR_REVISION Bump MoarVM to get debugserver additions, patrickb++ |
09:15 | |
| rakudo/main: a8b8c71327 | (Elizabeth Mattijsen)++ | tools/templates/NQP_REVISION Bump NQP to get debugserver additions, patrickb++ |
09:24 | ||
| [Coke] | for MoarVM, please consider updating docs/Changelog as you go | 13:10 | |
| (nqp has no changelog, and rakudo has the changelog draft wiki page) | 13:11 | ||
| releasable6: next | 14:01 | ||
| releasable6 | [Coke], Next release in ≈24 days and ≈4 hours. There are no known blockers. 20 out of 20 commits logged | ||
| [Coke] | let me know when to kick off an early blin run with all the PR merges. | 15:34 | |
| Looks like there's still some recent ones in the queue, didn't want to get ahead of it. | |||
| lizmat is still working on getting Baggies to handle coercive type parameterizations correctly | 15:36 | ||
| [Coke] | 👍 | 15:54 | |
| lizmat | m: dd <a b c>.Set (|) Map.new(:42a, :666d) | 19:00 | |
| evalable6 | Bool = Set.new("a","b","c","d") | ||
| [Coke] | wasn't 'make install' with RAKUDO_RAKUAST working a bit ago? | 23:01 | |
| Geth | rakudo/main: fa177fef93 | (Nick Logan)++ | 4 files Fix RakuAST .EVAL at BEGIN time losing access to setting symbols Calling .EVAL on a synthetic RakuAST node inside a BEGIN block failed with "Could not find a compile-time-value for lexical Nil" because the EVAL resolver couldn't reach CORE. At BEGIN time the caller's lexical context chain terminates at the partially-built CompUnit mainline and never links to the setting, so IMPL-SETTING-FROM-CONTEXT returns Nil ... (46 more lines) |
23:09 | |
| rakudo/main: 4666cd5684 | (Nick Logan)++ (committed using GitHub Web editor) | 4 files Merge pull request #6135 from ugexe/ugexe/issue-6134 Fix RakuAST .EVAL at BEGIN time losing access to setting symbols |
|||