| Geth | rakudo/main: b8b8874c7a | (Nick Logan)++ | 2 files RakuAST: apply the default parameter type to post-constraint sub-signatures A destructuring sub-signature binds like the enclosing routine's parameter list, so an unfilled parameter should bind to Any, the same default an ordinary routine parameter gets. The grammar arranges this by passing $*ON-ROUTINE into the nested signature, which lets the signature action install the Any default. ... (10 more lines) |
00:03 | |
| rakudo/main: 08237dcff5 | (Nick Logan)++ (committed using GitHub Web editor) | 2 files Merge pull request #6267 from ugexe/ugexe/rakuast-subsignature-default-type RakuAST: apply the default parameter type to post-constraint sub signatures |
|||
| rakudo/main: 2c52423ad9 | (Nick Logan)++ | 2 files RakuAST: accept lexical regexes in the grammar sanity check A capturing <foo> assertion resolves to a lexically visible regex when there is one, and compiles to a call to it rather than to a method of the grammar; Assertion::Named registers and resolves the &foo lookup for exactly this purpose. The grammar sanity check only consulted the composed methods, so a grammar using such an assertion ... (8 more lines) |
|||
| rakudo/main: 978e54b1b6 | (Nick Logan)++ (committed using GitHub Web editor) | 2 files Merge pull request #6263 from ugexe/ugexe/rakuast-grammar-lexical-assertions RakuAST: accept lexical regexes in the grammar sanity check |
|||
| rakudo/main: bf586a663e | (Nick Logan)++ | 2 files RakuAST: emit the lexical self when a self term has no resolution A role body is compiled at compose time, ahead of the compilation unit's check pass. A `self` term inside a regex in a role is reached through that early compilation, so it arrives at code generation without a resolution recorded by the check pass and dies with "This element has not been resolved". ... (6 more lines) |
|||
| rakudo/main: d80e464e3a | (Nick Logan)++ (committed using GitHub Web editor) | 2 files Merge pull request #6268 from ugexe/ugexe/rakuast-self-in-regex RakuAST: emit the lexical self when a self term has no resolution |
00:05 | ||
| rakudo/main: 76927f5833 | (Nick Logan)++ | 3 files RakuAST: align operator sink-warning purity with the legacy frontend The "useless use in sink context" warning asked the operator node's own `is-pure`, which answers from a table that treats every operator outside a small set as pure. So `temp %h;` as a statement was flagged useless even though it localizes its target and restores it when the scope leaves, and any user-defined prefix or infix without side-effect-free ... (11 more lines) |
|||
| rakudo/main: f7a1853247 | (Nick Logan)++ (committed using GitHub Web editor) | 3 files Merge pull request #6269 from ugexe/ugexe/rakuast-operator-sink-purity RakuAST: align operator sink-warning purity with the legacy frontend |
|||
| rakudo/main: 7a0cdb44ee | (Nick Logan)++ | 3 files RakuAST: don't drop a declaration when collapsing a constant conditional Previously the ternary (`?? !!`) and short-circuit (`&&`/`||` and their loose `and`/`or` forms) collapses replaced the whole expression with the branch the constant condition selected, but only checked that the dropped branch was droppable, not the condition they also removed. A condition like `(my @a := (3, 7))` declares and binds a variable, so folding it away lost the binding and let the empty declared array decide the branch instead of the bound value. This makes both collapses require the condition to be droppable too, so a condition that declares a variable is left for runtime. |
|||
| rakudo/main: 37fced81ad | (Nick Logan)++ (committed using GitHub Web editor) | 3 files Merge pull request #6273 from ugexe/ugexe/rakuast-collapse-keeps-declaration RakuAST: don't drop a declaration when collapsing a constant conditional |
|||
| rakudo/main: 4 commits pushed by (Nick Logan)++ | 00:08 | ||
| rakudo/main: 3adba69e5d | (Nick Logan)++ | 2 files Accept a Mu lhs and rhs in the test-assign metaops The test-assign metaops (//=, ||=, &&=, and the loose and control forms) take their left and right sides through a sigilless \lhs and a $rhs, both of which default to an Any type constraint. The legacy frontend inlines these so it never binds the parameters, but the RakuAST frontend calls the helper, so a Mu on either side died with "expected Any but got Mu". ... (9 more lines) |
00:09 | ||
| rakudo/main: f38ab67b32 | (Nick Logan)++ (committed using GitHub Web editor) | 2 files Merge pull request #6274 from ugexe/ugexe/test-assign-metaop-mu RakuAST: Accept a Mu lhs and rhs in the test-assign metaops |
|||
| rakudo/main: 14eaee6ddb | (Nick Logan)++ | 3 files RakuAST: don't redeclare a special variable bound in a sub-signature A block or routine declares an implicit $_, $/, and $! unless its signature already binds one. The check for an existing binding only looked at top-level parameters, so a special variable bound inside a destructuring sub-signature, such as the $_ in `-> (:value($_))`, went unnoticed. The implicit one was then declared alongside it, and the two declarations of the same lexical failed to ... (6 more lines) |
00:13 | ||
| rakudo/main: aa5e2499d0 | (Nick Logan)++ (committed using GitHub Web editor) | 3 files Merge pull request #6277 from ugexe/ugexe/rakuast-subsignature-special-vars RakuAST: don't redeclare a special variable bound in a sub signature |
|||
| ugexe | i wonder how far along in the ecosystem we actually are | 00:50 | |
| [Coke] | ===> Testing: ASTQuery:ver<0.0.7>:auth<zef:FCO> - that worked! | 02:14 | |
| ugexe: are we at a point where I could try a RAKUAST run (with a normal build of raku) and get you some useful output? | 02:15 | ||
| I think I'd have to do a modified run with maybe legacy 2026.05 and AST HEAD? | 02:16 | ||
| ugexe | couldnt hurt to try, although i see Blin has JSON::Fast as a dependency which still needs to be fixed to work on rakuast | 02:45 | |
|
08:59
finanalyst joined
09:15
finanalyst left
|
|||