Repository navigation
Conversation
|
Comment on behalf of stas at php.net: merged |
EdmondDantes
added a commit
to true-async/php-src
that referenced
this pull request
May 22, 2026
…uest completes Every coroutine doing I/O on one descriptor parks on the shared per-handle async-IO event (data->async_io->event), so any single operation's completion notifies them all. php_stdiop_write()/php_stdiop_read() did not re-check req->completed after waking, so a spuriously woken coroutine disposed its own request while that request's libuv write was still in flight — libuv then wrote into freed memory (heap corruption / SEGV in uv__async_io). Both await paths now loop, re-suspending until their own request completed. Found via the ext/async I/O chaos suite; regression test lives in ext/async (tests/io/083-concurrent_async_write.phpt). The structural fix — per-request completion events instead of the broadcast — is true-async/php-async#130.
EdmondDantes
added a commit
to true-async/php-src
that referenced
this pull request
May 22, 2026
heyJordanParker
added a commit
to heyJordanParker/php-sharp
that referenced
this pull request
Oct 11, 2026
Problem - master took #101, which pins Mago at #118 (90e8d64) and rewrites Sets.sharp:75 and Sets.inc:75 so a Set's changing method has a place. - This branch pinned php#129's head (bce91aa). Both sides changed the pin line. php#129 merged as 82593a4, which did not compile, and mago-sharp php#153 fixed it as 625c278. Solution - Take master's Sets.sharp and Sets.inc as they are. - SHARP_MAGO_COMMIT pins 625c278, the merge of php#153, which carries both #118 and php#129. The bridge sources the header is generated from are unchanged since bce91aa, so the pin is the only line the generated header changes. Zend/tests/sharp/ Sets.sharp, Sets.inc from master: the copied Set is held in a local before add ext/sharp/ sharp_unit.h conflict: SHARP_MAGO_COMMIT is 625c278, the merge of mago-sharp php#153
heyJordanParker
added a commit
to heyJordanParker/php-sharp
that referenced
this pull request
Oct 11, 2026
… to Mago fe442e9 Problem: Master gained #101, which pins Mago 90e8d64 and rewrites `Sets.sharp:75` so a `Set` that `copy()` returns is added to through a local. This branch had made the same fix on one line, so `Sets.sharp`, `Sets.inc` and the pin in `ext/sharp/sharp_unit.h` conflicted. The generics Mago branch now also holds Mago php#129, which refuses an `int` raised to a negative power the checker knows. Its flow typing knows `let b = -1;`, so it refuses `Overflow.constantNegativePower` and `Overflow.zeroNegativePower`, the two fixtures that test the engine's `ArithmeticError` when opcache propagates a negative exponent. Every pin that holds php#129 refuses them, on master too. Solution: - `Sets.sharp` and `Sets.inc` take master's two-line fix, which owns it. - `ext/sharp/sharp_unit.h` is the header Mago fe442e9 generates. - `Overflow.sharp` declares each exponent `int b = -1;` and `int n = -1;`. The checker then knows only `int`, and opcache still propagates the constant: the optimized `Overflow::constantNegativePower` is `T0 = POW int(2) int(-1)`, which throws. Each line keeps its line number. `sharp/bin/test --passes Zend/tests/sharp/overflow_pow.phpt` passes under every optimizer pass alone and both JITs. Zend/tests/sharp/ ├── Sets.sharp, Sets.inc <- conflict: master's fix └── Overflow.sharp <- the two negative exponents are typed int locals ext/sharp/sharp_unit.h <- conflict: pinned to Mago fe442e9
heyJordanParker
added a commit
to heyJordanParker/php-sharp
that referenced
this pull request
Oct 11, 2026
… to Mago 9cb1a28 Problem: Master gained #100, which pins Mago php#129 and keeps `Overflow.constantNegativePower` and `Overflow.zeroNegativePower` compiling under php#129's negative-power refusal by writing each exponent as `~0`. This branch had fixed the same two lines with `int b = -1;` and `int n = -1;`, so `Overflow.sharp` and the pin in `ext/sharp/sharp_unit.h` conflicted. Solution: - `Overflow.sharp` takes master's `~0`, which owns the fix. This branch's typed locals are gone. - `ext/sharp/sharp_unit.h` is the header Mago 9cb1a28 generates. `php -n --ri sharp` prints "Mago commit => 9cb1a28…". - `sharp/bin/test` passes 202 of 202 tests, and `sharp/bin/test --passes Zend/tests/sharp/overflow_pow.phpt` passes in all 19 modes. Zend/tests/sharp/Overflow.sharp <- conflict: master's `~0` ext/sharp/sharp_unit.h <- conflict: pinned to Mago 9cb1a28
heyJordanParker
added a commit
to heyJordanParker/php-sharp
that referenced
this pull request
Oct 13, 2026
Problem: Master gained php#129, which refuses an `int` where a `float` is expected in PHP# code. Both sides had renamed upstream's `strict_nonnull` switch to `sharp_rules`, and sixteen files conflicted where each side built a comparison result its own way. This branch has `ComparisonResult::for_dialect` and `ComparisonResult::nested`, and php#129 has `ComparisonResult::with_sharp_rules(bool)` at each call. Composed, three faults appeared: - A PHP# type parameter whose bound is `string|int` passed into a `float` was named as an int-for-float refusal. This branch's opaque type-parameter path in `atomic_comparator::is_contained_by` folded the bound's parts with `all()` into one shared result, so the first refused part, the `int`, left `int_for_float` set. php#129 fixed that fold for templates with `are_parts_contained_by`, and its codex test caught the gap. - Master itself does not compile. php#129 added `assigned_value` to `static_property_assignment::analyze`, and #118's `check_changed_collection` in `method_call.rs` still calls it with six arguments. Master CI at 82593a4 failed. - Local assignments lost this branch's `explain_blocked_substitution` in php#129's rewrite of the ordinary message. Solution: - One constructor path. `for_dialect` and `nested` stay, `with_sharp_rules` goes, and `int_for_float` joins the struct. The `sharp_rules` doc names both PHP# rules: the `float` that refuses an `int`, and the opaque type parameter. - `are_parts_contained_by` is `pub(super)`, and the PHP# type-parameter input folds its bound through it, so a bound names its `int` only when the `int` is the one reason. - `check_changed_collection` passes `None` as `assigned_value`, as the `property_assignment::analyze` call beside it does. A changed collection has no assigned expression. - Every php#129 report stays, and the ordinary message in each place keeps `explain_blocked_substitution`. crates/ ├── codex/src/ttype/comparator/ │ ├── mod.rs <- conflict: for_dialect, nested, int_for_float │ ├── atomic_comparator.rs <- conflict; a type parameter's bound folds through are_parts_contained_by │ ├── union_comparator.rs <- conflict; are_parts_contained_by is pub(super) │ └── array_, callable_, generic_, iterable_comparator.rs <- conflict: nested() └── analyzer/src/ ├── expression/assignment/{mod,array_assignment,property_assignment,static_property_assignment}.rs │ <- conflict: php#129's reports with for_dialect and explain_blocked_substitution ├── expression/call/method_call.rs <- master's arity break fixed ├── invocation/arguments.rs <- conflict: both imports, for_dialect └── statement/{return,function_like/mod,class_like/property,class_like/constant}.rs <- conflict: for_dialect
heyJordanParker
added a commit
to heyJordanParker/php-sharp
that referenced
this pull request
Oct 13, 2026
Problem: PR php#152 must be up to date with master. Master gained php#164, which leaves `unreachable!` lines out of patch-coverage. It also gained php#163, which adds List insert, remove and clear, and php#165, which fixes a project with a law that compiles twice. php#163's `takes_operands` calls `ComparisonResult::with_sharp_rules(true)`, which this branch replaced in its php#129 merge, so the merged tree did not compile. Solution: merge master. Git merged every file on its own. `takes_operands` builds its result with `ComparisonResult::for_dialect(Dialect::Sharp)`, the branch's one constructor path. Master's `with_sharp_rules(true)` made the same result. The analyzer, semantics, bridge and names suites and the root compile tests pass on the merged tree, and clippy reports nothing. crates/ ├── analyzer/src/expression/binary/utils.rs for_dialect(Dialect::Sharp) in takes_operands ├── analyzer/src/expression/call/method_call.rs master's search runs after the call's type is known ├── names/src/lib.rs master's insert, remove and clear └── sharp-bridge/ master's List changes beside this branch's checks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
No description provided.