$ the-wire · showcase
rust-analyzer kills a false `Sized` diagnostic, rust removes old solver coherence
By RepoJournal · Filed · About Rust · Composed from the cited sources · methodology
rust-analyzer's sizedness check now walks the subtyping table, and the compiler's old solver coherence support is gone; the rest of the day is interpreter and CI hygiene.
fix: use sub-unification roots in type_var_is_sized rust-lang/rust-analyzer
Sizedness checks previously only followed inference variables linked through the regular unification table, so variables related to a `Sized` obligation through subtyping went unnoticed and IDE users saw a bogus `expected i32, found dyn Send`. The fix introduces `UseSubtyping` and consults the sub-unification table; other callers keep existing behavior.
Rollup merge of #161491 - sjwang05:no-old-coherence, r=khyperia rust-lang/rust
The last remaining place the old solver was still used for coherence is removed, in a change best reviewed commit by commit with ignore-whitespace. If you maintain a type-system-adjacent fork or out-of-tree tool that leans on the older coherence path, this is the commit that ends it.
Auto merge of #163754 - JonathanBrouwer:rollup-fdgcPnD, r=JonathanBrouwer rust-lang/rust
A six-PR rollup brings a future-compat warning for `#[panic_handler]` on `unsafe fn`, stabilization of `CStr::display`, documentation for the `no_main` and `repr` attributes, and a new `--frontend-threads` option for `./x perf`. If you have an `unsafe fn` panic handler, expect the new FCW on your next nightly bump.
ops: Resolve `unpinned-uses` and `ref-version-mismatch` rust-lang/josh-sync
josh-sync's workflow files now satisfy the `unpinned-uses` and `ref-version-mismatch` rules as part of getting more `zizmor` lints enforced across `rust-lang` repos through `crabwatch`. The author also flagged inconsistent versions of `actions/checkout` (7.0.1 vs 4.4.0) and `actions/create-github-app-token` (2.2.2 vs 3.2.0) as follow-ups.
fix(report): Specify UTF-8 when `open`ing files rust-lang/surveys
The survey `report` library now specifies UTF-8 when opening files, fixing a case where Windows 11's CP-1252 default made it unable to parse survey data at all. Existing report scripts are deliberately left untouched, so they keep their historical output.
