$ the-wire · showcase
Span loses its ordering, MutTy disappears, libc ctest deprecates generate_test
By RepoJournal · Filed · About Rust · Composed from the cited sources · methodology
A day of compiler-internal subtraction: `Span` and `SpanData` lose their `Ord` impls, `MutTy` is removed from both AST and HIR, and libc's ctest grows a documented build path while retiring `generate_test`.
Rollup merge of #163303 - nnethercote:fix-Span-eq-ord, r=oli-obk rust-lang/rust
`PartialOrd` and `Ord` are gone from `Span` and `SpanData` because they disagreed with the `PartialEq`/`Eq` impls, and because span ordering is inherently dubious. Any code sorting or comparing spans by order now fails to compile; the removal landed alongside the same change in HIR-level `MutTy` cleanup.
remove `ast::MutTy` rust-lang/rust
`ast::MutTy` is removed outright, part of a two-step cleanup that also took `hir::MutTy` with it. Out-of-tree tools that match on the AST or HIR node will need to be updated against the new representation.
ctest: Add deprecate `generate_test` rust-lang/libc
`ctest`'s `generate_test` is deprecated in favor of the new `build_test`, and a top-level example now documents how to drive a build. `try_build_test` exists so the error can be printed with `Display` rather than the `Debug` output `.unwrap()` produces, replacing the `#[doc_hidden]` function that was previously the only way to build tests.
deps: Lock file maintenance (main) rust-lang/libc
Renovate refreshed all lock files on `main` to the latest dependency versions. Automerge is disabled by config, so this one needs a human merge.
test: Skip fdim bitwise NAN tests rust-lang/compiler-builtins
The `fdim` bitwise NaN tests are being skipped after `check_f128` panicked on a `-inf +qNaN` input, reproduced in unsigned `libm`. Unrelated to that, the crates.io index picked up routine version bumps for `greentic-sorla-pack`, `greentic-sorla-ir`, `greentic-sorla-lang`, `ox_content_profiler`, and `scv-channels`.