Fix panicking Option unwraping in match arm analysis
Hi, first PR here!
I've noticed my IDE sometimes briefly becoming pretty slow to respond while writing Rust. When checking the logs I found reams of this same error repeating itself.
```
thread 'Worker' panicked at 'called `Option::unwrap()` on a `None` value'
crates/ide-assists/src/handlers/convert_match_to_let_else.rs:90:46
```
RA seemed to have been panicking on virtually every keystroke I made whenever I was part way through writing/refactoring a match statement of relevance to this assist.
The fix in this PR should be self-explanatory.
Migrate `rustc_lint` lint diagnostics
Part 2 of [Migrate `rustc_lint` errors to `SessionDiagnostic`](https://github.com/rust-lang/rust/pull/100776)
r? `@davidtwco`
# TODO
- [x] Refactor some lints manually implementing `DecorateLint` to use `Option<Subdiagnostic>`.
- [x] Add `#[rustc_lint_diagnostics]` to lint functions in `context.rs`.
- [x] Migrate `hidden_unicode_codepoints.rs`.
- [x] Migrate `UnsafeCode` in `builtin.rs`.
- [x] Migrate the rest of `builtin.rs`.
interior-mutable types should be 'static' rather than 'const
with the 'tracking' feature enabled, `limit::Limit` is interior mutable, so the `const` `hir_def::nameres::mod_resolution::MOD_DEPTH_LIMIT` is probably meant to be `static` rather than `const`.
with `const` you get a new copy of it everywhere it's used, instead of the shared, global value that seems to be the intent.
This fix brought to you by the catchily-named `clippy::declare_interior_mutable_const` lint.
Use CI LLVM in `test-various` builder
It was disabled because it needs `lld`, but since #104748 was merged it is no longer needed.
This will speed this test, since it no longer needs to build LLVM.
Refine search for const and function assoc items
This changes our searching behavior, before we always associated all usages and definitions of associated items with all implementations of a trait and the trait itself. Now, when searching for references of a an associated trait item, we still do the same and consider all implementations, but when searching for an associated item of an implementation we now only consider the uses of that specific implementations associated item.
This does not affect associated type aliases as we unfortunately are missing information in the IDE layer here still.
Move src/test to the root
See MCP at rust-lang/compiler-team#573
There may be more changes needed.
The first commit is just the move of the files:
You can check that the first commit did not do anything else than renames by running
```
git diff --diff-filter=r -M100% <rust-lang remote>/master <first commit hash>
```
The output should be empty, because the filter excludes renames, and the match threshold for qualifying a rename is 100%.
The second one is mostly a "find and replace" of `src/test` to `tests` and whatever is needed to make CI pass.
What is left to do:
---
- [x] Move directory
- [ ] Change references to `src/test`
- [x] Change references in-tree
- [ ] Change references in submodules / out-of-tree docs
- [x] Make CI pass:
- [x] Fix tidy
- [x] Fix tests
- [x] Bless tests if needed (shouldn't normally)
- [ ] Merge it !
Fix stack overflow in recursive AST walk in early lint
The src/test/ui/issues/issue-74564-if-expr-stack-overflow.rs test case added to verify https://github.com/rust-lang/rust/issues/74564 still crashes with a stack overflow on s390x-ibm-linux.
Symptom is a very deep recursion in compiler/rustc_lint/src/early.rs:
fn visit_expr(&mut self, e: &'a ast::Expr) {
self.with_lint_attrs(e.id, &e.attrs, |cx| {
lint_callback!(cx, check_expr, e);
ast_visit::walk_expr(cx, e);
})
}
(where walk_expr recursively calls back into visit_expr). The crash happens at a nesting depth of over 17000 stack frames when using the default 8 MB stack size on s390x.
This patch fixes the problem by adding a ensure_sufficient_stack call to the with_lint_attrs routine (which also should take care of all the other mutually recursive visitors here).
Fixes part of https://github.com/rust-lang/rust/issues/105383.
minor: tweak wording and remove blank issue template
There's already a "blank issue" link at the bottom, and some users seem to pick this one because it's the first.
Revert "Use ZWNJ to prevent VSCode from forming ligatures between hints and code"
Doesn't actually seem to fix the bug, it seems to be theme dependent and I am not sure what causes it
Reverts rust-lang/rust-analyzer#13886
Consider return type when giving various method suggestions
1. Fix a bug in method probe where we weren't normalizing `xform_ret_ty` for non-`impl` method candidates. This shouldn't affect happy-path code, since we only use `xform_ret_ty` when probing methods for diagnostics (I think).
2. Pass the return type expectation down to `lookup_probe`/`probe_for_name` usages in diagnostics. Added a few UI tests to gate against bad suggestions.
3. Make a `FnCtxt::lookup_probe_for_diagnostic` which properly passes down `IsSuggestion(true)`. Should help suppress other weird notes in some corner cases.
Rollup of 9 pull requests
Successful merges:
- #105034 (Add example for iterator_flatten)
- #105708 (Enable atomic cas for bpf targets)
- #106175 (Fix bad import suggestion with nested `use` tree)
- #106204 (No need to take opaques in `check_type_bounds`)
- #106387 (Revert "bootstrap: Get rid of `tail_args` in `stream_cargo`")
- #106636 (Accept old spelling of Fuchsia target triples)
- #106639 (update Miri)
- #106640 (update test for inductive canonical cycles)
- #106647 (rustdoc: merge common CSS for `a`)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
