Rollup of 9 pull requests
Successful merges:
- #122842 (Don't emit an error about failing to produce a file with a specific name if user never gave an explicit name)
- #122881 (Delegation: fix ICE on `bound_vars` divergence)
- #122910 (Validate that we're only matching on unit struct for path pattern)
- #122970 (Use `chunk_by` when building `ReverseSccGraph`)
- #122988 (add even more tests! )
- #122999 (Fix unpretty UI test when /tmp does not exist)
- #123001 (Rename `{enter,exit}_lint_attrs` to `check_attributes{,_post}`)
- #123022 (Add `async-closures/once.rs` back to cranelift tests)
- #123034 (Add a bunch of needs-unwind annotations to tests)
r? `@ghost`
`@rustbot` modify labels: rollup
Rename `{enter,exit}_lint_attrs` to `check_attributes{,_post}`
Several places in Clippy want to check all the attributes of a node, we end up using `hir().attrs()` from several different `check_` functions (e.g. [in our doc lints](95c62ffae9/clippy_lints/src/doc/mod.rs (L396))) but this is error prone, we recently found that doc lints weren't triggering on struct fields for example
I went to add a `check_attributes` function but realised `enter_lint_attrs` is already this, the rename is to encourage their use
Also removes `LateContextAndPass::visit_attribute` since it's unused - `visit_attribute` for HIR visitors is only called by `hir().walk_attributes()` which lint passes do not use
Validate that we're only matching on unit struct for path pattern
Resolution doesn't validate that we only really take `CtorKind::Unit` in path patterns, since all it sees is `Res::SelfCtor(def_id)`. Check this instead during pattern typeck.
r? petrochenkov
Fixes#122809
Delegation: fix ICE on `bound_vars` divergence
Fixes https://github.com/rust-lang/rust/issues/122550.
Bug was caused by divergence between lowered type and corresponding `bound_vars` in `late_bound_vars_map`. In this patch `bound_vars` calculation for delegation item is moved from `lower_fn_ty` to `resolve_bound_vars` query.
r? `@petrochenkov`
Don't emit an error about failing to produce a file with a specific name if user never gave an explicit name
Fixes#122509
You can ask `rustc` to produce some intermediate results with `--emit foo`, this operation comes in two flavors: `--emit asm` and `--emit asm=foo.s`. First one produces one or more `.s` files without any name guarantees, second one renames it into `foo.s`. Second version only works when compiler produces a single file - for asm files this means using a single compilation unit for example.
In case compilation produced more than a single file `rustc` runs following check to emit some warnings:
```rust
if crate_output.outputs.contains_key(&output_type) {
// 2) Multiple codegen units, with `--emit foo=some_name`. We have
// no good solution for this case, so warn the user.
sess.dcx().emit_warn(errors::IgnoringEmitPath { extension });
} else if crate_output.single_output_file.is_some() {
// 3) Multiple codegen units, with `-o some_name`. We have
// no good solution for this case, so warn the user.
sess.dcx().emit_warn(errors::IgnoringOutput { extension });
} else {
// 4) Multiple codegen units, but no explicit name. We
// just leave the `foo.0.x` files in place.
// (We don't have to do any work in this case.)
}
```
Comment in the final `else` branch implies that if user didn't ask for a specific name - there's no need to emit warnings. However because of the internal representation of `crate_output.outputs` - this doesn't work as expected: if user asked to produce an asm file without giving it an implicit name it will contain `Some(None)`.
To fix the problem new code actually checks if user gave an explicit name. I think this was an original intentional behavior, at least comments imply that.
Move `--sysroot` argument out of the argument file to fix miri issue
Fixes https://github.com/rust-lang/miri/issues/3404.
For now, miri needs this argument to be moved out of the arg file so they can update it if needed.
cc `@RalfJung`
r? `@notriddle`
Temporarily remove nnethercote from the review rotation.
I will be on vacation for the next three weeks. I will re-add myself when I return.
r? `@nnethercote`
CFI: Support complex receivers
Right now, we only support rewriting `&self` and `&mut self` into `&dyn MyTrait` and `&mut dyn MyTrait`. This expands it to handle the full gamut of receivers by calculating the receiver based on *substitution* rather than based on a rewrite. This means that, for example, `Arc<Self>` will become `Arc<dyn MyTrait>` appropriately with this change.
This approach also allows us to support associated type constraints as well, so we will correctly rewrite `&self` into `&dyn MyTrait<T=i32>`, for example.
r? ```@workingjubilee```
CFI: Handle dyn with no principal
In user-facing Rust, `dyn` always has at least one predicate following it. Unfortunately, because we filter out marker traits from receivers at callsites and `dyn Sync` is, for example, legal, this results in us having `dyn` types with no predicates on occasion in our alias set encoding. This patch handles cases where there are no predicates in a `dyn` type which are relevant to its alias set.
Fixes#122998
r? workingjubilee
Clean up unnecessary headers/flags in coverage mir-opt tests
During #122542, I noticed that some of the headers and flags I had copied over from `tests/mir-opt/instrument_coverage.rs` were unnecessary. And while working to remove those, I noticed even more that could be removed or replaced.
Clarify transmute example
The example claims using an iterator will copy the entire vector, but this is not true in practice thanks to internal specializations in the stdlib (see https://godbolt.org/z/cnxo3MYs5 for confirmation that this doesn't reallocate nor iterate over the vec's elements). Since neither the copy nor the optimization is guaranteed I opted for saying that they _may_ happen.
Add more comments to the bootstrap code that handles `tests/coverage`
At the bootstrap level, coverage tests are a bit more complicated than other test suites, because we want to run the same set of test files in multiple different modes, in a way that's convenient and flexible when invoked manually.
This PR adds a few more comments to explain what's going on.
For the `MiddleDot` case, current behaviour:
- For a case like `1.2`, `sym1` is `1` and `sym2` is `2`, and `self.token`
holds `1.2`.
- It creates a new ident token from `sym1` that it puts into `self.token`.
- Then it does `bump_with` with a new dot token, which moves the `sym1`
token into `prev_token`.
- Then it does `bump_with` with a new ident token from `sym2`, which moves the
`dot` token into `prev_token` and discards the `sym1` token.
- Then it does `bump`, which puts whatever is next into `self.token`,
moves the `sym2` token into `prev_token`, and discards the `dot` token
altogether.
New behaviour:
- Skips creating and inserting the `sym1` and dot tokens, because they are
unnecessary.
- This also demonstrates that the comment about `Spacing::Alone` is
wrong -- that value is never used. That comment was added in #77250,
and AFAICT it has always been incorrect.
The commit also expands comments. I found this code hard to read
previously, the examples in comments make it easier.
Pass in the span for the field rather than using `prev_token`.
Also rename it `mk_expr_tuple_field_access`, because it doesn't do any
actual parsing, it just creates an expression with what it's given.
Not much of a clarity win by itself, but unlocks additional subsequent
simplifications.
Replace `mir_built` query with a hook and use mir_const everywhere instead
A small perf improvement due to less dep graph handling.
Mostly just a cleanup to get rid of one of our many mir queries
Previously, we only rewrote `&self` and `&mut self` receivers. By
instantiating the method from the trait definition, we can make this
work work with arbitrary legal receivers instead.
ci: Build gccjit from a git archive
A full `git clone` of GCC includes quite a lot of history, and it's
completely unnecessary for building it in CI. We can use a GitHub
archive URL to get a simple tarball that is much faster to download.
Also, the `gcc-build` directory can be removed after install to reduce
the image size even further.
In user-facing Rust, `dyn` always has at least one predicate following
it. Unfortunately, because we filter out marker traits from receivers at
callsites and `dyn Sync` is, for example, legal, this results in us
having `dyn` types with no predicates on occasion in our alias set
encoding. This patch handles cases where there are no predicates in a
`dyn` type which are relevant to its alias set.
Fixes#122998
Rollup of 10 pull requests
Successful merges:
- #122737 (conditionally ignore fatal diagnostic in the SilentEmitter)
- #122757 (Fixed the `private-dependency` bug)
- #122886 (add test for #90192)
- #122937 (Unbox and unwrap the contents of `StatementKind::Coverage`)
- #122949 (Add a regression test for #117310)
- #122962 (Track run-make-support lib in common inputs stamp)
- #122977 (Rename `Arguments::as_const_str` to `as_statically_known_str`)
- #122983 (Fix build failure on ARM/AArch64/PowerPC/RISC-V FreeBSD/NetBSD)
- #122984 (panic-in-panic-hook: formatting a message that's just a string is risk-free)
- #122992 (std:🧵 refine available_parallelism for solaris/illumos.)
r? `@ghost`
`@rustbot` modify labels: rollup
std:🧵 refine available_parallelism for solaris/illumos.
Rather than the system-wide available cpus fallback solution, we fetch the cpus bound to the current process.
panic-in-panic-hook: formatting a message that's just a string is risk-free
This slightly improves the output in the 'panic while processing panic' case if the panic message does not involve any formatting. Follow-up to https://github.com/rust-lang/rust/pull/122930.
r? ``@Amanieu``
Rename `Arguments::as_const_str` to `as_statically_known_str`
While `const` has a particular meaning about language guarantees, here
we need a fuzzier notion like whether constant propagation was
effective, and `statically_known` is the best term we have for now.
r? ``@RalfJung``
Track run-make-support lib in common inputs stamp
So that when the support library gets modified, `run-make` tests are forced to re-run instead of being still ignored as if nothing changed.
Fixes#122961.
Add a regression test for #117310Closes#117310
It seems to have been fixed in `rustc 1.79.0-nightly (1388d7a06 2024-03-20)` or before, so just adding a regression test for it.
Unbox and unwrap the contents of `StatementKind::Coverage`
The payload of coverage statements was historically a structure with several fields, so it was boxed to avoid bloating `StatementKind`.
Now that the payload is a single relatively-small enum, we can replace `Box<Coverage>` with just `CoverageKind`.
This patch also adds a size assertion for `StatementKind`, to avoid accidentally bloating it in the future.
``@rustbot`` label +A-code-coverage
Fixed the `private-dependency` bug
Fixed the private-dependency bug: If the directly dependent crate is loaded last and is not configured with `--extern`, it may be incorrectly set to `private-dependency`
Fixes#122756
conditionally ignore fatal diagnostic in the SilentEmitter
This change is primarily meant to allow rustfmt to ignore all diagnostics when using the `SilentEmitter`. Back in #121301 the `SilentEmitter` was shared between rustc and rustfmt. This changed rustfmt's behavior from ignoring all diagnostic to emitting fatal diagnostics, which lead to https://github.com/rust-lang/rustfmt/issues/6109.
These changes allow rustfmt to maintain its previous behaviour when using the `SilentEmitter`, while allowing rustc code to still emit fatal diagnostics.
