Polymorphization
This PR implements an analysis to detect when functions could remain polymorphic during code generation.
Fixes#46477
r? @eddyb
cc @rust-lang/wg-mir-opt @nikomatsakis @pnkfelix
disallow non-static lifetimes in const generics
Disallow non-static lifetimes in const generics in order to to patch over an ICE caused when we encounter a non-static lifetime in a const generic during borrow checking. This restriction may be relaxed in the future, but we need more discussion before then, and in the meantime we should still deal with this ICE.
Fixes issue #60814
wf: check foreign fn decls for well-formedness
Fixes#73252 and fixes#73253.
This PR extends current well-formedness checking to apply to foreign function declarations, re-using the existing machinery for regular functions. In doing this, later parts of the compiler (such as the `improper_ctypes` lint) can rely on being operations not failing as a result of invalid code which would normally be caught earlier.
This commit introduces a `FiniteBitSet` type which replaces the manual
bit manipulation which was being performed in polymorphization.
Signed-off-by: David Wood <david@davidtw.co>
This commit replaces the `-Z polymorphize-errors` debugging flag with a
`#[rustc_polymorphize_error]` attribute for use on functions.
Signed-off-by: David Wood <david@davidtw.co>
This commit normalizes function signatures for instances before
substituting, a workaround for polymorphization considering
parameters unused when they show up in the signature, but not the
body (due to being normalized).
Unfortunately, this causes test output to change with the parallel
compiler only.
Signed-off-by: David Wood <david@davidtw.co>
This commit implements the `unused_generic_params` query, an initial
version of polymorphization which detects when an item does not use
generic parameters and is being needlessly monomorphized as a result.
Signed-off-by: David Wood <david@davidtw.co>
This commit extends current well-formedness checking to apply to foreign
function declarations, re-using the existing machinery for regular
functions. In doing this, later parts of the compiler (such as the
`improper_ctypes` lint) can rely on being operations not failing as a
result of invalid code which would normally be caught earlier.
Signed-off-by: David Wood <david@davidtw.co>
Only skip impls of foreign unstable traits
Previously unstable impls were skipped, which meant that any impl with an unstable method would get skipped.
Fixes#74531.
do not try fetching the ancestors of errored trait impls
fixes#74483
While building the specialization graph, we use `tcx.all_impls` which discards impls with incorrect self types,
we do however call `trait_def.ancestors` with these impls which caused an ICE as they aren't part of the
specialization graph.
This has been put in place to patch over an ICE caused when we encounter
a non-static lifetime in a const generic during borrow checking. This
restriction may be relaxed in the future, but we need more discussion
before then, and in the meantime we should still deal with this ICE.
Fixes issue #60814
Generating the coverage map
@tmandry @wesleywiser
rustc now generates the coverage map and can support (limited)
coverage report generation, at the function level.
Example commands to generate a coverage report:
```shell
$ BUILD=$HOME/rust/build/x86_64-unknown-linux-gnu
$ $BUILD/stage1/bin/rustc -Zinstrument-coverage \
$HOME/rust/src/test/run-make-fulldeps/instrument-coverage/main.rs
$ LLVM_PROFILE_FILE="main.profraw" ./main
called
$ $BUILD/llvm/bin/llvm-profdata merge -sparse main.profraw -o main.profdata
$ $BUILD/llvm/bin/llvm-cov show --instr-profile=main.profdata main
```
r? @wesleywiser
Rust compiler MCP rust-lang/compiler-team#278
Relevant issue: #34701 - Implement support for LLVMs code coverage instrumentation
Make unreachable_unchecked a const fn
This PR makes `std::hint::unreachable_unchecked` a const fn so we can use it inside a const function.
r? @RalfJung
Fixes#53188.
rustc_metadata: Make crate loading fully speculative
Instead of reporting `span_err`s on the spot crate loading errors are now wrapped into the `CrateError` enum and returned, so they are reported only at the top level `resolve_crate` call, and not reported at all if we are resolving speculatively with `maybe_resolve_crate`.
As a result we can attempt loading crates for error recovery (e.g. import suggestions) without any risk of producing extra errors.
Also, this means better separation between error reporting and actual logic.
Fixes https://github.com/rust-lang/rust/issues/55103
Fixes https://github.com/rust-lang/rust/issues/56590
improper_ctypes_definitions: allow `Box`
Addresses https://github.com/rust-lang/rust/pull/72700#issuecomment-659449386.
This PR stops linting against `Box` in `extern "C" fn`s for the `improper_ctypes_definitions` lint - boxes are documented to be FFI-safe.
cc @alexcrichton @CryZe
Compare tagged/niche-filling layout and pick the best one
Finishes up #71045, and so fixes#63866.
cc @eddyb
r? @nikomatsakis (since @eddyb wrote the first commit)
intra-doc links: resolve modules in the type namespace
Fixes https://github.com/rust-lang/rust/issues/62830
Modules actually live in the type namespace, not all three, and it's not possible to clash a type with a module.
Add Arguments::as_str().
There exist quite a few macros in the Rust ecosystem which use `format_args!()` for formatting, but special case the one-argument case for optimization:
```rust
#[macro_export]
macro_rules! some_macro {
($s:expr) => { /* print &str directly, no formatting, no buffers */ };
($s:expr, $($tt:tt)*) => { /* use format_args to write to a buffer first */ }
}
```
E.g. [here](7a961f0fbe/src/macros.rs (L48-L58)), [here](20f9a9e223/src/macros.rs (L9-L17)), and [here](7b679cd6da/px4/src/logging.rs (L45-L52)).
The problem with these is that a forgotten argument such as in `some_macro!("{}")` will not be diagnosed, but just prints `"{}"`.
With this PR, it is possible to handle the no-arguments case separately *after* `format_args!()`, while simplifying the macro. Then these macros can give the proper error about a missing argument, just like `print!("{}")` does, while still using the same optimized implementation as before.
This is even more important with [RFC 2795](https://github.com/rust-lang/rfcs/pull/2795), to make sure `some_macro!("{some_variable}")` works as expected.
Fix MinGW `run-make-fulldeps` tests
`compiler-rt-works-on-mingw` and `libs-search-path` were not ran because `only-mingw` doesn't match any target.
Enabled and verified few ignored tests with `windows-gnu` toolchain. They are still ignored on MSVC since I'm not experienced with this target.
Make some Option methods const
Tracking issue: #67441
Constantify the following methods of `Option`:
- `as_ref`
- `is_some`
- `is_none`
- `iter` (not sure about this one, but it is possible, and will be useful when const traits are a thing)
cc @rust-lang/wg-const-eval @rust-lang/libs
This commit stops linting against `Box` in `extern "C" fn`s for the
`improper_ctypes_definitions` lint - boxes are documented to be
FFI-safe.
Signed-off-by: David Wood <david@davidtw.co>
When ran on Windows `cp` will follow symlink: `checkout/build/<target>/<stage>/lib/rustlib/src/rust`.
It points to `checkout` which means the test will get stuck in copying loop until there is no space left.
