Replace in-tree `rustc_apfloat` with the new version of the crate
Replace the in-tree version of `rustc_apfloat` with the new version of the crate which has been correctly licensed. The new crate incorporates upstream changes from LLVM since the original port was done including many correctness fixes and has been extensively fuzz tested to validate correctness.
Fixes#100233Fixes#102403Fixes#113407Fixes#113409Fixes#55993Fixes#93224Closes#93225Closes#109573
Remove -Z diagnostic-width
This removes the `-Z diagnostic-width` option since it is ignored and does nothing. `-Z diagnostic-width` was stabilized as `--diagnostic-width` in #95635. It is not entirely clear why the `-Z` flag was kept, but in part its final use was removed in #102216, but the `-Z` flag itself was not removed.
Squelch a noisy rustc_expand unittest
The test `rustc_parse::tests::bad_path_expr_1` prints an error message to stderr, circumventing libtest's stderr intercept. This causes noise when running tests, in particular they show up 16 times on the GitHub Actions summary page. The solution here is to not use an error emitter that prints to stderr, and instead check that the correct error is generated.
lint/ctypes: fix `()` return type checks
Fixes#113436.
`()` is normally FFI-unsafe, but is FFI-safe when used as a return type. It is also desirable that a transparent newtype for `()` is FFI-safe when used as a return type.
In order to support this, when a type was deemed FFI-unsafe, because of a `()` type, and was used in return type - then the type was considered FFI-safe. However, this was the wrong approach - it didn't check that the `()` was part of a transparent newtype! The consequence of this is that the presence of a `()` type in a more complex return type would make it the entire type be considered safe (as long as the `()` type was the first that the lint found) - which is obviously incorrect.
Instead, this logic is removed, and after [consultation with t-lang](https://github.com/rust-lang/rust/issues/113436#issuecomment-1640756721), I've fixed the bugs and inconsistencies and made `()` FFI-safe within types.
I also refactor a function, but that's not too exciting.
interpret: make read/write methods generic
Instead of always having to call `into()` to convert things to `PlaceTy`/`OpTy`, make the relevant methods generic. This also means that when we read from an `MPlaceTy`, we avoid creating an intermediate `PlaceTy`.
This makes it feasible to remove the `Copy` from `MPlaceTy`. All the other `*Ty` interpreter types already had their `Copy` removed a while ago so this is only consistent. (And in fact we had one function that accidentally took `MPlaceTy` instead of `&MPlaceTy`.)
Split some functions with many arguments into builder pattern functions
r? `@estebank`
This doesn't resolve all of the ones in rustc, mostly because I need to do other cleanups in order to be able to use some builder derives from crates.io
Works around https://github.com/rust-lang/rust/issues/90672 by making `x test rustfmt --bless` format itself instead of testing that it is formatted
new unstable option: -Zwrite-long-types-to-disk
This option guards the logic of writing long type names in files and instead using short forms in error messages in rustc_middle/ty/error behind a flag. The main motivation for this change is to disable this behaviour when running ui tests.
This logic can be triggered by running tests in a directory that has a long enough path, e.g. /my/very-long-path/where/rust-codebase/exists/
This means ui tests can fail depending on how long the path to their file is.
Some ui tests actually rely on this behaviour for their assertions, so for those we enable the flag manually.
Double check that hidden types match the expected hidden type
Fixes https://github.com/rust-lang/rust/issues/113278 specifically, but I left a TODO for where we should also add some hardening.
It feels a bit like papering over the issue, but at least this way we don't get unsoundness, but just surprising errors. Errors will be improved and given spans before this PR lands.
r? `@compiler-errors` `@lcnr`
Don't say that a type is uncallable if its fn signature has errors in it
This is fallout from #106309, where we don't consider param-env candidates that reference errors because they unify with everything. This means, however, that we don't consider an APIT like `impl Fn(MissingType)` isn't considered to implement `Fn`, for example.
We can double-check that with a weaker heuristic [`extract_callable_info`](https://doc.rust-lang.org/nightly/nightly-rustc/rustc_hir_typeck/fn_ctxt/struct.FnCtxt.html#method.extract_callable_info), and suppress the knock-down error using that.
Fixes#113566
Rollup of 7 pull requests
Successful merges:
- #114008 (coverage: Obtain the `__llvm_covfun` section name outside a per-function loop)
- #114014 (builtin_macros: expect raw strings too)
- #114043 (docs(LazyLock): add example pass local LazyLock variable to struct)
- #114051 (Add regression test for invalid "unused const" in method)
- #114052 (Suggest `{Option,Result}::as_ref()` instead of `cloned()` in some cases)
- #114058 (Add help for crate arg when crate name is invalid)
- #114060 (abi: unsized field in union - assert to delay bug )
r? `@ghost`
`@rustbot` modify labels: rollup
abi: unsized field in union - assert to delay bug
Fixes#113279.
> Unions cannot have unsized fields, and as such, layout computation for
unions asserts that each union field is sized (as this would normally
have halted compilation earlier).
>
> However, if a generator ends up with an unsized local - a circumstance
in which an error will always have been emitted earlier, for example, if
attempting to dereference a `&str` - then the generator transform will
produce a union with an unsized field.
>
> Since https://github.com/rust-lang/rust/pull/110107, later passes will be run, such as constant propagation,
and can attempt layout computation on the generator, which will result
in layout computation of `str` in the context of it being a field of a
union - and so the aforementioned assertion would cause an ICE.
>
> It didn't seem appropriate to try and detect this case in the MIR body
and skip this specific pass; tainting the MIR body or delaying a bug
from the generator transform (or elsewhere) wouldn't prevent this either
(as neither would prevent the later pass from running); and tainting when
the deref of `&str` is reported, if that's possible, would unnecessarily
prevent potential other errors from being reported later in compilation,
and is very tailored to this specific case of getting a unsized type in
a generator.
>
> Given that this circumstance can only happen when an error should have
already been reported, the correct fix appears to be just changing the
assert to a delayed bug. This will still assert if there is some
circumstance where this occurs and no error has been reported, but it
won't crash the compiler in this instance.
While debugging this, I noticed a translation ICE in a delayed bug, so I fixed that too:
> During borrowck, the `MultiSpan` from a buffered diagnostic is cloned and
used to emit a delayed bug indicating a diagnostic was buffered - when
the buffered diagnostic is translated, then the cloned `MultiSpan` may
contain labels which can only render with the diagnostic's arguments, but
the delayed bug being emitted won't have those arguments. Adds a function
which clones `MultiSpan` without also cloning the contained labels, and
use this function when creating the buffered diagnostic delayed bug.
Suggest `{Option,Result}::as_ref()` instead of `cloned()` in some cases
Fixes#114050
When we have an expr available that produces the type expectation, we can suggest appending `.as_ref()` to the span, instead of cloning the expr producing the mismatch
coverage: Obtain the `__llvm_covfun` section name outside a per-function loop
This section name is always constant for a given target, but obtaining it from LLVM requires a few intermediate allocations. There's no need to do so repeatedly from inside a per-function loop.
Normalize the RHS of an `Unsize` goal in the new solver
`Unsize` goals are... tricky. Not only do they structurally match on their self type, but they're also structural on their other type parameter. I'm pretty certain that it is both incomplete and also just plain undesirable to not consider normalizing the RHS of an unsize goal. More practically, I'd like for this code to work:
```rust
trait A {}
trait B: A {}
impl A for usize {}
impl B for usize {}
trait Mirror {
type Assoc: ?Sized;
}
impl<T: ?Sized> Mirror for T {
type Assoc = T;
}
fn main() {
// usize: Unsize<dyn B>
let x = Box::new(1usize) as Box<<dyn B as Mirror>::Assoc>;
// dyn A: Unsize<dyn B>
let y = x as Box<<dyn A as Mirror>::Assoc>;
}
```
---
In order to achieve this, we add `EvalCtxt::normalize_non_self_ty` (naming modulo bikeshedding), which *must* be used for all non-self type arguments that are structurally matched in candidate assembly. Currently this is only necessary for `Unsize`'s argument, but I could see future traits requiring this (hopefully rarely) in the future. It uses `repeat_while_none` to limit infinite looping, and normalizes the self type until it is no longer an alias.
Also, we need to fix feature gate detection for `trait_upcasting` and `unsized_tuple_coercion` when HIR typeck has unnormalized types. We can do that by checking the `ImplSource` returned by selection, which necessitates adding a new impl source for tuple upcasting.
Unions cannot have unsized fields, and as such, layout computation for
unions asserts that each union field is sized (as this would normally
have halted compilation earlier).
However, if a generator ends up with an unsized local - a circumstance
in which an error will always have been emitted earlier, for example, if
attempting to dereference a `&str` - then the generator transform will
produce a union with an unsized field.
Since #110107, later passes will be run, such as constant propagation,
and can attempt layout computation on the generator, which will result
in layout computation of `str` in the context of it being a field of a
union - and so the aforementioned assertion would cause an ICE.
It didn't seem appropriate to try and detect this case in the MIR body
and skip this specific pass; tainting the MIR body or delaying a bug
from the generator transform (or elsewhere) wouldn't prevent this either
(as neither would prevent the later pass from running); and tainting when
the deref of `&str` is reported, if that's possible, would unnecessarily
prevent potential other errors from being reported later in compilation,
and is very tailored to this specific case of getting a unsized type in
a generator.
Given that this circumstance can only happen when an error should have
already been reported, the correct fix appears to be just changing the
assert to a delayed bug. This will still assert if there is some
circumstance where this occurs and no error has been reported, but it
won't crash the compiler in this instance.
Signed-off-by: David Wood <david@davidtw.co>
interpret: Unify projections for MPlaceTy, PlaceTy, OpTy
For ~forever, we didn't really have proper shared code for handling projections into those three types. This is mostly because `PlaceTy` projections require `&mut self`: they might have to `force_allocate` to be able to represent a project part-way into a local.
This PR finally fixes that, by enhancing `Place::Local` with an `offset` so that such an optimized place can point into a part of a place without having requiring an in-memory representation. If we later write to that place, we will still do `force_allocate` -- for now we don't have an optimized path in `write_immediate` that would avoid allocation for partial overwrites of immediately stored locals. But in `write_immediate` we have `&mut self` so at least this no longer pollutes all our type signatures.
(Ironically, I seem to distantly remember that many years ago, `Place::Local` *did* have an `offset`, and I removed it to simplify things. I guess I didn't realize why it was so useful... I am also not sure if this was actually used to achieve place projection on `&self` back then.)
The `offset` had type `Option<Size>`, where `None` represent "no projection was applied". This is needed because locals *can* be unsized (when they are arguments) but `Place::Local` cannot store metadata: if the offset is `None`, this refers to the entire local, so we can use the metadata of the local itself (which must be indirect); if a projection gets applied, since the local is indirect, it will turn into a `Place::Ptr`. (Note that even for indirect locals we can have `Place::Local`: when the local appears in MIR, we always start with `Place::Local`, and only check `frame.locals` later. We could eagerly normalize to `Place::Ptr` but I don't think that would actually simplify things much.)
Having done all that, we can finally properly abstract projections: we have a new `Projectable` trait that has the basic methods required for projecting, and then all projection methods are implemented for anything that implements that trait. We can even implement it for `ImmTy`! (Not that we need that, but it seems neat.) The visitor can be greatly simplified; it doesn't need its own trait any more but it can use the `Projectable` trait. We also don't need the separate `Mut` visitor any more; that was required only to reflect that projections on `PlaceTy` needed `&mut self`.
It is possible that there are some more `&mut self` that can now become `&self`... I guess we'll notice that over time.
r? `@oli-obk`
Reimplement C-str literals
This reverts #113334, cc `@fmease.`
While converting lexer tokens to ast Tokens in `rustc_parse`, we check the edition of the span of the token. If the edition < 2021, we split the token into two, one being the identifier and other being the str literal.
