Region inference: Use outlives-static constraints in constraint search
Revise the extra `r: 'static` constraints added upon universe issues to add an explanation, and use that explanation during constraint blame search. This greatly simplifies the region inference logic, which now does not need to reverse-engineer the event that caused a region to outlive `'static`.
This cosmetically changes the output of two UI tests. I blessed them i separate commits with separate motivations, but that can of course be squashed as desired. We probably want that.
The PR was extracted out of rust-lang/rust#130227 and consists of one-third of its functional payload.
r? lcnr
Use -Zmir-opt-level=0 in tests for MIR building
The mir-opt test suite currently defaults all tests in it to `-Zmir-opt-level=4`, so if a test is trying to test MIR _building_ not optimizations and it is in that directory, it _must_ override the default mir-opt-level.
ci: Begin running ui tests with `rust.debuginfo-level-tests=1`
To reduce risk of regressing on generating debuginfo e.g. in the form of ICE:s. This will also ensure that future ui tests work with different debuginfo levels. See https://github.com/rust-lang/rust/issues/61117.
When I looked at run time for different CI jobs, **x86_64-gnu-debug** was far from the bottleneck, so it should be fine to make it perform more work.
A handful of tests are failing so we need to force debuginfo=0 on those for now.
We'll start small with debuginfo=1. We'll step up to debuginfo=2 once most (all?) tests can handle debuginfo=1. There are more failures with debuginfo=2 than with debuginfo=1.
make rustdoc::invalid_html_tags more robust
best reviewed a commit at a time.
I kept finding more edge case so I ended up having to make quite significant changes to the parser in order to make it preserve state across events and handle multiline attributes correctly.
fixesrust-lang/rust#145529
Implement some more checks in `ptr_guaranteed_cmp`.
* Pointers with different residues modulo their allocations' least common alignment are never equal.
* Pointers to the same static allocation are equal if and only if they have the same offset.
* Pointers to different non-zero-sized static allocations are unequal if both point within their allocation, and not on opposite ends.
Tracking issue for `const_raw_ptr_comparison`: <https://github.com/rust-lang/rust/issues/53020>
This should not affect `is_null`, the only usage of this intrinsic on stable.
Closes https://github.com/rust-lang/rust/issues/144584
Support non-defining uses in HIR typeck
This changes the impl of `NormalizesTo` for opaque types to be structural during HIR typeck. The previous impl equated region variables of the opaque type key with existing entries which can result in spurious leak check errors and also results in mismatches with MIR borrowck, theoretically causing ICE.
The approach is very similar to rust-lang/rust#145244 in MIR typeck:
- we collect all uses of opaque types during HIR typeck
- before writeback, we search for *defining uses*
- the opaque type key has fully universal generic args modulo regions
- the hidden type has no infer vars
- we use these defining uses to compute the concrete type for the opaque and map it to the definition site
- we use this concrete type to check the type of all uses of opaques during HIR typeck. This also constrains infer vars in non-defining uses
Fixes https://github.com/rust-lang/trait-system-refactor-initiative/issues/135, fixes https://github.com/rust-lang/trait-system-refactor-initiative/issues/49.
r? `@BoxyUwU`
Account for impossible bounds making seemingly unsatisfyable dyn-to-dyn casts
Fixes https://github.com/rust-lang/rust/issues/141806
When we have an impossible where clause like `dyn Trait<u8>: Sized`, this may make a dyn-to-dyn cast like `dyn Trait<()> -> dyn trait<u8>` to successfully type check as if it were a wide-to-thin ptr cast (discarding metadata):
16ad385579/compiler/rustc_hir_typeck/src/cast.rs (L862-L865)
In borrowck, we are expecting that the only meaningful dyn-to-dyn cast to be a metadata-preserving wide-to-wide ptr cast, which requires that the principals of the dyn pointers are equal. Borrowck additionally assumes that these principals have already been proven equal *modulo regions*, and we thus ICE since `Trait<u8>` and `Trait<()>` do not unify:
16ad385579/compiler/rustc_borrowck/src/type_check/mod.rs (L1481-L1524)
This PR fixes this ICE by checking whether the RHS of the cast is considered to be Sized in the environment of the MIR typeck, and if so then skipping over this dyn->dyn principal compatibility check.
r? `@lcnr` perhaps?
Make sure to treat only param where clauses as inherent
See the description in the test file.
This PR fixes a bug introduced by rust-lang/rust#141333, where we considered non-`Param` where clauses to be "inherent" for the purpose of method probing, which leads to both changes in method ambiguity (see test) and also import usage linting (and thus fixes https://github.com/rust-lang/rust/issues/145185).
r? `@lcnr`
All function pointers are currently treated as unaligned anyway;
any change implementing function pointer alignment during consteval should add
tests that it works properly on arm::t32 functions.
Use unnamed lifetime spans as primary spans for `MISMATCHED_LIFETIME_SYNTAXES`
Fixes https://github.com/rust-lang/rust/issues/145772
This PR changes the primary span(s) of the `MISMATCHED_LIFETIME_SYNTAXES` to point to the *unnamed* lifetime spans in both the inputs and *outputs* of the function signature. As reported in rust-lang/rust#145772, this should make it so that IDEs highlight the spans of the actionable part of this lint, rather than just the (possibly named) input spans like they do today.
This could be tweaked further perhaps, for example for `fn foo(_: T<'_>) -> T`, we don't need to highlight the elided lifetime if the actionable part is to change only the return type to `T<'_>`, but I think it's improvement on what's here today, so I think that should be follow-up since I think the logic might get a bit hairy.
cc ```@shepmaster```
Add lint against integer to pointer transmutes
# `integer_to_ptr_transmutes`
*warn-by-default*
The `integer_to_ptr_transmutes` lint detects integer to pointer transmutes where the resulting pointers are undefined behavior to dereference.
### Example
```rust
fn foo(a: usize) -> *const u8 {
unsafe {
std::mem::transmute::<usize, *const u8>(a)
}
}
```
```
warning: transmuting an integer to a pointer creates a pointer without provenance
--> a.rs:1:9
|
158 | std::mem::transmute::<usize, *const u8>(a)
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
= note: this is dangerous because dereferencing the resulting pointer is undefined behavior
= note: exposed provenance semantics can be used to create a pointer based on some previously exposed provenance
= help: if you truly mean to create a pointer without provenance, use `std::ptr::without_provenance_mut`
= help: for more information about transmute, see <https://doc.rust-lang.org/std/mem/fn.transmute.html#transmutation-between-pointers-and-integers>
= help: for more information about exposed provenance, see <https://doc.rust-lang.org/std/ptr/index.html#exposed-provenance>
= note: `#[warn(integer_to_ptr_transmutes)]` on by default
help: use `std::ptr::with_exposed_provenance` instead to use a previously exposed provenance
|
158 - std::mem::transmute::<usize, *const u8>(a)
158 + std::ptr::with_exposed_provenance::<u8>(a)
|
```
### Explanation
Any attempt to use the resulting pointers are undefined behavior as the resulting pointers won't have any provenance.
Alternatively, `std::ptr::with_exposed_provenance` should be used, as they do not carry the provenance requirement or if the wanting to create pointers without provenance `std::ptr::without_provenance_mut` should be used.
See [std::mem::transmute] in the reference for more details.
[std::mem::transmute]: https://doc.rust-lang.org/std/mem/fn.transmute.html
--------
People are getting tripped up on this, see https://github.com/rust-lang/rust/issues/128409 and https://github.com/rust-lang/rust/issues/141220. There are >90 cases like these on [GitHub search](https://github.com/search?q=lang%3Arust+%2Ftransmute%3A%3A%3Cu%5B0-9%5D*.*%2C+%5C*const%2F&type=code).
Fixes https://github.com/rust-lang/rust-clippy/issues/13140
Fixes https://github.com/rust-lang/rust/issues/141220
Fixes https://github.com/rust-lang/rust/issues/145523
`@rustbot` labels +I-lang-nominated +T-lang
cc `@traviscross`
r? compiler
Fix ICE when validating transmuting ZST to inhabited enum
MIR validation attempts to determine the number of bytes needed to represent the size of the source type to compute the discriminant for the inhabited target enum. For a ZST source, there is no source data to use as a discriminant so no proper runtime check can be generated.
Since that should never be possible, insert a delayed bug to ensure the problem has been properly reported to the user by the type checker.
Fixesrust-lang/rust#145786
match exhaustiveness diagnostics: show a trailing comma on singleton tuple consructors in witness patterns (and clean up a little)
Constructor patterns of type `(T,)` are written `(pat,)`, not `(pat)`. However, exhaustiveness/usefulness diagnostics would print them as `(pat)` when e.g. providing a witness of non-exhaustiveness and suggesting adding arms to make matches exhaustive; this would result in an error when applied.
rust-analyzer already prints the trailing comma, so it doesn't need changing.
This also includes some cleanup in the second commit, with justification in the commit message.
change HIR typeck region uniquification handling approach
rust-lang/rust#144405 causes structural lookup of opaque types to not work during HIR typeck, so instead avoid uniquifying goals and instead only reprove them if MIR borrowck actually encounters an error.
This doesn't perfectly maintain the property that HIR typeck succeeding implies that MIR typeck succeeds, instead weakening this check to only guarantee that HIR typeck implies that MIR typeck succeeds modulo region uniquification. This means we still get the actually desirable ICEs if we MIR building is broken or we forget to check some property in HIR typeck, without having to deal with the fallout of uniquification in HIR typeck itself.
We report errors using the original obligation sources of HIR typeck so diagnostics aren't that negatively impacted either.
Here's the history of region uniquification while working on the new trait solver:
- rust-lang/rust#107981
- rust-lang/rust#110180
- rust-lang/rust#114117
- rust-lang/rust#130821
- rust-lang/rust#144405
- rust-lang/rust#145706 <- we're here 🎉
r? `@BoxyUwU`
MIR validation attempts to determine the number of bytes needed to
represent the size of the source type to compute the discriminant for
the inhabited target enum. For a ZST source, there is no source data to
use as a discriminant so no proper runtime check can be generated.
Since that should never be possible, insert a delayed bug to ensure the
problem has been properly reported to the user by the type checker.
Pointers with different residues modulo their least common allocation alignment are never equal.
Pointers to the same static allocation are equal if and only if they have the same offset.
Strictly in-bounds (in-bounds and not one-past-the-end) pointers to different static allocations are always unequal.
A pointer cannot be equal to an integer if `ptr-int` cannot be null.
Also adds more tests for `ptr_guaranteed_cmp`.
Co-authored-by: Ralf Jung <post@ralfj.de>
