Split rustc_mir
The `rustc_mir` crate is the second largest in the compiler.
This PR splits it up into 5 crates:
- rustc_borrowck;
- rustc_const_eval;
- rustc_mir_dataflow;
- rustc_mir_transform;
- rustc_monomorphize.
Improve diagnostics for unary plus operators (#88276)
This pull request improves the diagnostics emitted on parsing a unary plus operator. See #88276.
Before:
```
error: expected expression, found `+`
--> src/main.rs:2:13
|
2 | let x = +1;
| ^ expected expression
```
After:
```
error: leading `+` is not supported
--> main.rs:2:13
|
2 | let x = +1;
| ^
| |
| unexpected `+`
| help: try removing the `+`
```
This expands the API to be more flexible, allowing for more visitation patterns
on graphs. This will be useful to avoid extra datasets (and allocations) in
cases where the expanded DFS API is sufficient.
This also fixes a bug with the previous DFS constructor, which left the start
node not marked as visited (even though it was immediately returned).
Suggest deriving traits if possible
This only applies to builtin derives as I don't think there is a
clean way to get the available derives in typeck.
Closes#85851
Remove `hir::GenericBound::Unsized`
Rather than "moving" the `?Sized` bounds to the param bounds, just also check where clauses in `astconv`. I also did some related cleanup here, but that's not strictly neccesary. Also going to do a perf run here.
r? `@estebank`
These were deleted in https://reviews.llvm.org/D108614, and in C++ I
definitely see the argument for their removal. I didn't try and
propagate the changes up into higher layers of rustc in this change
because my initial goal was to get rustc working against LLVM HEAD
promptly, but I'm happy to follow up with some refactoring to make the
API on the Rust side match the LLVM API more directly (though the way
the enum works in Rust makes the API less scary IMO).
r? @nagisa cc @nikic
Fix handling of +whole-archive native link modifier.
This PR fixes a bug in `add_upstream_native_libraries` that led to the `+whole-archive` modifier being ignored when linking in native libs.
~~Note that the PR does not address the situation when `+whole-archive` is combined with `+bundle`.~~
`@wesleywiser's` commit adds validation code that turns combining `+whole-archive` with `+bundle` into an error.
Fixes https://github.com/rust-lang/rust/issues/88085.
r? `@petrochenkov`
cc `@wesleywiser` `@gcoakes`
Mmap the incremental data instead of reading it.
Instead of reading the full incremental state using `fs::read_file`, we memmap it using a private read-only file-backed map.
This allows the system to reclaim any memory we are not using, while ensuring we are not polluted by
outside modifications to the file.
Suggested in https://github.com/rust-lang/rust/pull/83036#issuecomment-800458082 by `@bjorn3`
Change scope of temporaries in match guards
Each pattern in a match arm has its own copy of the match guard in MIR, with its own temporary, so it has to be dropped before the the guards are joined to the single copy of the arm. This PR changes `then_else_break` to allow it to put the temporary in the innermost scope possible. This change isn't done for `if` expressions because that affects a large number of mir-opt tests and could more significantly affect performance.
closes#88649
r? `@oli-obk`
Remove SmallVector mention
SmallVector is long gone, as it's been first replaced
by OneVector in commit e5e6375352,
which then has been removed entirely in favour of SmallVec in
commit 130a32fa72.
Improve structured tuple struct suggestion
Previously, the span was just for the constructor name, which meant it
would result in syntactically-invalid code when applied. Now, the span
is for the entire expression.
I also changed it to use `span_suggestion_verbose`, for two reasons:
1. Now that the suggestion span has been corrected, the output is a bit
cluttered and hard to read. Putting the suggestion its own window
creates more space.
2. It's easier to see what's being suggested, since now the version
after the suggestion is applied is shown.
r? `@davidtwco`
Avoid invoking the hir_crate query to traverse the HIR
Walking the HIR tree is done using the `hir_crate` query. However, this is unnecessary, since `hir_owner(CRATE_DEF_ID)` provides the same information. Since depending on `hir_crate` forces dependents to always be executed, this leads to unnecessary work.
By splitting HIR and attributes visits, we can avoid an edge to `hir_crate` when trying to visit the HIR tree.
Stop allocating vtable entries for non-object-safe methods
Current a vtable entry is allocated for all associated fns, even if the method is not object-safe: https://godbolt.org/z/h7vx6f35T
As a result, each vtable for `Iterator`' currently consumes 74 `usize`s. This PR stops allocating vtable entries for those methods, reducing vtable size of each `Iterator` vtable to 7 `usize`s.
Note that this PR introduces will cause more invocations of `is_vtable_safe_method`. So a perf run might be needed. If result isn't favorable then we might need to query-ify `is_vtable_safe_method`.
Each pattern in a match arm has its own copy of the match guard in MIR,
with its own temporary, so it has to be dropped before the the guards
are joined to the single copy of the arm.
Provide `layout_of` automatically (given tcx + param_env + error handling).
After #88337, there's no longer any uses of `LayoutOf` within `rustc_target` itself, so I realized I could move the trait to `rustc_middle::ty::layout` and redesign it a bit.
This is similar to #88338 (and supersedes it), but at no ergonomic loss, since there's no funky `C: LayoutOf<Ty = Ty>` -> `Ty: TyAbiInterface<C>` generic `impl` chain, and each `LayoutOf` still corresponds to one `impl` (of `LayoutOfHelpers`) for the specific context.
After this PR, this is what's needed to get `trait LayoutOf` (with the `layout_of` method) implemented on some context type:
* `TyCtxt`, via `HasTyCtxt`
* `ParamEnv`, via `HasParamEnv`
* a way to transform `LayoutError`s into the desired error type
* an error type of `!` can be paired with having `cx.layout_of(...)` return `TyAndLayout` *without* `Result<...>` around it, such as used by codegen
* this is done through a new `LayoutOfHelpers` trait (and so is specifying the type of `cx.layout_of(...)`)
When going through this path (and not bypassing it with a manual `impl` of `LayoutOf`), the end result is that only the error case can be customized, the query itself and the success paths are guaranteed to be uniform.
(**EDIT**: just noticed that because of the supertrait relationship, you cannot actually implement `LayoutOf` yourself, the blanket `impl` fully covers all possible context types that could ever implement it)
Part of the motivation for this shape of API is that I've been working on querifying `FnAbi::of_*`, and what I want/need to introduce for that looks a lot like the setup in this PR - in particular, it's harder to express the `FnAbi` methods in `rustc_target`, since they're much more tied to `rustc` concepts.
r? `@nagisa` cc `@oli-obk` `@bjorn3`
Commit 95e096d6 changed a bunch of size checks already, but more have
been added, so this fixes the new ones the same way: the various size
checks that are conditional on target_arch = "x86_64" were not intended
to apply to x86_64-unknown-linux-gnux32, so add
target_pointer_width = "64" to the conditions.
Fix 2021 `dyn` suggestion that used code as label
The arguments to `span_suggestion` were in the wrong order, so the error
looked like this:
error[E0783]: trait objects without an explicit `dyn` are deprecated
--> src/test/ui/editions/dyn-trait-sugg-2021.rs:10:5
|
10 | Foo::hi(123);
| ^^^ help: <dyn Foo>: `use `dyn``
Now the error looks like this, as expected:
error[E0783]: trait objects without an explicit `dyn` are deprecated
--> src/test/ui/editions/dyn-trait-sugg-2021.rs:10:5
|
10 | Foo::hi(123);
| ^^^ help: use `dyn`: `<dyn Foo>`
This issue was only present in the 2021 error; the 2018 lint was
correct.
r? `@m-ou-se`
SmallVector is long gone, as it's been first replaced
by OneVector in commit e5e6375352,
which then has been removed entirely in favour of SmallVec in
commit 130a32fa72.
The arguments to `span_suggestion` were in the wrong order, so the error
looked like this:
error[E0783]: trait objects without an explicit `dyn` are deprecated
--> src/test/ui/editions/dyn-trait-sugg-2021.rs:10:5
|
10 | Foo::hi(123);
| ^^^ help: <dyn Foo>: `use `dyn``
Now the error looks like this, as expected:
error[E0783]: trait objects without an explicit `dyn` are deprecated
--> src/test/ui/editions/dyn-trait-sugg-2021.rs:10:5
|
10 | Foo::hi(123);
| ^^^ help: use `dyn`: `<dyn Foo>`
This issue was only present in the 2021 error; the 2018 lint was
correct.
Include debug info for the allocator shim
Issue Details:
In some cases it is necessary to generate an "allocator shim" to forward various Rust allocation functions (e.g., `__rust_alloc`) to an underlying function (e.g., `malloc`). However, since this allocator shim is a manually created LLVM module it is not processed via the normal module processing code and so no debug info is generated for it (if debugging info is enabled).
Fix Details:
* Modify the `debuginfo` code to allow creating debug info for a module without a `CodegenCx` (since it is difficult, and expensive, to create one just to emit some debug info).
* After creating the allocator shim add in basic debug info.
CrateLocator refactorings
This makes the `CrateLocator` a lot cleaner IMHO and much more self-contained. The last commit removes `extra_filename` from the crate metadata. This is an **insta-stable** change as it allows a crate like `libfoo-abc.rlib` to be used as dependency and then be renamed as `libfoo-bcd.rlib` while still being found as indirect dependency. This may reduce performance when there are a lot of versions of the same crate available as the extra filename won't be used to do an early rejection of crates before trying to load metadata, but it makes the logic to find the right filename a lot cleaner.
Detect bare blocks with type ascription that were meant to be a `struct` literal
Address part of #34255.
Potential improvement: silence the other knock down errors in `issue-34255-1.rs`.
Fix drop handling for `if let` expressions
MIR lowering for `if let` expressions is now more complicated now that
`if let` exists in HIR. This PR adds a scope for the variables bound in
an `if let` expression and then uses an approach similar to how we
handle loops to ensure that we reliably drop the correct variables.
Closes#88307
cc `@flip1995` `@richkadel` `@c410-f3r`
expand: Treat more macro calls as statement macro calls
This PR implements the suggestion from https://github.com/rust-lang/rust/pull/87981#issuecomment-906641052 and treats fn-like macro calls inside `StmtKind::Item` and `StmtKind::Semi` as statement macro calls, which is consistent with treatment of attribute invocations in the same positions and with token-based macro expansion model in general.
This also allows to remove a special case in `NodeId` assignment (previously tried in #87779), and to use statement `NodeId`s for linting (`assign_id!`).
r? `@Aaron1011`
Point at unclosed delimiters as part of the primary MultiSpan
Both the place where the parser encounters a needed closed delimiter and
the unclosed opening delimiter are important, so they should get the
same level of highlighting in the output.
_Context: https://twitter.com/mwk4/status/1430631546432675840_
Path remapping: Make behavior of diagnostics output dependent on presence of --remap-path-prefix.
This PR fixes a regression (#87745) with `--remap-path-prefix` where the flag stopped causing diagnostic messages to be remapped as well. The regression was introduced in https://github.com/rust-lang/rust/pull/83813 where we erroneously assumed that remapping of diagnostic messages was not desired anymore (because #70642 partially undid that functionality with nobody objecting).
The issue is fixed by making `--remap-path-prefix` remap diagnostic messages again, including for paths that have been remapped in upstream crates (e.g. the standard library). This means that "sysroot-localization" (implemented in #70642) is also disabled if `rustc` is invoked with `--remap-path-prefix`. The assumption is that once someone starts explicitly remapping paths they also don't want paths to their local Rust installation in their build output.
In the future we might want to give more fine-grained control over this behavior via compiler flags (see https://github.com/rust-lang/rfcs/pull/3127 for a related RFC). For now this PR is intended as a regression fix.
This PR is an alternative to https://github.com/rust-lang/rust/pull/88191, which makes diagnostic messages be remapped unconditionally. That approach, however, would effectively revert #70642.
Fixes https://github.com/rust-lang/rust/issues/87745.
cc `@cbeuw`
r? `@ghost`
Preserve most sub-obligations in the projection cache
Fixes https://github.com/rust-lang/rust/issues/85360
When we evaluate a projection predicate, we may produce sub-obligations. During trait evaluation, evaluating these sub-obligations might cause us to produce `EvaluatedToOkModuloRegions`.
When we cache the result of projection in our projection cache, we try to throw away some of the sub-obligations, so that we don't need to re-evaluate/process them the next time we need to perform this particular projection. However, we may end up throwing away predicates that will (recursively) evaluate to `EvaluatedToOkModuloRegions`. If we do, then the result of evaluating a predicate will depend on the state of the predicate cache - this is global untracked state, which interacts badly with incremental compilation.
To fix this, we now only discard global predicates that evaluate to `EvaluatedToOk`. This ensures that any predicates that (may) evaluate to `EvaluatedToOkModuloRegions` are kept in the cache, and influence the results of any queries which perform this projection.
Correct doc comments inside `use_expr_visitor.rs`
Just a simple update. I haven't changed any content inside the comments, as they still seem correct. Have a wonderful rest of the day 🙃
Remove redundant `Span` in `QueryJobInfo`
Previously, `QueryJobInfo` was composed of two parts: a `QueryInfo` and
a `QueryJob`. However, both `QueryInfo` and `QueryJob` have a `span`
field, which seem to be the same. So, the `span` was recorded twice.
Now, `QueryJobInfo` is composed of a `QueryStackFrame` (the other field
of `QueryInfo`) and a `QueryJob`. So, now, the `span` is only recorded
once.
For two reasons:
1. Now that the suggestion span has been corrected, the output is a bit
cluttered and hard to read. Putting the suggestion its own window
creates more space.
2. It's easier to see what's being suggested, since now the version
after the suggestion is applied is shown.
(And same for tuple variants.)
Previously, the span was just for the constructor name, which meant it
would result in syntactically-invalid code when applied. Now, the span
is for the entire expression.
MIR lowering for `if let` expressions is now more complicated now that
`if let` exists in HIR. This PR adds a scope for the variables bound in
an `if let` expression and then uses an approach similar to how we
handle loops to ensure that we reliably drop the correct variables.
Previously, `QueryJobInfo` was composed of two parts: a `QueryInfo` and
a `QueryJob`. However, both `QueryInfo` and `QueryJob` have a `span`
field, which seem to be the same. So, the `span` was recorded twice.
Now, `QueryJobInfo` is composed of a `QueryStackFrame` (the other field
of `QueryInfo`) and a `QueryJob`. So, now, the `span` is only recorded
once.
Rollup of 9 pull requests
Successful merges:
- #86376 (Emit specific warning to clarify that `#[no_mangle]` should not be applied on foreign statics or functions)
- #88040 (BTree: remove Ord bound from new)
- #88053 (Fix the flock fallback implementation)
- #88350 (add support for clobbering xer, cr, and cr[0-7] for asm! on OpenPower/PowerPC)
- #88410 (Remove bolding on associated constants)
- #88525 (fix(rustc_typeck): produce better errors for dyn auto trait)
- #88542 (Use the return value of readdir_r() instead of errno)
- #88548 (Stabilize `Iterator::intersperse()`)
- #88551 (Stabilize `UnsafeCell::raw_get()`)
Failed merges:
r? `@ghost`
`@rustbot` modify labels: rollup
Emit specific warning to clarify that `#[no_mangle]` should not be applied on foreign statics or functions
Foreign statics and foreign functions should not have `#[no_mangle]` applied, as it does nothing to the name and has some extra hidden behavior that is normally unwanted. There was an existing warning for this, but it says the attribute is only allowed on "statics or functions", which to the user can be confusing.
This PR adds a specific version of the unused `#[no_mangle]` warning that explains that the target is a *foreign* static or function and that they do not need the attribute.
Fixes#78989
Add bit removal methods to SparseBitMatrix and factor *BitSet relational methods into more extensible trait
I need the ability to clear the bits out of a row from `SparseBitMatrix`. Currently, all the mutating methods only allow insertion of bits, and there is no way to get access to the underlying data.
One approach is simply to make `ensure_row` public, since it grants `&mut` access to the underlying `HybridBitSet`. This PR adds the `pub` modifier. However, presumably this method was private for a reason, so I'm open to other designs. I would prefer general mutable access to the rows, because that way I can add many mutating operations (`clear`, `intersect`, etc.) without filing a PR each time :-)
r? `@ecstatic-morse`
Introduce `let...else`
Tracking issue: #87335
The trickiest part for me was enforcing the diverging else block with clear diagnostics. Perhaps the obvious solution is to expand to `let _: ! = ..`, but I decided against this because, when a "mismatched type" error is found in typeck, there is no way to trace where in the HIR the expected type originated, AFAICT. In order to pass down this information, I believe we should introduce `Expectation::LetElseNever(HirId)` or maybe add `HirId` to `Expectation::HasType`, but I left that as a future enhancement. For now, I simply assert that the block is `!` with a custom `ObligationCauseCode`, and I think this is clear enough, at least to start. The downside here is that the error points at the entire block rather than the specific expression with the wrong type. I left a todo to this effect.
Overall, I believe this PR is feature-complete with regard to the RFC.
Issue Details:
In some cases it is necessary to generate an "allocator shim" to forward various Rust allocation functions (e.g., `__rust_alloc`) to an underlying function (e.g., `malloc`). However, since this allocator shim is a manually created LLVM module it is not processed via the normal module processing code and so no debug info is generated for it (if debugging info is enabled).
Fix Details:
* Modify the `debuginfo` code to allow creating debug info for a module without a `CodegenCx` (since it is difficult, and expensive, to create one just to emit some debug info).
* After creating the allocator shim add in basic debug info.
Concrete regions can show up in mir borrowck if the originated from there
We used to not encounter them here, because we took regions from typeck's opaque type resolution by renumbering them. We don't do that anymore. Instead mir borrock does all the logic, and it can handle concrete regions just fine, as long as it created them itself.
fixes#83190 which was introduced by #87287
r? `@spastorino`
Clean up the lowering of AST items
This PR simplifies and improves `rustc_ast_lowering::item` in various minor ways. The reasons for the changes should mostly be self evident, though I'm happy to specifically explain anything if needed.
These changes used to be part of #88019, but I removed them after it was pointed out that some of my other changes to `rustc_ast_lowering` were unnecessary. It felt like a bad idea to clean up code which I didn't even need to touch anymore.
r? `@cjgillot`
Warn when [T; N].into_iter() is ambiguous in the new edition.
Fixes https://github.com/rust-lang/rust/issues/88475
In https://github.com/rust-lang/rust/issues/88475, a situation was found where `[T; N].into_iter()` becomes *ambiguous* in the new edition. This is different than the case where `(&[T; N]).into_iter()` resolves differently, which was the only case handled by the `array_into_iter` lint. This is almost identical to the new-traits-in-the-prelude problem. Effectively, due to the array-into-iter hack disappearing in Rust 2021, we effectively added `IntoIterator` to the 'prelude' in Rust 2021 specifically for arrays.
This modifies the prelude collisions lint to detect that case and emit a `array_into_iter` lint in that case.
Don't use `guess_head_span` in `predicates_of` for foreign span
Previously, the result of `predicates_of` for a foreign trait
would depend on the *current* state of the corresponding source
file in the foreign crate. This could lead to ICEs during incremental
compilation, since the on-disk contents of the upstream source file
could potentially change without the upstream crate being recompiled.
Additionally, this ensure that that the metadata we produce for a crate
only depends on its *compiled* upstream dependencies (e.g an rlib or
rmeta file), *not* the current on-disk state of the upstream crate
source files.
update const generics feature gates
**tl;dr: split const generics into three features: `adt_const_params`, `const_generics_defaults` and `generic_const_exprs`**
continuing the work of `@BoxyUwU` in #88324, this PR
- renames `feature(const_evaluatable_checked)` to `feature(generic_const_exprs)` which now doesn't need any other feature gate to work. Previously `feature(const_evaluatable_checked)` was only useful in combination with `feature(const_generics)`.
- completely removes `feature(lazy_normalization_consts)`. This feature only supplied the parents generics to anonymous constants, which is pretty useless as generic anon consts are only allowed with `feature(generic_const_exprs)` anyways.
- moves the ability to use additional const param types from `feature(const_generics)` into `feature(adt_const_params)`. As `feature(const_generics)` is now mostly useless without `feature(generic_const_exprs)` we also remove that feature flag.
- updates tests, removing duplicates and unnecessary revisions in some cases and also deletes all unused `*.stderr` files.
I not also remove the ordering restriction for const and type parameters if any of the three const generics features is active.
This ordering restriction feels like the only "real" use of the current `feature(const_generics)` right now so this change isn't a perfect solution, but as I intend to stabilize the ordering - and `feature(const_generics_defaults)` - in the very near future, I think this is acceptable for now.
---
cc `@rust-lang/project-const-generics` about the new feature names and this change in general.
I don't think we need any external approval for this change but I do intend to publish an update to the const generics tracking issue the day this PR lands, so I don't want this merged yet.
Apologies to whoever ends up reviewing this PR 😅❤️
r? rust-lang/project-const-generics
