| Age | Commit message (Collapse) | Author | Lines |
|---|
|
tag/niche terminology cleanup
The term "discriminant" was used in two ways throughout the compiler:
* every enum variant has a corresponding discriminant, that can be given explicitly with `Variant = N`.
* that discriminant is then encoded in memory to store which variant is active -- but this encoded form of the discriminant was also often called "discriminant", even though it is conceptually quite different (e.g., it can be smaller in size, or even use niche-filling).
After discussion with @eddyb, this renames the second term to "tag". The way the tag is encoded can be either `TagEncoding::Direct` (formerly `DiscriminantKind::Tag`) or `TagEncoding::Niche` (formerly `DiscrimianntKind::Niche`).
This finally resolves some long-standing confusion I had about the handling of variant indices and discriminants, which surfaced in https://github.com/rust-lang/rust/pull/72419.
(There is also a `DiscriminantKind` type in libcore, it remains unaffected. I think this corresponds to the discriminant, not the tag, so that seems all right.)
r? @eddyb
|
|
remove visit_terminator_kind from MIR visitor
For some reason, we had both `visit_terminator` and `visit_terminator_kind`. In contrast, for `Statement` we just have `visit_statement`. So this cleans things up by removing `visit_terminator_kind` and porting its users to `visit_terminator`.
|
|
Make all uses of ty::Error delay a span bug
r? @eddyb
A second attempt at https://github.com/rust-lang/rust/pull/70245
resolves https://github.com/rust-lang/rust/issues/70866
|
|
Fix link error with #[thread_local] introduced by #71192
r? @oli-obk
|
|
Don't run generator transform when there's a TyErr
Not sure if this might cause any problems later on, but we shouldn't be hitting codegen or const eval for the produced MIR anyways, so it should be fine.
cc https://github.com/rust-lang/rust/issues/72685#issuecomment-643749020
|
|
|
|
|
|
|
|
|
|
store `ObligationCause` on the heap
Stores `ObligationCause` on the heap using an `Rc`.
This PR trades off some transient memory allocations to reduce the size of–and thus the number of instructions required to memcpy–a few widely used data structures in trait solving.
|
|
This mode is unnecessary because it's always ok to evaluate the
right-hand side of assignments even if the left-hand side should not
have reads propagated.
|
|
|
|
Addresses feedback from @oli-obk (Thanks!)
|
|
|
|
|
|
As suggested in PR feedback:
https://github.com/rust-lang/rust/pull/73011#discussion_r435728923
This allows count_code_region() to be handled like a normal intrinsic so
the InstanceDef::InjectedCode variant is no longer needed.
|
|
This initial version only injects counters at the top of each function.
Rust Coverage will require injecting additional counters at each
conditional code branch.
|
|
|
|
This commit intends to fix an accidental regression from #70846. The
goal of #70846 was to build compiler-builtins with a maximal number of
CGUs to ensure that each module in the source corresponds to an object
file. This high degree of control for compiler-builtins is desirable to
ensure that there's at most one exported symbol per CGU, ideally
enabling compiler-builtins to not conflict with the system libgcc as
often.
In #70846, however, only part of the compiler understands that
compiler-builtins is built with many CGUs. The rest of the compiler
thinks it's building with `sess.codegen_units()`. Notably the
calculation of `sess.lto()` consults `sess.codegen_units()`, which when
there's only one CGU it disables ThinLTO. This means that
compiler-builtins is built without ThinLTO, which is quite harmful to
performance! This is the root of the cause from #73135 where intrinsics
were found to not be inlining trivial functions.
The fix applied in this commit is to remove the special-casing of
compiler-builtins in the compiler. Instead the build system is now
responsible for special-casing compiler-builtins. It doesn't know
exactly how many CGUs will be needed but it passes a large number that
is assumed to be much greater than the number of source-level modules
needed. After reading the various locations in the compiler source, this
seemed like the best solution rather than adding more and more special
casing in the compiler for compiler-builtins.
Closes #73135
|
|
Rollup of 10 pull requests
Successful merges:
- #72707 (Use min_specialization in the remaining rustc crates)
- #72740 (On recursive ADT, provide indirection structured suggestion)
- #72879 (Miri: avoid tracking current location three times)
- #72938 (Stabilize Option::zip)
- #73086 (Rename "cyclone" to "apple-a7" per changes in upstream LLVM)
- #73104 (Example about explicit mutex dropping)
- #73139 (Add methods to go from a nul-terminated Vec<u8> to a CString)
- #73296 (Remove vestigial CI job msvc-aux.)
- #73304 (Revert heterogeneous SocketAddr PartialEq impls)
- #73331 (extend network support for HermitCore)
Failed merges:
r? @ghost
|
|
Miri: avoid tracking current location three times
Miri tracks the current instruction to execute in the call stack, but it also additionally has two `TyCtxtAt` that carry a `Span` that also tracks the current instruction. That is quite silly, so this PR uses `TyCtxt` instead, and then uses a method for computing the current span when a `TyCtxtAt` is needed. Having less redundant (semi-)global state seems like a good improvement to me. :D
To keep the ConstProp errors the same, I had to add the option to `error_to_const_error` to overwrite the span. Also for some reason this changes cycle errors a bit -- not sure if we are now better or worse as giving those queries the right span. (It is unfortunately quite easy to accidentally use `DUMMY_SP` by calling the query on a `TyCtxt` instead of a `TyCtxtAt`.)
r? @oli-obk @eddyb
|
|
Rollup of 10 pull requests
Successful merges:
- #71824 (Check for live drops in constants after drop elaboration)
- #72389 (Explain move errors that occur due to method calls involving `self`)
- #72556 (Fix trait alias inherent impl resolution)
- #72584 (Stabilize vec::Drain::as_slice)
- #72598 (Display information about captured variable in `FnMut` error)
- #73336 (Group `Pattern::strip_*` method together)
- #73341 (_match.rs: fix module doc comment)
- #73342 (Fix iterator copied() documentation example code)
- #73351 (Update E0446.md)
- #73353 (structural_match: non-structural-match ty closures)
Failed merges:
r? @ghost
|
|
Display information about captured variable in `FnMut` error
Fixes #69446
When we encounter a region error involving an `FnMut` closure, we
display a specialized error message. However, we currently do not
tell the user which upvar was captured. This makes it difficult to
determine the cause of the error, especially when the closure is large.
This commit records marks constraints involving closure upvars
with `ConstraintCategory::ClosureUpvar`. When we decide to 'blame'
a `ConstraintCategory::Return`, we additionall store
the captured upvar if we found a `ConstraintCategory::ClosureUpvar` in
the path.
When generating an error message, we point to relevant spans if we have
closure upvar information available. We further customize the message if
an `async` closure is being returned, to make it clear that the captured
variable is being returned indirectly.
|
|
Explain move errors that occur due to method calls involving `self`
When calling a method that takes `self` (e.g. `vec.into_iter()`), the method receiver is moved out of. If the method receiver is used again, a move error will be emitted::
```rust
fn main() {
let a = vec![true];
a.into_iter();
a;
}
```
emits
```
error[E0382]: use of moved value: `a`
--> src/main.rs:4:5
|
2 | let a = vec![true];
| - move occurs because `a` has type `std::vec::Vec<bool>`, which does not implement the `Copy` trait
3 | a.into_iter();
| - value moved here
4 | a;
| ^ value used here after move
```
However, the error message doesn't make it clear that the move is caused by the call to `into_iter`.
This PR adds additional messages to move errors when the move is caused by using a value as the receiver of a `self` method::
```
error[E0382]: use of moved value: `a`
--> vec.rs:4:5
|
2 | let a = vec![true];
| - move occurs because `a` has type `std::vec::Vec<bool>`, which does not implement the `Copy` trait
3 | a.into_iter();
| ------------- value moved due to this method call
4 | a;
| ^ value used here after move
|
note: this function takes `self`, which moves the receiver
--> /home/aaron/repos/rust/src/libcore/iter/traits/collect.rs:239:5
|
239 | fn into_iter(self) -> Self::IntoIter;
```
TODO:
- [x] Add special handling for `FnOnce/FnMut/Fn` - we probably don't want to point at the unstable trait methods
- [x] Consider adding additional context for operations (e.g. `Shr::shr`) when the call was generated using the operator syntax (e.g. `a >> b`)
- [x] Consider pointing to the method parent (impl or trait block) in addition to the method itself.
|
|
Check for live drops in constants after drop elaboration
Resolves #66753.
This PR splits the MIR "optimization" pass series in two and introduces a query–`mir_drops_elaborated_and_const_checked`–that holds the result of the `post_borrowck_cleanup` analyses and checks for live drops. This query is invoked in `rustc_interface` for all items requiring const-checking, which means we now do `post_borrowck_cleanup` for items even if they are unused in the crate.
As a result, we are now more precise about when drops are live. This is because drop elaboration can e.g. eliminate drops of a local when all its fields are moved from. This does not mean we are doing value-based analysis on move paths, however; Storing a `Some(CustomDropImpl)` into a field of a local will still set the qualifs for that entire local.
r? @oli-obk
|
|
Clean up type alias impl trait implementation
- Removes special case for top-level impl trait
- Removes associated opaque types
- Forbid lifetime elision in let position impl trait. This is consistent with the behavior for inferred types.
- Handle lifetimes in type alias impl trait more uniformly with other parameters
cc #69323
cc #63063
Closes #57188
Closes #62988
Closes #69136
Closes #73061
|
|
The `Call` terminator only works with `FnDef` and `FnPtr` types.
It happened to work with `Self` so far because it was always
substituted with the real type before being used.
|
|
|
|
It does the same thing as `Deref` now
|
|
|
|
Emit an error when incompatible sanitizer are configured through command
line options. Previously the last one configured prevailed and others
were silently ignored.
Additionally use a set to represent configured sanitizers, making it
possible to enable multiple sanitizers at once. At least in principle,
since currently all of them are considered to be incompatible with
others.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Querify whether a type has structural equality (Take 2)
Alternative to #72177.
Unlike in #72177, this helper method works for all types, falling back to a query for `TyKind::Adt`s that determines whether the `{Partial,}StructuralEq` traits are implemented.
This is my preferred interface for this method. I think this is better than just documenting that the helper only works for ADTs. If others disagree, we can just merge #72177 with the fixes applied. This has already taken far too long.
|
|
|
|
|
|
fix caller_location intrinsic for Miri
Fixes https://github.com/rust-lang/rust/issues/73272
r? @oli-obk Cc @Aaron1011
|
|
r=oli-obk
Disable the `SimplifyArmIdentity` pass
This pass is buggy so I'm disabling it to fix a stable-to-beta
regression.
Related to #73223
|
|
This helper method works for all types, falling back to a query for
`TyKind::Adt`s to determine whether the implement the
`{Partial,}StructuralEq` traits.
|
|
This pass is buggy so I'm disabling it to fix a stable-to-beta
regression.
Related to #73223
|
|
|
|
|
|
|
|
of the evaluation
|
|
|
|
|
