Remove `MaybeForgetReturn` suggestion
#115196 implemented a suggestion to add a missing `return` when there is an ambiguity error, when that ambiguity error could be constrained by the return type of the function.
I initially reviewed it and thought it could be useful; however, looking back at that code now, I feel like it's a bit too much of a hack to be worth keeping around in typeck, especially given how rare it's expected to fire in practice. This is especially true because it depends on `StashKey::MaybeForgetReturn`, which is only stashed when we have *Sized* obligation ambiguity errors. Let's remove it for now.
I'd like to note that it's basically impossible to get this suggestion to apply in its current state except for what I'd consider somewhat artificial examples, involving no generic trait bounds. For example, it's not triggered for:
```rust
struct W<T>(T);
fn bar<T: Default>() -> W<T> { todo!() }
fn foo() -> W<i32> {
if true {
bar();
}
W(0)
}
```
Nor is it triggered for:
```
fn foo() -> i32 {
if true {
Default::default();
}
0
}
```
It's basically only triggered iff there's only one ambiguity error on the type, which is `Sized`.
Generally, suggesting something that affects control flow is a pretty dramatic suggestion; therefore, both the accuracy and precision of this diagnostic should be pretty high.
One other, somewhat unrelated observation is that this might be using stashed diagnostics incorrectly (or at least unnecessarily). Stashed diagnostics are used when error detection is fragmented over several major stages of the compiler, like a parse or resolver error which later can be recovered in typeck. However, this one is a bit different since it is fully handled within typeck -- perhaps that suggests that if this were to be reimplemented, it wouldn't need to be so complicated of an implementation.
Document workings of successors more clearly
This is an attempt to fix#135087 together with https://github.com/rust-lang/rust/pull/135886, but I am not sure if I've succeeded in adding much clarity here, so don't be shy with your comments.
Implement `&pin const self` and `&pin mut self` sugars
This PR implements part of #130494.
It introduces the sugars `&pin const self` and `&pin mut self` for `self: Pin<&Self>` and `self: Pin<&mut Self>`.
handle forced compiler and revert #137476
Fixes https://github.com/rust-lang/rust/issues/138004
I would appreciate it if we could measure CI pipelines with the current changes to see if this reduces recent CI overhead. cc `@rust-lang/infra`
try-job: dist-powerpc64le-linux
Inline `FnOnce`/`FnMut`/`Fn` shims once again
This PR fixes the argument checking for `extern "rust-call"` ABI functions with a spread arg, which do no expect their arguments to be exploded from a tuple like closures do.
Secondly, it removes the hack that prevented them from being inlined. This results in more work done by the compiler, but it does end up allowing us to inline functions we didn't before.
Fixes#137901
Rollup of 20 pull requests
Successful merges:
- #134063 (dec2flt: Clean up float parsing modules)
- #136581 (Retire the legacy `Makefile`-based `run-make` test infra)
- #136662 (Count char width at most once in `Formatter::pad`)
- #136764 (Make `ptr_cast_add_auto_to_object` lint into hard error)
- #136798 (Added documentation for flushing per #74348)
- #136865 (Perform deeper compiletest path normalization for `$TEST_BUILD_DIR` to account for compare-mode/debugger cases, and normalize long type file filename hashes)
- #136975 (Look for `python3` first on MacOS, not `py`)
- #136977 (Upload Datadog metrics with citool)
- #137240 (Slightly reformat `std::fs::remove_dir_all` error docs)
- #137298 (Check signature WF when lowering MIR body)
- #137463 ([illumos] attempt to use posix_spawn to spawn processes)
- #137477 (uefi: Add Service Binding Protocol abstraction)
- #137569 (Stabilize `string_extend_from_within`)
- #137633 (Only use implied bounds hack if bevy, and use deeply normalize in implied bounds hack)
- #137679 (Various coretests improvements)
- #137723 (Make `rust.description` more general-purpose and pass `CFG_VER_DESCRIPTION`)
- #137728 (Remove unsizing coercions for tuples)
- #137731 (Resume one waiter at once in deadlock handler)
- #137875 (mir_build: Integrate "simplification" steps into match-pair-tree creation)
- #138028 (compiler: add `ExternAbi::is_rustic_abi`)
r? `@ghost`
`@rustbot` modify labels: rollup
compiler: add `ExternAbi::is_rustic_abi`
Various parts of the compiler were hand-rolling this extremely simple check that is nonetheless easy to get wrong as the compiler evolves over time. Discourage them from being so "original" again by replacing it with a single implementation on the type that represents these ABIs. This simplifies a surprising amount of code as a result.
Also fixes#132981, an ICE that emerged due to other checks being made stricter.
mir_build: Integrate "simplification" steps into match-pair-tree creation
The “simplification” step helps to prepare THIR patterns for lowering into MIR, and originally dates back to the earliest days of MIR in the compiler.
Over time, various intermediate data structures have been introduced (e.g. `MatchPair`, later renamed to `MatchPairTree`) that reduce the need for a separate simplification step, because some of the necessary simplifications can be built into the construction of those intermediate structures instead. This PR continues that process to its logical conclusion and removes the simplification step entirely, by integrating its remaining responsibilities into match-pair-tree creation: flattening “irrefutable” nodes, collecting bindings/ascriptions in flat lists, and sorting or-patterns after other subpatterns.
This has a few immediate benefits:
- We can remove `TestCase::Irrefutable`, which was not allowed to exist after simplification, and was much larger than other test-case variants.
- We can make `MatchPairTree::place` non-optional, because only irrefutable nodes could fail to have a place.
In the future, this should also help with some ideas I have for simplifying how `AscribeUserType` and `ExpandedConstant` nodes are handled, by representing them as side-data keyed by THIR pattern ID, so that they are no longer their own kinds of THIR pattern node.
Resume one waiter at once in deadlock handler
When multiple query loop errors occur in the code, only one waiter should be resumed at a time to avoid waking up multiple waiters at the same time and causing deadlock due to thread grabbing.
This fixes the UI failures in #132051
cc `@Zoxc` `@cjgillot` `@nnethercote` `@bjorn3` `@Kobzol`
Zulip discussion [here](https://rust-lang.zulipchat.com/#narrow/channel/187679-t-compiler.2Fwg-parallel-rustc/topic/Deadlocks.20and.20Rayon)
Edit: We can't reproduce these bugs with the existing test suits, so we keep them until we merge #132051
UPDATES #129912
UPDATES #120757
UPDATES #129911
Various coretests improvements
The first commit is not yet strictly necessary as directly testing libcore works though useless work, but will be necessary once https://github.com/rust-lang/rust/pull/136642 migrates the liballoc tests into a separate package. The second commit fixes https://github.com/rust-lang/rust/issues/137478 and ensures that coretests actually gets tested on all CI job. The third commit fixes an error that didn't get caught because coretests doesn't run on the wasm32 CI job.
uefi: Add Service Binding Protocol abstraction
- Some UEFI protocols such as TCP4, TCP6, UDP4, UDP6, etc are managed by service binding protocol.
- A new instance of such protocols is created and destroyed using the corresponding service binding protocol.
- This PR adds abstractions to make using such protocols simpler using Rust Drop trait.
- The reason to add these abstractions in a seperate PR from TCP4 Protocol is to make review easier.
[EFI_SERVICE_BINDING_PROTCOL](https://uefi.org/specs/UEFI/2.11/11_Protocols_UEFI_Driver_Model.html#efi-service-binding-protocol)
cc ````@nicholasbishop````
[illumos] attempt to use posix_spawn to spawn processes
illumos has `posix_spawn`, and the very newest versions also have `_addchdir`, so use that. POSIX standardized this function so I also added a weak symbol lookup for the non `_np` version. (illumos has both.)
This probably also works on Solaris, but I don't have access to an installation to validate this so I decided to focus on illumos instead.
This is a nice ~4x performance improvement for process creation. My go-to as usual is nextest against the clap repo, which acts as a stress test for process creation -- with [this commit]:
```console
$ cargo nextest run -E 'not test(ui_tests) and not test(example_tests)'
before: Summary [ 1.747s] 879 tests run: 879 passed, 2 skipped
after: Summary [ 0.445s] 879 tests run: 879 passed, 2 skipped
```
[this commit]: fde45f9aea
Slightly reformat `std::fs::remove_dir_all` error docs
To make the error cases easier to spot on a quick glance, as I've been bitten by this a couple of times already 💀
cc #137230.
Look for `python3` first on MacOS, not `py`
`py` is not installed by default *and* trying to run it results in a popup asking if you want to install it. `python3` is installed by default.
This hopefully should not be too disruptive to people on Windows, since they should be going through `x.ps1` instead anyway. Just in case, I've added a check for Cygwin and Msys (i'm not sure how else you'd get a bash shell on windows).
I've tested this on macOS, WSL, "git bash" (which appears to be MSYS), and MSVC with `bash -c ./x` (which runs in WSL, but differently than interactive bash 🙃). I haven't tested Cygwin, which is unsupported anyway, but the code tries to make it work.
Perform deeper compiletest path normalization for `$TEST_BUILD_DIR` to account for compare-mode/debugger cases, and normalize long type file filename hashes
Fixes#136510.
### Summary
- Whereas previously `$TEST_BUILD_DIR` is a normalization of `/path/to/build/test/<test_suite_name>/`, we now more deeply normalize. `$TEST_BUILD_DIR` now becomes a normalization of `/path/to/build/test/<test_suite_name>/<subdirs>/$name.$revision.$compare_mode.$debugger/` to normalize away path name differences when `--compare-mode` and/or `--debugger` are specified.
- We also centralize the normalization of long type name hashes
cf. https://github.com/rust-lang/rust/pull/136328#discussion_r1936760908.
### Review advice
- Best reviewed commit-by-commit.
- Split into 3 commits:
- **Commit 1**: compiletest changes to have `$TEST_BUILD_DIR` more deeply normalize.
- **Commit 2**: remove per-test hacks for long type path hash normalizations, and rebless tests *specifically* affected by that.
- **Commit 3**: rebless other tests that were changed as a side-effect of deeper `$TEST_BUILD_DIR` normalizations.
**Commit 2** is created via first finding tests that try to perform long type file hash normalizations on an ad hoc, per-test basis:
```
rg --no-ignore -l --no-ignore -F -e "long-type" tests/ui/**/*.rs
```
<details>
<summary>Tests with ad hoc long-type hash normalizations</summary>
```
tests/ui/type_length_limit.rs
tests/ui/traits/on_unimplemented_long_types.rs
tests/ui/regions/issue-102374.rs
tests/ui/recursion/recursion.rs
tests/ui/recursion/issue-83150.rs
tests/ui/recursion/issue-23122-2.rs
tests/ui/methods/inherent-bound-in-probe.rs
tests/ui/issues/issue-67552.rs
tests/ui/issues/issue-37311-type-length-limit/issue-37311.rs
tests/ui/issues/issue-20413.rs
tests/ui/issues/issue-8727.rs
tests/ui/infinite/infinite-instantiation.rs
tests/ui/infinite/infinite-instantiation-struct-tail-ice-114484.rs
tests/ui/higher-ranked/trait-bounds/hrtb-doesnt-borrow-self-1.rs
tests/ui/higher-ranked/trait-bounds/hrtb-doesnt-borrow-self-2.rs
tests/ui/higher-ranked/trait-bounds/hang-on-deeply-nested-dyn.rs
tests/ui/error-codes/E0275.rs
tests/ui/diagnostic-width/secondary-label-with-long-type.rs
tests/ui/diagnostic-width/long-e0277.rs
tests/ui/diagnostic-width/non-copy-type-moved.rs
tests/ui/diagnostic-width/long-E0308.rs
tests/ui/diagnostic-width/E0271.rs
tests/ui/diagnostic-width/binop.rs
```
</details>
These ad hoc normalizations were removed, and they are reblessed.
r? `````@lqd`````
Make `ptr_cast_add_auto_to_object` lint into hard error
In Rust 1.81, we added a FCW lint (including linting in dependencies) against pointer casts that add an auto trait to dyn bounds. This was part of work making casts of pointers involving trait objects stricter, and was part of the work needed to restabilize trait upcasting.
We considered just making this a hard error, but opted against it at that time due to breakage found by crater. This breakage was mostly due to the `anymap` crate which has been a persistent problem for us.
It's now a year later, and the fact that this is not yet a hard error is giving us pause about stabilizing arbitrary self types and `derive(CoercePointee)`. So let's see about making a hard error of this.
r? ghost
cc ```@adetaylor``` ```@Darksonn``` ```@BoxyUwU``` ```@RalfJung``` ```@compiler-errors``` ```@oli-obk``` ```@WaffleLapkin```
Related:
- https://github.com/rust-lang/rust/pull/135881
- https://github.com/rust-lang/rust/issues/136702
- https://github.com/rust-lang/rust/pull/136776
Tracking:
- https://github.com/rust-lang/rust/issues/127323
- https://github.com/rust-lang/rust/issues/44874
- https://github.com/rust-lang/rust/issues/123430
Count char width at most once in `Formatter::pad`
When both width and precision flags are specified, then `Formatter::pad` counts the character width twice. Instead, record the character width when truncating it to the precision, so it does not need to be recomputed. Simplify control flow so the cases are more clear.
Related:
- 6c9e708f4be (`fmt::Formatter::pad`: don't call chars().count() more than one time, 2021-09-01): Reduce counting chars from thrice to twice in worst case
- ede39aeb331 (feat: reinterpret `precision` field for strings, 2016-06-29): Change meaning of precision for strings
- b820748ff5a (Implement formatting arguments for strings and integers, 2013-08-10): Implement `Formatter::pad`
Retire the legacy `Makefile`-based `run-make` test infra
The final piece of [porting run-make tests to use Rust #121876](https://github.com/rust-lang/rust/issues/121876).
Closes#121876.
Closes#40713.
Closes#81791 (no longer using `wc`).
Closes#56475 (no longer a problem in current form of that test; we don't ignore the test on `aarch64-unknown-linux-gnu`).
### Summary
This PR removes the legacy `Makefile`-based `run-make` test infra which has served us well over the years. The legacy infra is no longer needed since we ported all of `Makefile`-based `run-make` tests to the new `rmake.rs` infra.
Additionally, this PR:
- Removes `tests/run-make/tools.mk` since no more `Makefile`-based tests remain.
- Updates `tests/run-make/README.md` and rustc-dev-guide docs to remove mention about `Makefile`-based `run-make` tests
- Update test suite requirements in rustc-dev-guide on Windows to no longer need MSYS2 (they should also now run successfully on native Windows MSVC).
- Update `triagebot.toml` to stop backlinking to #121876.
**Thanks to everyone who helped in this effort to modernize the `run-make` test infra and test suite!**
r? bootstrap
dec2flt: Clean up float parsing modules
This is the first portion of my work adding support for parsing and printing `f16`. Changes in `float.rs` replace the magic constants with expressions and add some use of generics to better support the new float types. Everything else is related to documentation or naming; there are no functional changes in this PR.
This can be reviewed by commit.
Use `std::mem::{size_of, size_of_val, align_of, align_of_val}` from the
prelude instead of importing or qualifying them.
These functions were added to all preludes in Rust 1.80.
mgca: Lower all const paths as `ConstArgKind::Path`
When `#![feature(min_generic_const_args)]` is enabled, we now lower all
const paths in generic arg position to `hir::ConstArgKind::Path`. We
then lower assoc const paths to `ty::ConstKind::Unevaluated` since we
can no longer use the anon const expression lowering machinery. In the
process of implementing this, I factored out `hir_ty_lowering` code that
is now shared between lowering assoc types and assoc consts.
This PR also introduces a `#[type_const]` attribute for trait assoc
consts that are allowed as const args. However, we still need to
implement code to check that assoc const definitions satisfy
`#[type_const]` if present (basically is it a const path or a
monomorphic anon const).
r? `@BoxyUwU`
Support raw-dylib link kind on ELF
raw-dylib is a link kind that allows rustc to link against a library without having any library files present.
This currently only exists on Windows. rustc will take all the symbols from raw-dylib link blocks and put them in an import library, where they can then be resolved by the linker.
While import libraries don't exist on ELF, it would still be convenient to have this same functionality. Not having the libraries present at build-time can be convenient for several reasons, especially cross-compilation. With raw-dylib, code linking against a library can be cross-compiled without needing to have these libraries available on the build machine. If the libc crate makes use of this, it would allow cross-compilation without having any libc available on the build machine. This is not yet possible with this implementation, at least against libc's like glibc that use symbol versioning. The raw-dylib kind could be extended with support for symbol versioning in the future.
This implementation is very experimental and I have not tested it very well. I have tested it for a toy example and the lz4-sys crate, where it was able to successfully link a binary despite not having a corresponding library at build-time.
I was inspired by Björn's comments in https://internals.rust-lang.org/t/bundle-zig-cc-in-rustup-by-default/22096/27
Tracking issue: #135694
r? bjorn3
try-job: aarch64-apple
try-job: x86_64-msvc-1
try-job: x86_64-msvc-2
try-job: test-various
Rollup of 12 pull requests
Successful merges:
- #135767 (Future incompatibility warning `unsupported_fn_ptr_calling_conventions`: Also warn in dependencies)
- #137852 (Remove layouting dead code for non-array SIMD types.)
- #137863 (Fix pretty printing of unsafe binders)
- #137882 (do not build additional stage on compiler paths)
- #137894 (Revert "store ScalarPair via memset when one side is undef and the other side can be memset")
- #137902 (Make `ast::TokenKind` more like `lexer::TokenKind`)
- #137921 (Subtree update of `rust-analyzer`)
- #137922 (A few cleanups after the removal of `cfg(not(parallel))`)
- #137939 (fix order on shl impl)
- #137946 (Fix docker run-local docs)
- #137955 (Always allow rustdoc-json tests to contain long lines)
- #137958 (triagebot.toml: Don't label `test/rustdoc-json` as A-rustdoc-search)
r? `@ghost`
`@rustbot` modify labels: rollup
