Document Poisoning in `LazyCell` and `LazyLock`
Currently, there is no documentation of poisoning behavior in either `LazyCell` or `LazyLock`, even though both of them can be observed as poisoned by users.
`LazyCell` [plagyround example](https://play.rust-lang.org/?version=stable&mode=debug&edition=2024&gist=9cf38b8dc56db100848f54085c2c697d)
`LazyLock` [playground example](https://play.rust-lang.org/?version=stable&mode=debug&edition=2024&gist=f1cd6f9fe16636e347ebb695a0ce30c0)
# Open Questions
- [x] Is it worth making the implementation of `LazyLock` more complicated to ensure that the the panic message is `"LazyLock instance has previously been poisoned"` instead of `"Once instance has previously been poisoned"`? See the `LazyLock` playground link above for more context.
- [x] Does it make sense to move `LazyLock` into the `poison` module? It is certainly a poison-able type, but at the same time it is slightly different from the 4 other types currently in the `poison` module in that it is unrecoverable. I think this is more of a libs-api question.
``@rustbot`` label +T-libs-api
Please let me know if these open questions deserve a separate issue / PR!
Use `as_array` in PartialEq for arrays
Now that `as_array` exists we might as well use it here, since it's a bit more convenient than getting the correct type out of `try_into`.
Remove unnecessary `rust_` prefixes
part of https://github.com/rust-lang/rust/issues/116005
Honestly, not sure if this can affect linking somehow, also I didn't touched things like `__rust_panic_cleanup` and `__rust_start_panic` which very likely will break something, so just small cleanup here
also didn't changed `rust_panic_without_hook` because it was renamed here https://github.com/rust-lang/rust/pull/144852
r? libs
`AlignmentEnum` should just be `repr(usize)` now
These used to use specific sizes because they were compiled on all widths. But now that the types themselves are `#[cfg]`'d, we can save some conversions by having it always be `repr(usize)`.
Update safety comment for new_unchecked in niche_types
Change the safety comment on `new_unchecked` to mention the valid range instead of 0. I noticed this while working on https://github.com/model-checking/verify-rust-std
Improve formatting of doc code blocks
We don't currently apply automatic formatting to doc comment code blocks. As a
result, it has built up various idiosyncracies, which make such automatic
formatting difficult. Some of those idiosyncracies also make things harder for
human readers or other tools.
This PR makes a few improvements to doc code formatting, in the hopes of making
future automatic formatting easier, as well as in many cases providing net
readability improvements.
I would suggest reading each commit separately, as each commit contains one
class of changes.
Document guarantees of poisoning
This mostly documents the current behavior of `Mutex` and `RwLock` (rust-lang/rust#143471) as imperfect. It's unlikely that the situation improves significantly in the future, and even if it does, the rules will probably be more complicated than "poisoning is completely reliable", so this is a conservative guarantee.
We also explicitly specify that `OnceLock` never poisons, even though it has an API similar to mutexes.
Fixesrust-lang/rust#143471 by improving documentation.
r? ``@Amanieu``
Simplify library dependencies on `compiler-builtins`
The three panic-related library crates need to have access to `core`, and `compiler-builtins` needs to be in the crate graph. Rather than specifying both dependencies, switch these crates to use `rustc-std-workspace-core` which already does this.
This means there is now a single place that the `compiler-builtins` dependency needs to get configured, for everything other than `alloc` and `std`.
The second commit removes `compiler-builtins` from `std` (more details in the message).
`compiler-builtins` is already in the crate graph via `alloc`, and all
features related to `compiler-builtins` goes through `alloc`. There
isn't any reason that `std` needs this direct dependency, so remove it.
The three panic-related library crates need to have access to `core`,
and `compiler-builtins` needs to be in the crate graph. Rather than
specifying both dependencies, switch these crates to use
`rustc-std-workspace-core` which already does this.
This means there is now a single place that the `compiler-builtins`
dependency needs to get configured, for everything other than `alloc`
and `std`.
`compiler_builtins` shouldn't be called directly. Change the `PartialEq`
implementation for `DevicePathNode` to use slice equality instead, which
will call `memcmp`/`bcmp` via the intrinsic.
Make `libtest::ERROR_EXIT_CODE` const public to not redefine it in rustdoc
I think it's better to make this constant public so it can be used by crates using `libtest` as dependency.
As a side-note, I will update https://github.com/rust-lang/rust/pull/143900 to make use of this constant once this is current PR is merged.
Fix Ord, Eq and Hash implementation of panic::Location
Fixes https://github.com/rust-lang/rust/issues/144486.
Now properly compares/hashes the filename rather than the pointer to the string.
Fix typo in `DropGuard` doc
Follows-up rust-lang/rust#144236 (I happened to see the typo yesterday but didn’t think it should delay the PR’s merge so I kept quiet, sorryyyyy).
thread name in stack overflow message
Fixesrust-lang/rust#144481, which is caused by the thread name not being initialised yet when setting up the stack overflow information. Unfortunately, the stack overflow UI test did not test for the correct thread name being present, and testing this separately didn't occur to me when writing https://github.com/rust-lang/rust/pull/140628.
This PR contains the smallest possible fix I could think of: passing the thread name explicitly to the platform thread creation function. In the future I'd very much like to explore some possibilities around merging the thread packet and thread handle into one structure and using that in the platform code instead – but that's best left for another PR.
This PR also amends the stack overflow test to check for thread names, so we don't run into this again.
``@rustbot`` label +beta-nominated
Document why `Range*<&T> as RangeBounds<T>` impls are not `T: ?Sized`, and give an alternative.
`Range*<&T> as RangeBounds<T>` impls have been tried to be relaxed to `T: ?Sized` at least twice:
* https://github.com/rust-lang/rust/pull/61584
* https://github.com/rust-lang/rust/pull/64327
I also was just about to make another PR to do it again until I `./x.py test library/alloc` and rediscovered the type inference regression, then searched around and found the previous PRs. Hence this PR instead so hopefully that doesn't keep happening 😛.
These impls cannot be relaxed for two reasons:
1. Type inference regressions: See ``@SimonSapin's`` explanation from a previous PR: https://github.com/rust-lang/rust/pull/61584#issuecomment-499601046
2. It's a breaking change: `impl RangeBounds<MyUnsizedType> for std::ops::Range<&MyUnsizedType>` is allowed after the coherence rebalance ([playground link](https://play.rust-lang.org/?version=stable&mode=debug&edition=2024&gist=f704a6fe53bfc33e55b2fc246d895ec2)), and relaxing these impls would conflict with that downstream impl.
This PR adds doc-comments explaining that not having `T: ?Sized` is intentional[^1], and gives an explicit alternative: `(Bound<&T>, Bound<&T>)`.
Technically, the impls for the unstable new `std::range` types could be relaxed, as they are still unstable so the change would not be breaking, but having them be different in this regard seems worse (and the non-iterable `RangeTo/RangeToInclusive` range types are shared between the "new" and "old" so cannot be changed anyway), and then the type inference regression would pop up in whatever edition the new range types stabilize in.
The "see \<link\> for discussion of those issues" is intentionally left as a non-doc comment just for whoever may try to relax these impls again in the future, but if it is preferred to have the link in the docs I can add that.
Closes https://github.com/rust-lang/rust/issues/107196 (as wontfix)
CC https://github.com/rust-lang/rust/issues/64027
[^1]: "intentional" is maybe a bit of strong wording, should it instead say something like "was stabilized without it and it would be breaking to change it now"?
Implementation: `#[feature(sync_nonpoison)]`, `#[feature(nonpoison_mutex)]`
Continuation of https://github.com/rust-lang/rust/pull/134663
Tracking Issue: https://github.com/rust-lang/rust/issues/134645
This PR implements a new `sync/nonpoison` module, as well as the `nonpoison` variant of the `Mutex` lock.
There are 2 main changes here, the first is the new `nonpoison::mutex` module, and the second is the `mutex` integration tests.
For the `nonpoison::mutex` module, I did my best to align it with the current state of the `poison::mutex` module. This means that several unstable features (`mapped_lock_guards`, `lock_value_accessors`, and `mutex_data_ptr`) are also in the new `nonpoison::mutex` module, under their respective feature gates. Everything else in that file is under the correct feature gate (`#[unstable(feature = "nonpoison_mutex", issue = "134645")]`).
Everything in the `nonpoison::mutex` file is essentially identical in spirit, as we are simply removing the error case from the original `poison::mutex`.
The second big change is in the integration tests. I created a macro called that allows us to duplicate tests that are "generic" over the different mutex types, in that the poison mutex is always `unwrap`ped.
~~I think that there is an argument against doing this, as it can make the tests a bit harder to understand (and language server capabilities are weaker within macros), but I think the benefit of code deduplication here is worth it. Note that it is definitely possible to generalize this (with a few tweaks) to testing the other `nonpoison` locks when they eventually get implemented, but I'll leave that for a later discussion.~~
Adds tests for the `nonpoison::Mutex` variant by using a macro to
duplicate the existing `poison` tests.
Note that all of the tests here are adapted from the existing `poison`
tests.
Adds the equivalent `nonpoison` types to the `poison::mutex` module.
These types and implementations are gated under the `nonpoison_mutex`
feature gate.
Also blesses the ui tests that now have a name conflicts (because these
types no longer have unique names). The full path distinguishes the
different types.
Co-authored-by: Aandreba <aandreba@gmail.com>
Co-authored-by: Trevor Gross <tmgross@umich.edu>
