CI: rfl: use rust-next temporary commit
Commit c95bbb59a9b22f9b838b15d28319185c1c884329 within rust-next contains some changes required to be compatible with upcoming arbitraty self types work. Roll RFL CI forward to the latest rust-next to include that work.
Related:
https://github.com/rust-lang/rust/pull/130225https://github.com/rust-lang/rust/issues/44874
r? ``@ojeda``
try-job: x86_64-rust-for-linux
do not remove `.cargo` directory
If vendoring isn't used bootstrap removes `.cargo` directory, which prevents developers from setting certain options in the `.cargo/config.toml` file. This was introduced in https://github.com/rust-lang/rust/pull/97513 (specifically in [this commit](345eb14f6c)). Also, since https://github.com/rust-lang/rust/pull/123942, vendoring is now possible even in git sources, which means we shouldn't remove `.cargo` directory in git sources anymore.
Add wasm32v1-none target (compiler-team/#791)
This is a preliminary implementation of the MCP discussed in [compiler-team#791](https://github.com/rust-lang/compiler-team/issues/791). It's not especially "major" but you know, process! Anyway it adds a new wasm32v1-none target which just pins down a set of wasm features. I think this is close to the consensus that emerged when discussing it on Zulip so I figured I'd sketch to see how hard it is. Turns out not very.
Compiletest: Custom differ
This adds support for a custom differ for compiletests. It’s purely visual and helps produce cleaner output when UI tests fail.
I’m using an environment variable for now since it’s experimental and I don’t want to drill the cli arguments all the way down. Also did a bit of general cleanup while I was at it.
This is how it looks [with debug info silenced](https://github.com/rust-lang/rust/pull/131182) (#131182)
`COMPILETEST_DIFF_TOOL="/usr/bin/env difft --color always --background light --display side-by-side" ./x test tests/ui/parser`
Refactor change detection for rustdoc and download-rustc
This pull request refactors the change detection logic in the build process by consolidating redundant code into a new helper method. The key changes include the removal of duplicate logic for checking changes in directories and the addition of a new method to handle this functionality.
Refactoring and code simplification:
* [`src/bootstrap/src/core/build_steps/tool.rs`](diffhunk://#diff-dc86e288bcf7b3ca3f8c127d3568fbafc785704883bc7fc336bd185910aed5daL588-R593): Removed redundant change detection logic and replaced it with a call to the new `check_for_changes` method.
* [`src/bootstrap/src/core/config/config.rs`](diffhunk://#diff-5f5330cfcdb0a89b85ac3547b761c3a45c2534a85c4aaae8fea88c711a7a65b2R2837-R2872): Added a new method `check_for_changes` to centralize the logic for detecting changes in specified directories since a given commit.
* [`src/bootstrap/src/core/config/config.rs`](diffhunk://#diff-5f5330cfcdb0a89b85ac3547b761c3a45c2534a85c4aaae8fea88c711a7a65b2L2728-R2740): Updated the existing change detection code to use the new `check_for_changes` method.
Cleanup:
* [`src/bootstrap/src/core/build_steps/tool.rs`](diffhunk://#diff-dc86e288bcf7b3ca3f8c127d3568fbafc785704883bc7fc336bd185910aed5daL13-R13): Removed the unused import `git` from the helpers module.
r? ``@AlbertLarsan68``
nightly feature tracking: get rid of the per-feature bool fields
The `struct Features` that tracks which features are enabled has a ton of public `bool`-typed fields that are basically caching the result of looking up the corresponding feature in `enabled_lang_features`. Having public fields with an invariant is not great, so at least they should be made private. However, it turns out caching these lookups is actually [not worth it](https://github.com/rust-lang/rust/pull/131321#issuecomment-2402068336), so this PR just entirely gets rid of these fields. (The alternative would be to make them private and have a method for each of them to expose them in a read-only way. Most of the diff of this PR would be the same in that case.)
r? `@nnethercote`
do not implement `Relate` for "boring" types
and update some macros while we're at it. This means we don't have to implement `TypeVisitable` for them.
r? `@compiler-errors`
Optimize `Rc<T>::default`
The missing piece of https://github.com/rust-lang/rust/pull/131460.
Also refactored `Arc<T>::default` by using a safe `NonNull::from(Box::leak(_))` to replace the unnecessarily unsafe call to `NonNull::new_unchecked(Box::into_raw(_))`. The remaining unsafety is coming from `[Rc|Arc]::from_inner`, which is safe from the construction of `[Rc|Arc]Inner`.
analyse: remove unused uncanonicalized field
This field is unused and was only relevant when actually printing proof trees. Right now this simply causes proof tree building to leak a bunch of inference vars 😁
r? ``@compiler-errors``
Run most `core::num` tests in const context too
This adds some infrastructure for something I was going to use in #131566, but it felt worthwhile enough on its own to merge/discuss separately.
Essentially, right now we tend to rely on UI tests to ensure that things work in const context, rather than just using library tests. This uses a few simple macro tricks to make it *relatively* painless to execute tests in both runtime and compile-time context. And this only applies to the numeric tests, and not anything else.
Recommended to review without whitespace in the diff.
cc `@RalfJung`
Represent `hir::TraitBoundModifiers` as distinct parts in HIR
Stop squashing distinct `polarity` and `constness` into a single `hir::TraitBoundModifier`.
This PR doesn't attempt to handle all the corner cases correctly, since the old code certainly did not either; but it should be much easier for, e.g., rustc devs working on diagnostics, or clippy devs, to actually handle constness and polarity correctly.
try-job: x86_64-gnu-stable
better default capacity for str::replace
Adds smarter capacity for str::replace in cases where we know that the output will be at least as long as the original string.
x86-32 float return for 'Rust' ABI: treat all float types consistently
This helps with https://github.com/rust-lang/rust/issues/131819: for our own ABI on x86-32, we want to *never* use the float registers. The previous logic only considered F32 and F64, but skipped F16 and F128. So I made the logic just apply to all float types.
try-job: i686-gnu
try-job: i686-gnu-nopt
Rollup of 8 pull requests
Successful merges:
- #125205 (Fixup Windows verbatim paths when used with the `include!` macro)
- #131049 (Validate args are correct for `UnevaluatedConst`, `ExistentialTraitRef`/`ExistentialProjection`)
- #131549 (Add a note for `?` on a `impl Future<Output = Result<..>>` in sync function)
- #131731 (add `TestFloatParse` to `tools.rs` for bootstrap)
- #131732 (Add doc(plugins), doc(passes), etc. to INVALID_DOC_ATTRIBUTES)
- #132006 (don't stage-off to previous compiler when CI rustc is available)
- #132022 (Move `cmp_in_dominator_order` out of graph dominator computation)
- #132033 (compiletest: Make `line_directive` return a `DirectiveLine`)
r? `@ghost`
`@rustbot` modify labels: rollup
compiletest: Make `line_directive` return a `DirectiveLine`
This reduces the need to juggle raw tuples, and opens up the possibility of moving more parts of directive parsing into `line_directive`.
In order to make the main change possible, this PR also (partly) separates the debugger-command parsing from the main directive parser. That cleanup removes support for `[rev]` in debugger commands, which is not used by any tests.
Move `cmp_in_dominator_order` out of graph dominator computation
Dominator-order information is only needed for coverage graphs, and is easy enough to collect by just traversing the graph again.
This avoids wasted work when computing graph dominators for any other purpose.
add `TestFloatParse` to `tools.rs` for bootstrap
add TestFloatParse to tools for bootstrap, I am not sure this is what the issue https://github.com/rust-lang/rust/issues/128012 discussion wants.
try-job: aarch64-apple
Add a note for `?` on a `impl Future<Output = Result<..>>` in sync function
It's confusing to `?` a future of a result in a sync function. We have a suggestion to `.await` it if we're in an async function, but not a sync function. Note that this is the case for sync functions, at least.
Let's be a bit more vague about a fix, since it's somewhat context dependent. For example, you could block on it, or you could make your function asynchronous. 🤷
Validate args are correct for `UnevaluatedConst`, `ExistentialTraitRef`/`ExistentialProjection`
For the `Existential*` ones, we have to do some adjustment to the args list to deal with the missing `Self` type, so we introduce a `debug_assert_existential_args_compatible` function to the interner as well.
Fixup Windows verbatim paths when used with the `include!` macro
On Windows, the following code can fail if the `OUT_DIR` environment variable is a [verbatim path](https://doc.rust-lang.org/std/path/enum.Prefix.html) (i.e. begins with `\\?\`):
```rust
include!(concat!(env!("OUT_DIR"), "/src/repro.rs"));
```
This is because verbatim paths treat `/` literally, as if it were just another character in the file name.
The good news is that the standard library already has code to fix this. We can simply use `components` to normalize the path so it works as intended.
