Rollup of 5 pull requests
Successful merges:
- #126186 (Migrate `run-make/multiple-emits` to `rmake.rs`)
- #126236 (Delegation: fix ICE on recursive delegation)
- #126254 (Remove ignore-cross-compile directive from ui/macros/proc_macro)
- #126258 (Do not define opaque types when selecting impls)
- #126265 (interpret: ensure we check bool/char for validity when they are used in a cast)
r? `@ghost`
`@rustbot` modify labels: rollup
interpret: ensure we check bool/char for validity when they are used in a cast
In general, `Scalar::to_bits` is a bit dangerous as it bypasses all type information. We should usually prefer matching on the type and acting according to that. So I also refactored `unary_op` handling of integers to do that. The remaining `to_bits` uses are operations that just fundamentally don't care about the sign (and only work on integers).
invalid_char_cast.rs is the key new test, the others already passed before this PR.
r? `@oli-obk`
Remove ignore-cross-compile directive from ui/macros/proc_macro
All the other proc-macro tests don't have this, presumably this was forgotten when the restriction got lifted as it does test just fine
r? `@pietroalbini`
run-make: arm command wrappers with drop bombs
This PR is one in a series of cleanups to run-make tests and the run-make-support library.
### Summary
It's easy to forget to actually executed constructed command wrappers, e.g. `rustc().input("foo.rs")` but forget the `run()`, so to help catch these mistakes, we arm command wrappers with drop bombs on construction to force them to be executed by test code.
This PR also removes the `Deref`/`DerefMut` impl for our custom `Command` which derefs to `std::process::Command` because it can cause issues when trying to use a custom command:
```rs
htmldocck().arg().run()
```
fails to compile because the `arg()` is resolved to `std::process::Command::arg`, which returns `&mut std::process::Command` that doesn't have a `run()` command.
This PR also:
- Removes `env_var` on the `impl_common_helper` macro that was wrongly named and is a footgun (no users).
- Bumps the run-make-support library to version `0.1.0`.
- Adds a changelog to the support library.
### Details
Especially for command wrappers like `Rustc`, it's very easy to build up
a command invocation but forget to actually execute it, e.g. by using
`run()`. This commit adds "drop bombs" to command wrappers, which are
armed on command wrapper construction, and only defused if the command
is executed (through `run`, `run_fail`).
If the test writer forgets to execute the command, the drop bomb will
"explode" and panic with an error message. This is so that tests don't
silently pass with constructed-but-not-executed command wrappers.
This PR is best reviewed commit-by-commit.
try-job: x86_64-msvc
Handle single chars with `to_string()` for `single_char_add_str`
Add support for single chars / literals with `to_string()` call for `push_str()` and `insert_str()`.
changelog: [`single_char_add_str`]: handle single chars with `to_string()` call
Closes#12775
Normalize type aliases when checking significant drops.
fixes#10750
No test as I'm not sure exactly how the type alias caused a stack overflow here. Also removes an unnecessary `HashSet`.
changelog: none
- Update all command wrappers and command construction helpers with
`#[track_caller]` where suitable to help the drop bomb panic message.
- Remove `Deref`/`DerefMut` for `Command` because it was causing issues
with resolving to `std::process::Command` in a method call chain.
Rollup of 5 pull requests
Successful merges:
- #125913 (Spruce up the diagnostics of some early lints)
- #126234 (Delegation: fix ICE on late diagnostics)
- #126253 (Simplify assert matchers in `run-make-support`)
- #126257 (Rename `needs-matching-clang` to `needs-force-clang-based-tests`)
- #126259 (reachable computation: clarify comments around consts)
r? `@ghost`
`@rustbot` modify labels: rollup
People keep making fun of this signature for being so gnarly.
Associated type bounds lend it a much simpler scribbling.
ChangeOutputType can also come along for the ride.
This is incorrectly named (it's actually `env_clear`), and is itself
a gigantic footgun: removing `TMPDIR` on Unix and `TMP`/`TEMP` on
Windows basically wrecks anything that relies on `std::env::temp_dir`
from functioning correctly. For example, this includes rustc's codegen.
Rename `needs-matching-clang` to `needs-force-clang-based-tests`
This header is much more restrictive than its old name would suggest. As a result, most of the tests that use it don't actually run in any CI jobs.
Mitigation for #126180, though at some point we still need to go back fix the affected tests to actually run.
Spruce up the diagnostics of some early lints
Implement the various "*(note to myself) in a follow-up PR we should turn parts of this message into a subdiagnostic (help msg or even struct sugg)*" drive-by comments I left in #124417 during my review.
For context, before #124417, only a few early lints touched/decorated/customized their diagnostic because the former API made it a bit awkward. Likely because of that, things that should've been subdiagnostics were just crammed into the primary message. This PR rectifies this.
Only compute `specializes` query if (min)specialization is enabled in the crate of the specializing impl
Fixes (after backport) https://github.com/rust-lang/rust/issues/125197
### What
https://github.com/rust-lang/rust/pull/122791 makes it so that inductive cycles are no longer hard errors. That means that when we are testing, for example, whether these impls overlap:
```rust
impl PartialEq<Self> for AnyId {
fn eq(&self, _: &Self) -> bool {
todo!()
}
}
impl<T: Identifier> PartialEq<T> for AnyId {
fn eq(&self, _: &T) -> bool {
todo!()
}
}
```
...given...
```rust
pub trait Identifier: Display + 'static {}
impl<T> Identifier for T where T: PartialEq + Display + 'static {}
```
Then we try to see if the second impl holds given `T = AnyId`. That requires `AnyId: Identifier`, which requires that `AnyId: PartialEq`, which is satisfied by these two impl candidates... The `PartialEq<T>` impl is a cycle, and we used to winnow it when we used to treat inductive cycles as errors.
However, now that we don't winnow it, this means that we *now* try calling `candidate_should_be_dropped_in_favor_of`, which tries to check whether one of the impls specializes the other: the `specializes` query. In that query, we currently bail early if the impl is local.
However, in a foreign crate, we try to compute if the two impls specialize each other by doing trait solving. This may itself lead to the same situation where we call `specializes`, which will lead to a query cycle.
### How does this fix the problem
We now record whether specialization is enabled in foreign crates, and extend this early-return behavior to foreign impls too. This means that we can only encounter these cycles if we truly have a specializing impl from a crate with specialization enabled.
-----
r? `@oli-obk` or `@lcnr`
