| Age | Commit message (Collapse) | Author | Lines |
|---|
|
|
|
|
|
Uplift `clippy::invalid_null_ptr_usage` lint as `invalid_null_arguments`
This PR aims at uplifting the `clippy::invalid_null_ptr_usage` lint into rustc, this is similar to the [`clippy::invalid_utf8_in_unchecked` uplift](https://github.com/rust-lang/rust/pull/111543) a few months ago, in the sense that those two lints lint on invalid parameter(s), here a null pointer where it is unexpected and UB to pass one.
*For context: GitHub Search reveals that just for `slice::from_raw_parts{_mut}` [~20 invalid usages](hhttps://github.com/search?q=lang%3Arust+%2Fslice%3A%3Afrom_raw_parts%28_mut%29%3F%5C%28ptr%3A%3Anull%2F+NOT+path%3A%2F%5Eclippy_lints%5C%2Fsrc%5C%2F%2F+NOT+path%3A%2F%5Erust%5C%2Fsrc%5C%2Ftools%5C%2Fclippy%5C%2Fclippy_lints%5C%2Fsrc%5C%2F%2F+NOT+path%3A%2F%5Esrc%5C%2Ftools%5C%2Fclippy%5C%2Fclippy_lints%5C%2Fsrc%5C%2F%2F&type=code) with `ptr::null` and an additional [4 invalid usages](https://github.com/search?q=lang%3Arust+%2Fslice%3A%3Afrom_raw_parts%5C%280%28%5C%29%7C+as%29%2F+NOT+path%3A%2F%5Eclippy_lints%5C%2Fsrc%5C%2F%2F+NOT+path%3A%2F%5Erust%5C%2Fsrc%5C%2Ftools%5C%2Fclippy%5C%2Fclippy_lints%5C%2Fsrc%5C%2F%2F+NOT+path%3A%2F%5Esrc%5C%2Ftools%5C%2Fclippy%5C%2Fclippy_lints%5C%2Fsrc%5C%2F%2F+NOT+path%3A%2F%5Eutils%5C%2Ftinystr%5C%2Fsrc%5C%2F%2F+NOT+path%3A%2F%5Eutils%5C%2Fzerovec%5C%2Fsrc%5C%2F%2F+NOT+path%3A%2F%5Eprovider%5C%2Fcore%5C%2Fsrc%5C%2F%2F&type=code) with `0 as *const ...`-ish casts.*
-----
## `invalid_null_arguments`
(deny-by-default)
The `invalid_null_arguments` lint checks for invalid usage of null pointers.
### Example
```rust
// Undefined behavior
unsafe { std::slice::from_raw_parts(ptr::null(), 1); }
```
Produces:
```
error: calling this function with a null pointer is Undefined Behavior, even if the result of the function is unused
--> $DIR/invalid_null_args.rs:21:23
|
LL | let _: &[usize] = std::slice::from_raw_parts(ptr::null_mut(), 0);
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^---------------^^^^
| |
| null pointer originates from here
|
= help: for more information, visit <https://doc.rust-lang.org/std/ptr/index.html> and <https://doc.rust-lang.org/reference/behavior-considered-undefined.html>
```
### Explanation
Calling methods whose safety invariants requires non-null pointer with a null pointer is undefined behavior.
-----
The lint use a list of functions to know which functions and arguments to checks, this could be improved in the future with a rustc attribute, or maybe even with a `#[diagnostic]` attribute.
This PR also includes some small refactoring to avoid some ambiguities in naming, those can be done in another PR is desired.
`@rustbot` label: +I-lang-nominated
r? compiler
|
|
Remove attribute `#[rustc_error]`
It was an ancient way to write `check-pass` tests, but now it's no longer necessary (except for the `delayed_bug_from_inside_query` flavor, which is retained).
|
|
bootstrap: Avoid cloning `change-id` list
Inspired by [recent discussion](https://rust-lang.zulipchat.com/#narrow/channel/326414-t-infra.2Fbootstrap/topic/Collecting.20some.20Real.20Configs.20for.20testing/near/507845657) on the bootstrap `change-id` field, I took a look at the code and found this little optimization. It does not change behavior.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
This was removed, likely by mistake, during a refactor.
|
|
|
|
Fix compile errors of all the examples
|
|
compiletest: Support matching diagnostics on lines below
Using `//~vvv ERROR`.
This is not needed often, but it's easy to support, and it allows to eliminate a class of `error-pattern`s that cannot be eliminated in any other way.
See the diff for the examples of such patterns coming from parser.
Some of them can be matched by `//~ ERROR` or `//~^ ERROR` as well (when the final newline is allowed), but it changes the shape of reported spans, so I chose to keep the spans by using `//~v ERROR`.
|
|
mention that known-bug test directive takes arguments
|
|
|
|
- Fix #18782
|
|
Reject `{true,false}` as revision names
Because they would imply `--cfg={true,false}` otherwise, and the test writer has to use `cfg(r#true)` and `cfg(r#false)` in the test.
Closes #138663.
|
|
|
|
|
|
Bump to 1.88
https://forge.rust-lang.org/release/process.html#bump-the-stable-version-number-friday-the-week-before
r? ghost
|
|
Bump boostrap `cc` to 1.2.17 and `cmake` to 0.1.54
The `cc` version in `bootstrap` was reverted down to 1.1.22 in https://github.com/rust-lang/rust/pull/137460 (previously at 1.2.0). The offending issue has since then been resolved in https://github.com/rust-lang/cc-rs/pull/1413, and a new version of `cc` has been released in https://github.com/rust-lang/cc-rs/pull/1435, so let's try to update the version again.
See [the `cc-rs` changelog](https://github.com/rust-lang/cc-rs/blob/d9dd20e376368c7535f6ef89b809098f5f203c1a/CHANGELOG.md) and [the `cmake-rs` changelog](https://github.com/rust-lang/cmake-rs/blob/fd56c5a6b4ecda8815c863eb5b12d7b3f0391197/CHANGELOG.md) for details on what has changed here.
r? jieyouxu who tried this last in https://github.com/rust-lang/rust/pull/137022.
`@rustbot` label T-bootstrap
try-job: *apple*
|
|
|
|
Miri subtree update
r? `@ghost`
|
|
|
|
|
|
|
|
machine clock: make 'monotonic' explicit
|
|
|
|
bootstrap: update `test_find` test
`cc::Build::get_archiver` is noisy on the `arm-linux-androideabi` target and constantly printing `llvm-ar --version` output during bootstrap tests on all platforms.
|
|
Update target maintainers for thumb targets to reflect new REWG Arm team name
Closes #139027
The name of the team responsible for these targets has changed as the team was merged with other Arm-related teams (see https://github.com/rust-embedded/wg/pull/818). The link gives an up-to-date list of github usernames that can be pinged, whereas the old email address is not very actively maintained or tracked.
|
|
r=Kobzol
feat(config): Add ChangeId enum for suppressing warnings
closes: #138925
|
|
rustc_resolve: fix instability in lib.rmeta contents
rust-lang/rust@23032f31c91f2 accidentally introduced some nondeterminism in the ordering of lib.rmeta files, which we caught in our bazel-based builds only recently due to being further behind than normal. In my testing, this fixes the issue.
|
|
Greatly simplify doctest parsing and information extraction
The original process was pretty terrible, as it tried to extract information such as attributes by performing matches over tokens like `#!`, which doesn't work very well considering you can have `# ! [`, which is valid.
Also, it now does it in one pass: if the parser is happy, then we try to extract information, otherwise we return early.
r? `@fmease`
|
|
|
|
|
|
|
|
feat: Allow crate authors to control completion of their things
|
|
`cc::Build::get_archiver` is noisy on the `arm-linux-androideabi` target and
constantly printing `llvm-ar --version` output during bootstrap tests on all platforms.
Signed-off-by: onur-ozkan <work@onurozkan.dev>
|
|
This PR makes a fairly large version update to CMake and cc, so it is
likely that LLVM is built differently.
|
|
Similarly to what was previously done for the `llvm` step.
|
|
compiler-rt's CMake setup seems to have special logic for Apple
platforms that works poorly when this is not set.
|
|
|
|
To avoid a panic in cmake-rs that was introduced in:
https://github.com/rust-lang/cmake-rs/pull/158
|
|
|
|
description references
|
|
Where a fundamental type applied twice wasn't considered local.
|
