Stop `llvm.expect`ing assert terminators
We're putting `llvm.expect` calls before the <https://doc.rust-lang.org/nightly/nightly-rustc/rustc_middle/mir/enum.TerminatorKind.html#variant.Assert> terminators.
But we don't need them. One of the arms is always to a panic function that's marked `#[cold]`, which is `cold` <https://llvm.org/docs/LangRef.html#function-attributes> in LLVM, which
> When computing edge weights, basic blocks post-dominated by a cold function call are also considered to be cold; and, thus, given low weight.
So even without us emitting the extra intrinsic call, LLVM knows what to expect for the `br`. Thus we can save the (small) effort of emitting it and then LLVM optimizing it out.
r? compiler
Rollup of 10 pull requests
Successful merges:
- #124501 (add support to override lldb binary path for ./x test)
- #124573 (add a reference link to the comment of the "cc" and "cmake".)
- #124663 (Enable reusing CI Docker cache when running CI images locally)
- #124690 (Only consider ambiguous goals when finding best obligation for ambiguities)
- #124713 (Update Cargo specific diagnostics in check-cfg)
- #124717 (Implement `do_not_recommend` in the new solver)
- #124718 (Record impl args in the proof tree)
- #124720 (interpret: Drop: always evaluate place)
- #124721 (library/std: Fix build for NetBSD targets with 32-bit `c_long`)
- #124723 (Use correct Hermit links in The `rustc` Book)
r? `@ghost`
`@rustbot` modify labels: rollup
Use correct Hermit links in The `rustc` Book
As is, this documentation links to the old Hermit organization, `hermitcore`, which isn't used anymore.
I've updated the links to point to the new organization. This PR also changes the incorrect "rusty loader" link to point to the new `hermit-rs-template` repo.
(fixes#124722)
library/std: Fix build for NetBSD targets with 32-bit `c_long`
This fixes building `std` for targets like `mipsel-unknown-netbsd`.
If `c_long` is an `i64`, this conversion works with `Into`. But if it's an `i32`, this failed to convert a `u32` to an `i32`.
Implement `do_not_recommend` in the new solver
Put the test into `diagnostic_namespace` test folder even though it's not in the diagnostic namespace, because it should be soon.
r? lcnr
cc `@weiznich`
Update Cargo specific diagnostics in check-cfg
This PR updates the Cargo specific diagnostics for check-cfg/`unexpected_cfgs` lint.
Specifically it update to new url and use the double-column (instead of one) in the Cargo directive suggestion.
`@rustbot` label +F-check-cfg
cc `@weihanglo`
Only consider ambiguous goals when finding best obligation for ambiguities
We don't care about ambiguous goals when reporting true errors, and vice versa for ambiguities.
r? lcnr
Enable reusing CI Docker cache when running CI images locally
When running a CI image locally, e.g. using `DEPLOY=1 src/ci/docker/run.sh dist-x86_64-linux`, it can take a long time until the Docker image is built, which is annoying.
Since we now use proper Docker caching on CI, it should be possible to just `docker pull` the prebuilt image to reuse the cache. We didn't want to do this on CI, since our caching key isn't perfect and it's possible that we can miss some changes, but I think that for local usage it is fine (we could introduce some env. var. to force disable the image download, if needed).
r? `@Mark-Simulacrum`
add support to override lldb binary path for ./x test
When running debuginfo tests I couldn't set custom build of lldb. The `src/bootstrap/src/core/build_steps/test.rs` has "lldb" hardcoded. I ended up hacking `src/bootstrap/src/core/build_steps/test.rs` just to get the tests running the way I wanted.
Then I've found out that we can override `gdb` under [build] section. This PR enables the same for `lldb`
This fixes building `std` for targets like `mipsel-unknown-netbsd`.
If `c_long` is an `i64`, this conversion works with `Into`. But if it's
an `i32`, this failed to convert a `u32` to an `i32`.
interpret, miri: uniform treatments of intrinsics/functions with and without return block
A long time ago we didn't have a `dest: &MPlaceTy<'tcx, Self::Provenance>` for diverging functions, and since `dest` is used so often we special-cased these non-returning intrinsics and functions so that we'd have `dest` available everywhere else. But this has changed a while ago, now only the return block `ret` is optional, and there's a convenient `return_to_block` function for dealing with the `None` case.
So there no longer is any reason to treat diverging intrinsics/functions any different from those that do return.
Use `unchecked_sub` in `split_at`
LLVM currently isn't figuring it out on its own, even in the checked version where it hypothetically could.
Before: <https://rust.godbolt.org/z/PEY38YrKs>
```llvm
bb1: ; preds = %start
%4 = getelementptr inbounds float, ptr %x.0, i64 %n
%5 = sub i64 %x.1, %n
```
After:
```llvm
bb1: ; preds = %start
%4 = getelementptr inbounds float, ptr %x.0, i64 %n
%5 = sub nuw i64 %x.1, %n
```
This is not using the wrapper because there's already a ubcheck covering it, so I don't want this to get a second one once #121571 lands.
---
This is basically the same as #108763, since `split_at` is essentially doing two `get_unchecked`s.
Various improvements to entrypoint code
This moves some code around and adds some documentation comments to make it easier to understand what's going on with the entrypoint logic, which is a bit complicated.
The only change in behavior is consolidating the error messages for unix_sigpipe to make the code slightly simpler.
Improve several `Read` implementations
- `read_to_end` and `read_to_string` for `Cursor`
- Error on OOM in `read_to_string` of `&[u8]` and `VecDeque<u8>`
- Avoid making the slices contiguous in `VecDeque::read_to_string`
- ~`read_exact` and (unstable) `read_buf_exact` for `Take`~
- ~`read_buf` for `UnixStream` and `&UnixStream`~ (moved to #123084)
- `read_to_end` for `ChildStdErr`
Fix ignored tests for formatting
This PR fixes the ignored rules in `rustfmt.toml` that were changed in https://github.com/rust-lang/rust/pull/124613 to allow formatting `rmake.rs` but ended up allowing formatting every Rust files in `tests/`.
The fix is a bit involved since we need to workaround a [`.gitignore` pattern limitation](https://git-scm.com/docs/gitignore#_pattern_format):
> An optional prefix "!" which negates the pattern; any matching file excluded by a previous pattern will become included again. It is not possible to re-include a file if a parent directory of that file is excluded. Git doesn’t list excluded directories for performance reasons, so any patterns on contained files have no effect, no matter where they are defined.
Workaround using https://stackoverflow.com/a/5534865
I tested the fix by changing the formatting in an `rmake.rs` and UI test, and verifying that only the `rmake.rs` files were formatted.
Fixes https://github.com/rust-lang/rust/pull/124613#issuecomment-2094094670
cc `@GuillaumeGomez`
r? `@onur-ozkan`
This moves some code around and adds some documentation comments to make
it easier to understand what's going on with the entrypoint logic, which
is a bit complicated.
The only change in behavior is consolidating the error messages for
unix_sigpipe to make the code slightly simpler.
Docs: suggest `uN::checked_sub` instead of check-then-unchecked
As of #124114 it's exactly the same in codegen, so might as well not use `unsafe`.
Note that this is only for *unsigned*, since the overflow conditions for `iN::checked_sub` are more complicated.
Remove an unnecessary cast
Very minor thing, obviously, but I randomly saw this unnecessary cast showing up in the UbChecks, so might as well get rid of it.
We do not coerce `&mut &mut T -> *mut mut T`
Resolves#34117 by declaring it to be "working as intended" until someone RFCs it or whatever other lang proposal would be required. It seems a bit of a footgun, but perhaps there are strong reasons to allow it anyways. Seeing as how I often have to be mindful to not allow a pointer to coerce the wrong way in my FFI work, I am inclined to think not, but perhaps it's fine in some use-case and that's actually more common?
Set non-leaf frame pointers on Fuchsia targets
This is part of our work to enable shadow call stack sanitization on Fuchsia, see [this Fuchsia issue](https://g-issues.fuchsia.dev/issues/327643884).
r? ``@tmandry``
Move thread parking to `sys::sync`
Part of #117276.
I'll leave the platform-specific API abstractions in `sys::pal`, as per the initial proposal. I'm not entirely sure whether we'll want to keep it that way, but that remains to be seen.
r? ``@ChrisDenton`` (if you have time)
Reduce code size of `thread::set_current`
#123265 introduced a rather large binary size regression, because it added an `unwrap()` call on a `Result<(), Thread>`, which in turn pulled its rather heavy `Debug` implementation. This PR fixes this by readding the `rtassert!` that was removed.
Enable `--check-cfg` by default in UI tests
This PR enables-by-default `--check-cfg` in UI tests, now that it has become stable.
To do so this PR does 2 main things:
- it introduce the `no-auto-check-cfg` directive to `compiletest`, to prevent any `--check-cfg` args (only to be used for `--check-cfg` tests)
- it updates the _remaining_[^1] UI tests by either:
- allowing the lint when neither expecting the lint nor giving the check-cfg args make sense
- give the appropriate check-cfg args
- or expect the lint, when it useful
[^1]: some preparation work was done in #123577#123702
I highly recommend reviewing this PR commit-by-commit.
r? `@jieyouxu`
As of 124114 it's exactly the same in codegen, so might as well not use `unsafe`.
Note that this is only for *unsigned*, since the overflow conditions for `iN::checked_sub` are more complicated.
