Optimize pattern matching
These commits speed up the `match-stress-enum` benchmark, which is very artificial, but the changes are simple enough that it's probably worth doing.
r? `@Nadrieril`
As discussed here
https://github.com/rust-lang/rust/pull/88300#issuecomment-936097710
I felt this was the best place to put this (rather than next to
ExitStatusExt). After all, it's a property of the ExitStatus type on
Unix.
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
As discussed here
https://github.com/rust-lang/rust/pull/88300#issuecomment-936085371
exit is (conventionally) a library function, with _exit being the
actual system call.
I have checked the other references and they say "if the process
terminated by calling `exti`". I think despite the slight
imprecision (strictly, it should read iff ... `_exit`), this is
clearer. Anyone who knows about the distinction between `exit` and
`_exit` will not be confused.
`_exit` is the correct traditional name for the system call, despite
Linux calling it `exit_group` or `exit`:
https://www.freebsd.org/cgi/man.cgi?query=_exit&sektion=2&n=1
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
The `unreachable!` docs previously did not mention that there was a second
form, `unreachable!("message")` that could be used to specify a custom panic
message,
The docs now mention this in the same style as currently used for `unimplemented!`:
https://doc.rust-lang.org/core/macro.unimplemented.html#panics
Fix assertion failures in `OwnedHandle` with `windows_subsystem`.
As discussed in #88576, raw handle values in Windows can be null, such
as in `windows_subsystem` mode, or when consoles are detached from a
process. So, don't use `NonNull` to hold them, don't assert that they're
not null, and remove `OwnedHandle`'s `repr(transparent)`. Introduce a
new `HandleOrNull` type, similar to `HandleOrInvalid`, to cover the FFI
use case.
r? `@joshtriplett`
Added the --temps-dir option
Fixes#10971.
The new `--temps-dir` option puts intermediate files in a user-specified directory. This provides a fix for the issue where parallel invocations of rustc would overwrite each other's intermediate files.
No files are kept in the intermediate directory unless `-C save-temps=yes`.
If additional files are specifically requested using `--emit asm,llvm-bc,llvm-ir,obj,metadata,link,dep-info,mir`, these will be put in the output directory rather than the intermediate directory.
This is a backward-compatible change, i.e. if `--temps-dir` is not specified, the behavior is the same as before.
We would like to check for errors with AVX512,
but we don't pick our CPU. So, detect available features.
This variance in checks stochastically reveals issues.
Nondeterminism is acceptable as our goal is protecting downstream.
adjust documented inline-asm register constraints
This change more clearly specifies how `reg` and `reg_thumb` work with ARM, Thumb2, and Thumb1 code.
Based upon the [llvm documentation](https://llvm.org/docs/LangRef.html#supported-constraint-code-list) for register constraint codes.
To be clear, this just updates the docs to match what already happens with rustc/llvm.
No change in the compiler is required to make it match this new documentation.
Only use `clone3` when needed for pidfd
In #89522 we learned that `clone3` is interacting poorly with Gentoo's
`sandbox` tool. We only need that for the unstable pidfd extensions, so
otherwise avoid that and use a normal `fork`.
This is a re-application of beta #89924, now that we're aware that we need
more than just a temporary release fix. I also reverted 12fbabd27f, as
that was just fallout from using `clone3` instead of `fork`.
r? `@Mark-Simulacrum`
cc `@joshtriplett`
Remove potential useless data for search index
I uncovered this case when working on https://github.com/rust-lang/rust/pull/90726 to debug https://github.com/rust-lang/rust/pull/90385.
Explanations: if we have a full generic, we check if it has generics then we do the following:
* If it has only one generic, we remove one nested level in order to not keep the "parent" generic (since it has empty name, it's useless after all).
* Otherwise we add it alongside its generics.
However, I didn't handle the case where a generic had no generics. Meaning that we were adding items with empty names in the search index. So basically useless data in the search index.
r? `@camelid`
Use computed visibility in rustdoc
This PR changes `librustdoc` to use computed visibility instead of syntactic visibility. It was initially part of #88019, but was separated due to concerns that it might cause a regression somewhere we couldn't predict.
r? `@jyn514`
cc `@cjgillot` `@petrochenkov`
Update llvm submodule
This includes debug info generation fix, that fixes#90301.
Also includes WASM backend related fix for https://bugs.llvm.org/show_bug.cgi?id=52352
(I haven't found a corresponding Rust bug).
Because after PR 86041, the optimizer no longer load-merges at the LLVM IR level, which might be part of the perf loss. (I'll run perf and see if this makes a difference.)
Also I added a codegen test so this hopefully won't regress in future -- it passes on stable and with my change here, but not on the 2021-11-09 nightly.
