Reduce CString allocations in std as much as possible
Currently, every operation involving paths in `fs` allocates memory to hold the path before sending it through the syscall. This PR instead uses a stack allocation (chosen size is somewhat arbitrary) when the path is short before falling back to heap allocations for long paths.
Benchmarks show that the stack allocation is ~2x faster for short paths:
```
test sys::unix::fd::tests::bench_heap_path_alloc ... bench: 34 ns/iter (+/- 2)
test sys::unix::fd::tests::bench_stack_path_alloc ... bench: 15 ns/iter (+/- 1)
```
For long paths, I couldn't find any measurable difference.
---
I'd be surprised if I was the first to think of this, so I didn't fully flush out the PR. If this change is desirable, I'll make use of `run_with_cstr` across all platforms in every fs method (currently just unix open for testing). I also added an `impl From<FromBytesWithNulError>` which is presumably a no-no (or at least needs to be done in another PR).
---
Also see https://github.com/nix-rust/nix/pull/1655 with a bunch of discussion where I'm doing something similar.
I refactored the code:
- Removed handling of methods, as it felt entirely unnecessary
- Removed clippy utils (obviously...)
- Used some shiny compiler features
(let-else is very handy for lints 👀)
- I also renamed the lint to `for_loop_over_fallibles` (note: no `s`).
I'm not sure what's the naming convention here, so maybe I'm wrong.
Update CI to use Android NDK r25b
This commit updates the CI definitions to use the most recent Android LTS NDK release: r25b. Changes since the last NDK used by Rust negate the need to generate "standalone toolchains" and newer NDKs can be used in-place.
See https://developer.android.com/ndk/guides/other_build_systems#overview
Remove empty core::lazy and std::lazy
PR #98165 with commits 7c360dc117d554a11f7193505da0835c4b890c6f and c1a2db3372a4d6896744919284f3287650a38ab7 has moved all of the components of these modules into different places, namely {std,core}::sync and {std,core}::cell. The empty modules remained. As they are unstable, we can simply remove them.
Fix llvm-tblgen for cross compiling
- Let llvm-config tell us where to find its tools
- Add llvm-tblgen to rust-dev for cross-compiling
Fixes#86890.
r? ````@jyn514````
rustdoc: Document effect of fundamental types
This was implemented in https://github.com/rust-lang/rust/pull/96565, but not documented. But it's a useful feature for everyone who implements own wrapper (especially wrappers-around-pointers) types, so that they can behave like pointers (and stdlib wrappers) do -- so here goes a mention in the unstable section of the rustdoc book.
(That is where I initially looked to find tricks for making my own wrapper types be transparent to documentation).
don't ICE when normalizing closure input tys
We were ICEing while rendering diagnostics because `universe_causes` is expected to track every universe created in the typeck's infcx.
`normalize_and_add_constraints` doesn't update `universe_causes`
when creating new universes, causing an ICE. Remove it!
Add spans to better track normalization constraints.
Fix couple places where `universe_causes` is not updated correctly to
track newly added universes.
Fixes#102800
~Fixess #99665~ (UPDATE: no longer true; the issue has a different failure path than when this PR was created and should be fixed by #101708, but the changes in this PR are still correct and should prevent potential future ICEs)
VS Code problem matcher are restricted to be static "regexes". You can't
create a problem matcher dynamically, and you can't use custom code in
lieu of problem matcher.
This creates a problem for rust/cargo compiler errors. They use paths
relative to the root of the Cargo workspace, but VS Code doesn't
necessary know where that root is.
Luckily, there's a way out: our current problem matcher is defined like
this:
"fileLocation": [ "autoDetect", "${workspaceRoot}" ],
That means that relative pahts would be resoleved relative to workspace
root. VS Code allows to specify a command inside `${}`. So we can plug
custom logic there to fetch Cargo's workspace root!
And that's exactly what this PR is doing!
test: run-make: skip when cross-compiling
This test fails when targeting aarch64 Android. Instead of adding yet another architecture here (and one that's increasingly more common as the host), let's replace the growing list of architectures with ignore-cross-compile.
Make the `config.src` handling for downloadable bootstrap more conservative
In particular, this supports build directories within an unrelated git repository. Fixes https://github.com/rust-lang/rust/issues/102562.
As a side effect, it will fall back to the old logic when the source directory is being built from a tarball within an unrelated git repository. However, that second case is unsupported and untested; we reserve the right to break it in the future.
`@cr1901` can you confirm this fixes your problem?
cc `@kleisauke,` I believe this will also fix your issue (although your use case still isn't supported).
r? `@Mark-Simulacrum`
`EscapeAscii` is not an `ExactSizeIterator`
Fixes#99878
Do we want/need `EscapeAscii` to be an `ExactSizeIterator`? I guess we could precompute the length of the output if so?
Fix gdb-cmd for rust-gdbgui
With https://github.com/cs01/gdbgui/pull/198, the way that gdbgui arguments were specified changed. I've tested this with program generated from `cargo new --bin` and it worked as gdbgui should.
Closes#76383.
