rustc: Remove soft-float from MIPS targets
Right now two MIPS targets in the compiler, `mips-unknown-linux-{gnu,musl}` both
generate object files using the soft-float ABI through LLVM by default. This is
also expressed as the `-C soft-float` codegen option and otherwise isn't used
for any other target in the compiler. This option was added quite some time ago
(back in #9617), and nowadays it's more appropriate to be done through a codegen
option.
This is motivated by #34743 which necessitated an upgrade in the CMake
installation on our bots which necessitated an upgrade in the Ubuntu version
which invalidated the MIPS compilers we were using. The new MIPS compilers
(coming from Debian I believe) all have hard float enabled by default and soft
float support not built in. This meant that we couldn't upgrade the bots
until #34841 landed because otherwise we would fail to compile C code as the
`-msoft-float` option wouldn't work.
Unfortunately, though, this means that once we upgrade the bots the C code we're
compiling will be compiled for hard float and the Rust code will be compiled
for soft float, a bad mismatch! This PR remedies the situation such that Rust
will compile with hard float as well.
If this lands it will likely produce broken nightlies for a day or two while we
get around to upgrading the bots because the current C toolchain only produces
soft-float binaries, and now rust will be hard-float. Hopefully, though, the
upgrade can go smoothly!
implement AddAssign for String
Currently `String` implements `Add` but not `AddAssign`. This PR fills in that gap.
I played around with having `AddAssign` (and `Add` and `push_str`) take `AsRef<str>` instead of `&str`, but it looks like that breaks arguments that implement `Deref<Target=str>` and not `AsRef<str>`. Comments in [`libcore/convert.rs`](https://github.com/rust-lang/rust/blob/master/src/libcore/convert.rs#L207-L213) make it sound like we could fix this with a blanket impl eventually. Does anyone know what's blocking that?
core: impl From<T> for Option<T>
First, the semantics of this `impl` seem spot on. If I have a value `T`, and I wish to make a `Option<T>`, then `Option::from(val)` should always give `Some(val)`.
Second, this allows improvement for several APIs that currently take `Option<T>` as arguments. Consider:
```rust
fn set_read_timeout(&mut self, timeout: Option<u32>) {
// ...
}
x.set_read_timeout(Some(30));
x.set_read_timeout(Some(10));
x.set_read_timeout(None);
```
With this `impl`:
```rust
fn set_read_timeout<T: Into<Option<u32>>>(&mut self, timeout: T) {
let timeout = timeout.into();
// ...
}
x.set_read_timeout(30);
x.set_read_timeout(10);
x.set_read_timeout(Some(10)); // backwards compatible
x.set_read_timeout(None);
```
The change to those methods aren't included, but could be modified later.
r? @sfackler
mk: Stop using cmake for compiler-rt
The compiler-rt build system has been a never ending cause of pain for Rust
unfortunately:
* The build system is very difficult to invoke and configure to only build
compiler-rt, especially across platforms.
* The standard build system doesn't actually do what we want, not working for
some of our platforms and requiring a significant number of patches on our end
which are difficult to apply when updating compiler-rt.
* Compiling compiler-rt requires LLVM to be compiled, which... is a big
dependency! This also means that over time compiler-rt is not guaranteed to
build against older versions of LLVM (or newer versions), and we often want to
work with multiple versions of LLVM simultaneously.
The makefiles and rustbuild already know how to compile C code, the code here is
far from the *only* C code we're compiling. This patch jettisons all logic to
work with compiler-rt's build system and just goes straight to the source. We
just list all files manually (copied from compiler-rt's
lib/builtins/CMakeLists.txt) and compile them into an archive.
It's likely that this means we'll fail to pick up new files when we upgrade
compiler-rt, but that seems like a much less significant cost to pay than what
we're currently paying.
cc #34400, first steps towards that
The compiler-rt build system has been a never ending cause of pain for Rust
unfortunately:
* The build system is very difficult to invoke and configure to only build
compiler-rt, especially across platforms.
* The standard build system doesn't actually do what we want, not working for
some of our platforms and requiring a significant number of patches on our end
which are difficult to apply when updating compiler-rt.
* Compiling compiler-rt requires LLVM to be compiled, which... is a big
dependency! This also means that over time compiler-rt is not guaranteed to
build against older versions of LLVM (or newer versions), and we often want to
work with multiple versions of LLVM simultaneously.
The makefiles and rustbuild already know how to compile C code, the code here is
far from the *only* C code we're compiling. This patch jettisons all logic to
work with compiler-rt's build system and just goes straight to the source. We
just list all files manually (copied from compiler-rt's
lib/builtins/CMakeLists.txt) and compile them into an archive.
It's likely that this means we'll fail to pick up new files when we upgrade
compiler-rt, but that seems like a much less significant cost to pay than what
we're currently paying.
cc #34400, first steps towards that
Add IpAddr common methods
Per https://github.com/rust-lang/rfcs/pull/1668#issuecomment-230867962 no RFC is needed here.
The generated documentation for these methods is being weird. It shows a deprecation message referencing #27709 for each of them even though two of the referenced methods were stabilized as part of that issue. I don't know how best to address that.
Add some warnings to std::env::current_exe
/cc #21889 @rust-lang/libs @semarie
I started writing this up. I'm not sure if we want to go into other things and in what depth; we don't currently have a lot of security-specific documentation to model after.
Thoughts?
Retry on EINTR in Bytes and Chars.
>Since Bytes and Chars called directly into Read::read, they didn't use any of the retrying wrappers. This allows both iterator types to retry.
Right now two MIPS targets in the compiler, `mips-unknown-linux-{gnu,musl}` both
generate object files using the soft-float ABI through LLVM by default. This is
also expressed as the `-C soft-float` codegen option and otherwise isn't used
for any other target in the compiler. This option was added quite some time ago
(back in #9617), and nowadays it's more appropriate to be done through a codegen
option.
This is motivated by #34743 which necessitated an upgrade in the CMake
installation on our bots which necessitated an upgrade in the Ubuntu version
which invalidated the MIPS compilers we were using. The new MIPS compilers
(coming from Debian I believe) all have hard float enabled by default and soft
float support not built in. This meant that we couldn't upgrade the bots
until #34841 landed because otherwise we would fail to compile C code as the
`-msoft-float` option wouldn't work.
Unfortunately, though, this means that once we upgrade the bots the C code we're
compiling will be compiled for hard float and the Rust code will be compiled
for soft float, a bad mismatch! This PR remedies the situation such that Rust
will compile with hard float as well.
If this lands it will likely produce broken nightlies for a day or two while we
get around to upgrading the bots because the current C toolchain only produces
soft-float binaries, and now rust will be hard-float. Hopefully, though, the
upgrade can go smoothly!
Run base::internalize_symbols() even for single-codegen-unit crates.
The initial linkage-assignment (especially for closures) is a conservative one that makes some symbols more visible than they need to be. While this is not a correctness problem, it does force the LLVM inliner to be more conservative too, which results in poor performance. Once translation is based solely on MIR, it will be easier to also make the initial linkage assignment a better fitting one. Until then `internalize_symbols()` does a good job of preventing most performance regressions.
This should solve the regressions reported in https://github.com/rust-lang/rust/issues/34891 and maybe also those in https://github.com/rust-lang/rust/issues/34831.
As a side-effect, this will also solve most of the problematic cases described in https://github.com/rust-lang/rust/issues/34793. Not reliably so, however. For that, we still need a solution like the one implement in https://github.com/rust-lang/rust/pull/34830.
cc @rust-lang/compiler
The initial linkage-assignment (especially for closures) is a conservative one that makes some symbols more visible than they need to be. While this is not a correctness problem, it does force the LLVM inliner to be more conservative too, which results in poor performance. Once translation is based solely on MIR, it will be easier to also make the initial linkage assignment a better fitting one. Until then `internalize_symbols()` does a good job of preventing most performance regressions.
macros: fix bug in `stmt` matchers
Today, `stmt` matchers stop too early when parsing expression statements that begin with non-braced macro invocations. For example,
```rust
fn main() {
macro_rules! m { ($s:stmt;) => { $s } }
id!(vec![].push(0););
//^ Before this PR, the `stmt` matcher only consumes "vec![]", so this is an error.
//| After this PR, the `stmt` matcher consumes "vec![].push(0)", so this compiles.
}
```
This change is backwards compatible due to the follow set for `stmt`.
r? @eddyb
Guillaume Gomez
