This reverts commit 7ce620dd7c.
The const-hacks introduces bugs, and they make the code harder to maintain.
Let's wait until we can constify these functions without changing their implementation.
std: optimize `dlsym!` macro and add a test for it
The `dlsym!` macro always ensures that the name string is nul-terminated, so there is no need to perform the check at runtime. Also, acquire loads are generally faster than a load and a barrier, so use them. This is only false in the case where the symbol is missing, but that shouldn't matter too much.
std: make address resolution weirdness local to SGX
Currently, the implementations of `TcpStream::connect` and its cousins take an `io::Result<&SocketAddr>` as argument, which is very weird, as most of them then `?`-try the result immediately to access the actual address. This weirdness is however necessitated by a peculiarity of the SGX networking implementation:
SGX doesn't support DNS resolution but rather accepts hostnames in the same place as socket addresses. So, to make e.g.
```rust
TcpStream::connect("example.com:80")`
```
work, the DNS lookup returns a special error (`NonIpSockAddr`) instead, which contains the hostname being looked up. When `.to_socket_addrs()` fails, the `each_addr` function used to select an address will pass the error to the inner `TcpStream::connect` implementation, which in SGX's case will inspect the error and try recover the hostname from it. If
that succeeds, it continues with the found hostname.
This is pretty obviously a terrible hack and leads to buggy code (for instance, when users use the result of `.to_socket_addrs()` in their own `ToSocketAddrs` implementation to select from a list of possible URLs, the only URL used will be that of the last item tried). Still, without changes to the SGX usercall ABI, it cannot be avoided.
Therefore, this PR aims to minimise the impact of that weirdness and remove it from all non-SGX platforms. The inner `TcpStream::connect`, et al. functions now receive the `ToSocketAddrs` type directly and call `each_addr` (which is moved to `sys::net::connection`) themselves. On SGX, the implementation uses a special `each_addr` which contains the whole pass-hostname-through-error hack.
As well as making the code cleaner, this also opens up the possibility of reusing newly created sockets even if a connection request fails – but I've left that for another PR.
CC `@raoulstrackx`
Add `__isPlatformVersionAtLeast` and `__isOSVersionAtLeast` symbols
## Motivation
When Objective-C code uses ```@available(...)`,`` Clang inserts a call to [`__isPlatformVersionAtLeast`](https://github.com/llvm/llvm-project/blob/llvmorg-20.1.0/compiler-rt/lib/builtins/os_version_check.c#L276) (`__isOSVersionAtLeast` in older Clang versions). These symbols not being available sometimes ends up causing linker errors. See the new test `tests/run-make/apple-c-available-links` for a minimal reproducer.
The workaround is to link `libclang_rt.osx.a`, see e.g. https://github.com/alexcrichton/curl-rust/issues/279. But that's very difficult for users to figure out (and the backreferences to that issue indicates that people are still running into this in their own projects every so often).
For another recent example, this is preventing `rustc` from using LLVM assertions on macOS, see https://github.com/rust-lang/rust/pull/62592#issuecomment-510670657 and https://github.com/rust-lang/rust/pull/134275#issuecomment-2543067830.
It is also a blocker for [setting the correct minimum OS version in `cc-rs`](https://github.com/rust-lang/rust/issues/136113), since fixing this in `cc-rs` might end up introducing linker errors in places where we weren't before (by default, if using e.g. ```@available(macos`` 10.15, *)`, the symbol usually happens to be left out, since `clang` defaults to compiling for the host macOS version, and thus things _seem_ to work - but the availability check actually compiles down to nothing, which is a huge correctness footgun for running on older OSes).
(My super secret evil agenda is also to expose some variant of ```@available``` in Rust's `std` after https://github.com/rust-lang/rfcs/pull/3750 progresses further, will probably file an ACP for this later. But I believe this PR has value regardless of those future plans, since we'd be making C/Objective-C/Swift interop easier).
## Solution
Implement `__isPlatformVersionAtLeast` and `__isOSVersionAtLeast` as part of the "public ABI" that `std` exposes.
**This is insta-stable**, in the same sense that additions to `compiler-builtins` are insta-stable, though the availability of these symbols can probably be considered a "quality of implementation" detail rather than a stable promise.
I originally proposed to implement this in `compiler-builtins`, see https://github.com/rust-lang/compiler-builtins/pull/794, but we discussed moving it to `std` instead ([Zulip thread](https://rust-lang.zulipchat.com/#narrow/channel/219381-t-libs/topic/Provide.20.60__isPlatformVersionAtLeast.60.20in.20.60std.60.3F/with/507880717)), which makes the implementation substantially simpler, and we avoid gnarly issues with requiring the user to link `libSystem.dylib` (since `std` unconditionally does that).
Note that this does not solve the linker errors for (pure) `#![no_std]` users, but that's _probably_ fine, if you are using ```@available``` to test the OS version on Apple platforms, you're likely also using `std` (and it is still possible to work around by linking `libclang_rt.*.a`).
A thing to note about the implementation, I've choosen to stray a bit from LLVM's upstream implementation, and not use `_availability_version_check` since [it has problems when compiling with an older SDK](https://github.com/llvm/llvm-project/issues/64227). Instead, we use `sysctl kern.osproductversion` when available to still avoid the costly PList lookup in most cases, but still with a fall back to the PList lookup when that is not available (with the PList fallback being is similar to LLVM's implementation).
## Testing
Apple has a lot of different "modes" that they can run binaries in, which can be a bit difficult to find your bearings in, but I've tried to be as thorough as I could in testing them all.
Tested using roughly the equivalent of `./x test library/std -- platform_version` on the following configurations:
- macOS 14.7.3 on a Macbook Pro M2
- `aarch64-apple-darwin`
- `x86_64-apple-darwin` (under Rosetta)
- `aarch64-apple-ios-macabi`
- `x86_64-apple-ios-macabi` (under Rosetta)
- `aarch64-apple-ios` (using Xcode's "Designed for iPad" setting)
- `aarch64-apple-ios-sim` (in iOS Simulator, as iPhone with iOS 17.5)
- `aarch64-apple-ios-sim` (in iOS Simulator, as iPad with iOS 18.2)
- `aarch64-apple-tvos-sim` (in tvOS Simulator)
- `aarch64-apple-watchos-sim` (in watchOS Simulator)
- `aarch64-apple-ios-sim` (in visionOS simulator, using Xcode's "Designed for iPad" setting)
- `aarch64-apple-visionos-sim` (in visionOS Simulator)
- macOS 15.3.1 VM
- `aarch64-apple-darwin`
- `aarch64-apple-ios-macabi`
- macOS 10.12.6 on an Intel Macbook from 2013
- `x86_64-apple-darwin`
- `i686-apple-darwin`
- `x86_64-apple-ios` (in iOS Simulator)
- iOS 9.3.6 on a 1st generation iPad Mini
- `armv7-apple-ios` with an older compiler
Along with manually inspecting the output of `version_from_sysctl()` and `version_from_plist()`, and verifying that they actually match what's expected.
I believe the only real omissions here would be:
- `aarch64-apple-ios` on a newer iPhone that has `sysctl` available (iOS 11.4 or above).
- `aarch64-apple-ios` on a Vision Pro using Xcode's "Designed for iPad" setting.
But I don't have the hardware available to test those.
``@rustbot`` label O-apple A-linkage -T-compiler -A-meta -A-run-make
try-job: aarch64-apple
std: Implement WASIp2-specific stdio routines
This commit is an extension of rust-lang/rust#145944 but applied to stdio specifically. The stdio routines are updated away from WASIp1 APIs to using WASIp2 APIs natively. The end goal is to eventually drop the dependency on WASIp1 APIs in the standard library entirely in favor of exclusively depending on WASIp2.
This commit is an extension of previous libstd support but applied to stdio
specifically. The stdio routines are updated away from WASIp1 APIs to using
WASIp2 APIs natively. The end goal is to eventually drop the dependency on
WASIp1 APIs in the standard library entirely in favor of exclusively depending
on WASIp2.
Adds the following items under the `read_buf_at` feature:
- `std::os::unix::fs::FileExt::read_buf_at`
- `std::os::unix::fs::FileExt::read_buf_exact_at`
- `std::os::windows::fs::FileExt::seek_read_buf`
* Ensure nul-termination of the symbol name at compile-time
* Use an acquire load instead of a relaxed load and acquire fence
* Properly use `unsafe` and add safety comments
* Add tests
std: Start supporting WASIp2 natively
This commit is the start of an effort to support WASIp2 natively in the
standard library. Before this commit the `wasm32-wasip2` target behaved
exactly like `wasm32-wasip1` target by importing APIs from the core wasm
module `wasi_snapshot_preview1`. These APIs are satisfied by the
`wasm-component-ld` target by using an [adapter] which implements WASIp1
in terms of WASIp2. This adapter comes at a cost, however, in terms of
runtime indirection and instantiation cost, so ideally the adapter would
be removed entirely. The purpose of this adapter was to provide a
smoother on-ramp from WASIp1 to WASIp2 when it was originally created.
The `wasm32-wasip2` target has been around for long enough now that it's
much more established. Additionally the only thing historically blocking
using WASIp2 directly was implementation effort. Work is now underway to
migrate wasi-libc itself to using WASIp2 directly and now seems as good
a time as any to migrate the Rust standard library too.
Implementation-wise the milestones here are:
* The `wasm32-wasip2` target now also depends on the `wasi` crate at
version 0.14.* in addition to the preexisting dependency of 0.11.*.
The 0.14.* release series binds WASIp2 APIs instead of WASIp1 APIs.
* Some preexisting naming around `mod wasi` or `wasi.rs` was renamed to
`wasip1` where appropriate. For example `std::sys::pal::wasi` is now
called `std::sys::pal::wasip1`.
* More platform-specific WASI modules are now split between WASIp1 and
WASIp2. For example getting the current time, randomness, and
process arguments now use WASIp2 APIs directly instead of using WASIp1
APIs that require an adapter.
It's worth pointing out that this PR does not migrate the entire
standard library away from using WASIp1 APIs on the `wasm32-wasip2`
target. Everything related to file descriptors and filesystem APIs is
still using WASIp1. Migrating that is left for a future PR. In the
meantime the goal of this change is to lay the groundwork necessary for
migrating in the future. Eventually the goal is to drop the `wasi`
0.11.* dependency on the `wasm32-wasip2` target (the `wasm32-wasip1`
target will continue to retain this dependency).
[adapter]: https://github.com/bytecodealliance/wasmtime/blob/main/crates/wasi-preview1-component-adapter/README.md
Fix `sys::process::windows::tests::test_thread_handle` spurious failure
Instead of sleeping, wait for the process to finish so that we can be sure it's done. We use a timeout because otherwise this test can be stuck indefinitely if it fails (unfortunately std doesn't currently have a way to wait with a timeout so a manual OS API call is necessary).
I also changed the test to run `whoami` and pipe the output to null so that it doesn't clutter up the test output.
Fixesrust-lang/rust#146024
Previously, attempting to create/truncate a file without write/append access
would result in platform-specific error messages:
- Unix: "Invalid argument"
- Windows: raw OS error code 87
These error codes look like system errors, which could waste hours
of debugging for what is actually an API misuse issue.
fix(std): Fix undefined reference to __my_thread_exit on QNX 8.0
When cross-compiling for the x86_64/aarch64-unknown-nto-qnx800 target (QNX SDP 8.0), the build fails during the final link stage with the error:
```
error: linking with `qcc` failed: exit status: 1
...
= note: undefined reference to `__my_thread_exit'
```
- **On QNX 7.1**: The __my_thread_exit symbol is defined and exported by the main C library (libc.a/libc.so). The std backtrace code can therefore successfully take its address at compile time.
- **On QNX 8.0**: As part of a toolchain modernization, this symbol has been refactored. It is no longer present in any of the standard system libraries (.a or .so).
This patch addresses the problem at its source by conditionally compiling the problematic code.
Fixesrust-lang/rust#142726
This commit is the start of an effort to support WASIp2 natively in the
standard library. Before this commit the `wasm32-wasip2` target behaved
exactly like `wasm32-wasip1` target by importing APIs from the core wasm
module `wasi_snapshot_preview1`. These APIs are satisfied by the
`wasm-component-ld` target by using an [adapter] which implements WASIp1
in terms of WASIp2. This adapter comes at a cost, however, in terms of
runtime indirection and instantiation cost, so ideally the adapter would
be removed entirely. The purpose of this adapter was to provide a
smoother on-ramp from WASIp1 to WASIp2 when it was originally created.
The `wasm32-wasip2` target has been around for long enough now that it's
much more established. Additionally the only thing historically blocking
using WASIp2 directly was implementation effort. Work is now underway to
migrate wasi-libc itself to using WASIp2 directly and now seems as good
a time as any to migrate the Rust standard library too.
Implementation-wise the milestones here are:
* The `wasm32-wasip2` target now also depends on the `wasi` crate at
version 0.14.* in addition to the preexisting dependency of 0.11.*.
The 0.14.* release series binds WASIp2 APIs instead of WASIp1 APIs.
* Some preexisting naming around `mod wasi` or `wasi.rs` was renamed to
`wasip1` where appropriate. For example `std::sys::pal::wasi` is now
called `std::sys::pal::wasip1`.
* More platform-specific WASI modules are now split between WASIp1 and
WASIp2. For example getting the current time, randomness, and
process arguments now use WASIp2 APIs directly instead of using WASIp1
APIs that require an adapter.
It's worth pointing out that this PR does not migrate the entire
standard library away from using WASIp1 APIs on the `wasm32-wasip2`
target. Everything related to file descriptors and filesystem APIs is
still using WASIp1. Migrating that is left for a future PR. In the
meantime the goal of this change is to lay the groundwork necessary for
migrating in the future. Eventually the goal is to drop the `wasi`
0.11.* dependency on the `wasm32-wasip2` target (the `wasm32-wasip1`
target will continue to retain this dependency).
[adapter]: https://github.com/bytecodealliance/wasmtime/blob/main/crates/wasi-preview1-component-adapter/README.md
Fix STD build failing for target_os = "espidf"
A regression from rust-lang/rust#142938
cc `@lolbinarycat`
cc `@ibraheemdev`
ESP-IDF (and a few other embedded Tier-3 systems) is considered `cfg(unix)`, but it does not have the `O_NOFOLLOW` flag because neither of its three supported filesystems (FATFS, LitteLF and Spiffs) has symbolic links in the first place.
What this fix does is to keep the `set_permissions_nofollow` method available and non-failing for ESP-IDF, but it behaves as if no `O_NONFOLLOW` was set. This should be fine as there is nothing to follow in the first place, as there are no symbolic links there.
EDIT: Also added the same fix for Horizon, as requested by `@Meziu.`
Move WTF-8 code from std into core and alloc
This is basically a small portion of rust-lang/rust#129411 with a smaller scope. It *does not*\* affect any public APIs; this code is still internal to the standard library. It just moves the WTF-8 code into `core` and `alloc` so it can be accessed by `no_std` crates like `backtrace`.
> \* The only public API this affects is by adding a `Debug` implementation to `std::os::windows::ffi::EncodeWide`, which was not present before. This is due to the fact that `core` requires `Debug` implementations for all types, but `std` does not (yet) require this. Even though this was ultimately changed to be a wrapper over the original type, not a re-export, I decided to keep the `Debug` implementation so it remains useful.
Like we do with ordinary strings, the tests are still located entirely in `alloc`, rather than splitting them into `core` and `alloc`.
----
Reviewer note: for ease of review, this is split into three commits:
1. Moving the original files into their new "locations"
2. Actually modifying the code to compile.
3. Removing aesthetic changes that were made so that the diff for commit 2 was readable.
You can review commits 1 and 3 to verify these claims, but commit 2 contains the majority of the changes you should care about.
----
API changes: `impl Debug for std::os::windows::ffi::EncodeWide`
This commit adds an empty stub for the function
for QNX 8 targets. This symbol is required by the unwinder but is
not present, causing a linking failure when building with the
standard library.
Address review feedback: use whitelist for QNX versions
Ralf Jung