Document memory orderings of `thread::{park, unpark}`
Document `thread::park/unpark` as having acquire/release synchronization. Without that guarantee, even the example in the documentation can deadlock:
```rust
let flag = Arc::new(AtomicBool::new(false));
let t2 = thread::spawn(move || {
while !flag.load(Ordering::Acquire) {
thread::park();
}
});
flag.store(true, Ordering::Release);
t2.thread().unpark();
// t1: flag.store(true)
// t1: thread.unpark()
// t2: flag.load() == false
// t2 now parks, is immediately unblocked but never
// acquires the flag, and thus spins forever
```
Multiple calls to `unpark` should also maintain a release sequence to make sure operations released by previous `unpark`s are not lost:
```rust
let a = Arc::new(AtomicBool::new(false));
let b = Arc::new(AtomicBool::new(false));
let t2 = thread::spawn(move || {
while !a.load(Ordering::Acquire) || !b.load(Ordering::Acquire) {
thread::park();
}
});
thread::spawn(move || {
a.store(true, Ordering::Release);
t2.thread().unpark();
});
b.store(true, Ordering::Release);
t2.thread().unpark();
// t1: a.store(true)
// t1: t2.unpark()
// t3: b.store(true)
// t3: t2.unpark()
// t2 now parks, is immediately unblocked but never
// acquires the store of `a`, only the store of `b` which
// was released by the most recent unpark, and thus spins forever
```
This is of course a contrived example, but is reasonable to rely upon in real code.
Note that all implementations of park/unpark already comply with the rules, it's just undocumented.
std: only depend on dlmalloc for wasm*-unknown
It was already filtered out for emscripten, but wasi doesn't need dlmalloc
either since it reuses `unix/alloc.rs`.
io: soften ‘at most one write attempt’ requirement in io::Write::write
At the moment, documentation of std::io::Write::write indicates that
call to it ‘represents at most one attempt to write to any wrapped
object’. It seems that such wording was put there to contrast it with
pre-1.0 interface which attempted to write all the data (it has since
been changed in [RFC 517]).
However, the requirement puts unnecessary constraints and may
complicate adaptors which perform non-trivial transformations on the
data. For example, they may maintain an internal buffer which needs
to be written out before the write method accepts more data. It might
be natural to code the method such that it flushes the buffer and then
grabs another chunk of user data. With the current wording in the
documentation, the adaptor would be forced to return Ok(0).
This commit softens the wording such that implementations can choose
code structure which makes most sense for their particular use case.
While at it, elaborate on the meaning of `Ok(0)` return pointing out
that the write_all methods interprets it as an error.
[RFC 517]: https://rust-lang.github.io/rfcs/0517-io-os-reform.html
fix: get the l4re target working again
This is based on work from https://github.com/rust-lang/rust/pull/103966, addressing the review comment by `@m-ou-se` at the time and "fixing" the (probably newly) missing read_buf.
Rollup of 6 pull requests
Successful merges:
- #112352 (Fix documentation build on FreeBSD)
- #112644 (Correct types in method descriptions of `NonZero*` types)
- #112683 (fix ICE on specific malformed asm clobber_abi)
- #112707 ([rustdoc] Fix invalid handling of "going back in history" when "go to only search result" setting is enabled)
- #112719 (Replace fvdl with ffx, allow test without install)
- #112728 (Add `<meta charset="utf-8">` to `-Zdump-mir-spanview` output)
r? `@ghost`
`@rustbot` modify labels: rollup
Fix documentation build on FreeBSD
After the socket ancillary data implementation was introduced, the documentation build was broken on FreeBSD hosts, add the same workaround as for the existing implementations.
Fixes the doc build after #91793
previously it was only able to use BufWriter. This was due to a limitation in the
BufReader generics that prevented specialization. This change works around the issue
by using `where Self: Read` instead of `where I: Read`. This limits our options, e.g.
we can't access BufRead methods, but it happens to work out if we rely on some
implementation details.
Fix building libstd documentation on FreeBSD.
It fixes the following error:
```
error[E0412]: cannot find type `sockcred2` in module `libc`
--> library/std/src/os/unix/net/ancillary.rs:211:29
|
211 | pub struct SocketCred(libc::sockcred2);
| ^^^^^^^^^ not found in `libc`
```
Implement `TryFrom<&OsStr>` for `&str`
Recently when trying to work with `&OsStr` I was surprised to find this `impl` missing.
Since the `to_str` method already existed the actual implementation is fairly non-controversial, except for maybe the choice of the error type. I chose an opaque error here instead of something like `std::str::Utf8Error`, since that would already make a number of assumption about the underlying implementation of `OsStr`.
As this is a trait implementation, it is insta-stable, if I'm not mistaken?
Either way this will need an FCP.
I chose "1.64.0" as the version, since this is unlikely to land before the beta cut-off.
`@rustbot` modify labels: +T-libs-api
API Change Proposal: rust-lang/rust#99031 (accepted)
Add support for targets without unwinding in `mir-opt`, and improve `--bless` for it
The main goal of this PR is to add support for targets without unwinding support in the `mir-opt` test suite, by adding the `EMIT_MIR_FOR_EACH_PANIC_STRATEGY` comment. Similarly to 32bit vs 64bit, when that comment is present, blessed output files will have the `.panic-unwind` or `.panic-abort` suffix, and the right one will be chosen depending on the target's panic strategy.
The `EMIT_MIR_FOR_EACH_PANIC_STRATEGY` comment replaced all the `ignore-wasm32` comments in the `mir-opt` test suite, as those comments were added due to `wasm32` being a target without unwinding support. The comment was also added on other tests that were only executed on x86 but were still panic strategy dependent.
The `mir-opt` suite was then blessed, which caused a ton of churn as most of the existing output files had to be renamed and (mostly) duplicated with the abort strategy.
---
After [asking on Zulip](https://rust-lang.zulipchat.com/#narrow/stream/131828-t-compiler/topic/mir-opt.20tests.20and.20panic.3Dabort), the main concern about this change is it'd make blessing the `mir-opt` suite even harder, as you'd need to both bless it with an unwinding target and an aborting target. This exacerbated the current situation, where you'd need to bless it with a 32bit and a 64bit target already.
Because of that, this PR also makes significant enhancements to `--bless` for the `mir-opt` suite, where it will automatically bless the suite four times with different targets, while requiring minimal cross-compilation.
To handle the 32bit vs 64bit blessing, there is now an hardcoded list of target mapping between 32bit and 64bit. The goal of the list is to find a related target that will *probably* work without requiring additional cross-compilation toolchains on the system. If a mapping is found, bootstrap will bless the suite with both targets, otherwise just with the current target.
To handle the panic strategy blessing (abort vs unwind), I had to resort to what I call "synthetic targets". For each of the target we're blessing (so either the current one, or a 32bit and a 64bit depending on the previous paragraph), bootstrap will extract the JSON spec of the target and change it to include `"panic-strategy": "abort"`. It will then build the standard library with this synthetic target, and bless the `mir-opt` suite with it.
As a result of these changes, blessing the `mir-opt` suite will actually bless it two or four times with different targets, ensuring all possible variants are actually blessed.
---
This PR is best reviewed commit-by-commit.
r? `@jyn514`
cc `@saethlin` `@oli-obk`
It fixes the following error:
error[E0412]: cannot find type `sockcred2` in module `libc`
--> library/std/src/os/unix/net/ancillary.rs:211:29
|
211 | pub struct SocketCred(libc::sockcred2);
| ^^^^^^^^^ not found in `libc`
It might happen that a synthetic target name does not match one of the
hardcoded ones in std's build script, causing std to fail to build. This
commit changes the std build script avoid including the restricted-std
feature unconditionally when a synthetic target is being built.
Avoid unwind across `extern "C"` in `thread_local::fast_local`
This is a minimal fix for #112285, in case we want a simple patch that can be easily to backported if that's desirable.
*(Note: I have another broader cleanup which I've mostly omitted from here to avoid clutter, except for the `Cell` change, which isn't needed to fix UB, but simplifies safety comments).*
The only tier-1 target that this occurs on in a way that seems likely to cause problems in practice linux-gnu, although I believe some folks care about that platform somewhat 😉. I'm unsure how big of an issue this is. I've seen stuff like this behave quite badly, but there's a number of reasons to think this might actually be "fine in practice".
I've hedged my bets and assumed we'll backport this at least to beta but my feeling is that there's not enough evidence this is a problem worth backporting further than that.
### More details
This issue seems to have existed since `thread_local!`'s `const` init functionality was added. It occurs if you have a `const`-initialized thread local for a type that `needs_drop`, the drop panics, and you're on a target with support for static thread locals. In this case, we will end up defining an `extern "C"` function in the user crate rather than in libstd, and because the user crate will not have `#![feature(c_unwind)]` enabled, their panic will not be caught by an auto-inserted abort guard.
In practice, the actual situation where problems are likely[^ub] is somewhat narrower.
On most targets with static thread locals, we manage the TLS dtor list by hand (for reentrancy reasons among others). In these cases, while the users code may panic, we're calling it inside our own `extern "C"` (or `extern "system"`) function, which seems to (at least in practice) catch the panic and convert it to an abort.
However, on a few targets, most notably linux-gnu with recent glibc (but also fuchsia and redox), a tls dtor registration mechanism exists which we can actually use directly, [`__cxa_thread_atexit_impl`](https://github.com/rust-lang/rust/blob/master/library/std/src/sys/unix/thread_local_dtor.rs#L26-L36).
This is the case that seems most likely to be a cause for concern, as now we're passing a function to the system library and panicking out of it in a case where there are may not be Rust frames above it on the call stack (since it's running thread shutdown), and even if there were, it may not be prepared to handle such unwinding. If that's the case, it'd be bad.
Is it? Dunno. The fact that it's a `__cxa_*` function makes me think they probably have considered that the callback could throw but I have no evidence here and it doesn't seem to be written down anywhere, so it's just a guess. (I would not be surprised if someone comes into this thread to tell me how definitely-bad-news it is).
That said, as I said, all this is actually UB! If this isn't a "technically UB but fine in practice", but all bets are off if this is the kind of thing we are telling LLVM about.
[^ub]: This is UB so take that with a grain of salt -- I'm absolutely making assumptions about how the UB will behave "in practice" here, which is almost certainly a mistake.
After the socket ancillary data implementation was introduced, the
build was broken on FreeBSD, add the same workaround as for the
existing implementations.
