Stacked Borrows: make scalar field retagging the default
I think it is time to finally close this soundness gap. Any objections? :)
Unfortunately the latest released versions of hashbrown and scopeguard can fail under full field retagging. The fixes have landed in the git repos but have not been released yet. I don't know if scalar field retagging as enabled by this PR is sufficient to cause problems with these crates, but it seems likely that this would be the case -- e.g. if both `value` and `dropfn` are scalars, the entire scopeguard struct will be a `ScalarPair` and thus get field retagging.
However, given that we actually generate LLVM `noalias` for these cases, it seems prudent to inform users of this risk. They can easily set `-Zmiri-field-retag=none` to opt-out of this change.
Cc https://github.com/rust-lang/miri/issues/2528
./miri updates: toolchain, rustc-push, rustc-pull
This merges the `./rustup-toolchain` script into `./miri` as `./miri toolchain`, and adds two new commands for josh-based syncing.
r? `@rust-lang/miri`
Support timeouts with monotonic clocks even when isolation is enabled
With the deterministic monotonic clock support we now have, we can allow some synchronization primitives with timeouts even under isolation:
- Linux futex waiting (when set to the monotonic clock)
- pthread_cond_timedwait (when set to the monotonic clock)
- Windows WaitOnAddress
Unfortunately none of these exist on macOS -- the standard library always uses the system clock for timeouts on macOS, so that will still require `-Zmiri-disable-isolation`.
Implement thread parking for Windows
Cc https://github.com/rust-lang/miri/issues/2628
Based on code by `@DrMeepster.` However I adjusted `WakeByAddressSingle`: I don't think the futex value is compared *again* after the thread is woken up. I see nothing in the Windows docs indicating such a comparison, and the Linux futex does not behave like that either. So we only check the value before sleeping, same as on Linux.
test on windows-gnu target
The windows-gnu target for an open-source windows toolchain is slightly different in some low-level aspects of the standard library, such as TLS handling. So let's separately ensure that this works. (Also tests a 64bit windows target on a windows host, which we didn't have so far.)
Improve miri_print_borrow_stacks
Per post-merge review on https://github.com/rust-lang/miri/pull/2322
* `miri_print_stacks` renamed to `miri_print_borrow_stacks`
* A bit more details in docs, clarified how unstable these functions are meant to be
* Print an `unknown_bottom` if one exists
Open question: Currently `miri_get_alloc_id` gets the expected `AllocId` for `Wildcard` pointers, but for pointers with no provenance, the function reports UB and halts the interpreter. That's definitely wrong. But what _should_ we do? Is it reasonable to check if the pointer has `None` provenance and try to get an `AllocId` for its address? That still leaves us with a failure path, which in this case might be best-handled as an ICE? I'm just not sure that changing the return type of `miri_get_alloc_id` to `Option` is a win because it complicates all normal uses of this.
update Miri
I had to use a hacked version of josh to create this, so let's be careful with merging this and maybe wait a bit to see if the josh issue becomes more clear. But the history looks good to me, we are not adding duplicates of rustc commits that were previously mirrored to Miri.
Also I want to add some cross-testing of Miri in x.py.
Truncate thread names on Linux and Apple targets
These targets have system limits on the thread names, 16 and 64 bytes
respectively, and `pthread_setname_np` returns an error if the name is
longer. However, we're not in a context that can propagate errors when
we call this, and we used to implicitly truncate on Linux with `prctl`,
so now we manually truncate these names ahead of time.
r? ``````@thomcc``````
rustc: Use `unix_sigpipe` instead of `rustc_driver::set_sigpipe_handler`
This is the first (known) step towards starting to use `unix_sigpipe` in the wild. Eventually, `rustc_driver::set_sigpipe_handler` can be removed and all clients can use `unix_sigpipe` instead.
For now we just start using `unix_sigpipe` in one place: `rustc` itself.
It is easy to manually verify this change. If you remove `#[unix_sigpipe = "sig_dfl"]` and run `./x.py build` you will get an ICE when you do `./build/x86_64-unknown-linux-gnu/stage1/bin/rustc --help | false`. Add back `#[unix_sigpipe = "sig_dfl"]` and the ICE disappears again.
PR that added `set_sigpipe_handler`: https://github.com/rust-lang/rust/pull/49606
Tracking issue for `unix_sigpipe`: #97889
Not sure exactly how to label this PR. Going with T-libs for now since this is a T-libs feature.
````@rustdoc```` labels +T-libs
Support raw-dylib functions being used inside inlined functions
Fixes#102714
Issue Details:
When generating the import library for `raw-dylib` symbols, we currently only use the functions and variables declared within the current crate. This works fine if all crates are static libraries or `rlib`s as the generated import library will be contained in the static library or `rlib` itself, but if a dependency is a dynamic library AND the use of a `raw-dylib` function or variable is inlined or part of a generic instantiation then the current crate won't see its dependency's import library and so linking will fail.
Fix Details:
Instead, when we generate the import library for a `dylib` or `bin` crate, we will now generate it for the symbols both for the current crate and all upstream crates. We do this in two steps so that the import library for the current crate is passed into the linker first, thus it is preferred if there are any ambiguous symbols.
