TLS tests have been deadlocking on the OSX bots for quite some time now and this
commit is the result of the investigation into what's going on. It turns out
that a value in TLS which is being destroyed (e.g. the destructor is run) can be
reset back to the initial state **while the destructor is running** if TLS is
re-accessed.
To fix this we stop calling drop_in_place on OSX and instead move the data to a
temporary location on the stack.
The `thread::scoped` function will never be stabilized as-is and the API will
likely change significantly if it does, so this function is deprecated for
removal.
Currently the compiler has no knowledge of `#[thread_local]` which forces users
to take on two burdens of unsafety:
* The lifetime of the borrow of a `#[thread_local]` static is **not** `'static`
* Types in `static`s are required to be `Sync`
The thread-local modules mostly curb these facets of unsafety by only allowing
very limited scopes of borrows as well as allowing all types to be stored in a
thread-local key (regardless of whether they are `Sync`) through an `unsafe
impl`.
Unfortunately these measures have the consequence of being able to take the
address of the key itself and send it to another thread, allowing the same key
to be accessed from two different threads. This is clearly unsafe, and this
commit fixes this problem with the same trick used by `LocalKey`, which is to
have an indirect function call to find the address of the *current thread's*
thread local. This way the address of thread local keys can safely be sent among
threads as their lifetime truly is `'static`.
This commit will reduce the performance of cross-crate scoped thread locals as
it now requires an indirect function call, but this can likely be overcome in a
future commit.
Closes#25894
Currently the compiler has no knowledge of `#[thread_local]` which forces users
to take on two burdens of unsafety:
* The lifetime of the borrow of a `#[thread_local]` static is **not** `'static`
* Types in `static`s are required to be `Sync`
The thread-local modules mostly curb these facets of unsafety by only allowing
very limited scopes of borrows as well as allowing all types to be stored in a
thread-local key (regardless of whether they are `Sync`) through an `unsafe
impl`.
Unfortunately these measures have the consequence of being able to take the
address of the key itself and send it to another thread, allowing the same key
to be accessed from two different threads. This is clearly unsafe, and this
commit fixes this problem with the same trick used by `LocalKey`, which is to
have an indirect function call to find the address of the *current thread's*
thread local. This way the address of thread local keys can safely be sent among
threads as their lifetime truly is `'static`.
This commit will reduce the performance of cross-crate scoped thread locals as
it now requires an indirect function call, but this can likely be overcome in a
future commit.
Closes#25894
This commit adds a ./configure option called `--disable-elf-tls` which disables
ELF based TLS (that which is communicated to LLVM) on platforms which already
support it. OSX 10.6 does not support this form of TLS, and some users of Rust
need to target 10.6 and are unable to do so due to the usage of TLS. The
standard library will continue to use ELF based TLS on OSX by default (as the
officially supported platform is 10.7+), but this adds an option to compile the
standard library in a way that is compatible with 10.6.
This commit is an implementation of [RFC 1040][rfc] which is a redesign of the
currently-unstable `Duration` type. The API of the type has been scaled back to
be more conservative and it also no longer supports negative durations.
[rfc]: https://github.com/rust-lang/rfcs/blob/master/text/1040-duration-reform.md
The inner `duration` module of the `time` module has now been hidden (as
`Duration` is reexported) and the feature name for this type has changed from
`std_misc` to `duration`. All APIs accepting durations have also been audited to
take a more flavorful feature name instead of `std_misc`.
Closes#24874
This commit is an implementation of [RFC 1040][rfc] which is a redesign of the
currently-unstable `Duration` type. The API of the type has been scaled back to
be more conservative and it also no longer supports negative durations.
[rfc]: https://github.com/rust-lang/rfcs/blob/master/text/1040-duration-reform.md
The inner `duration` module of the `time` module has now been hidden (as
`Duration` is reexported) and the feature name for this type has changed from
`std_misc` to `duration`. All APIs accepting durations have also been audited to
take a more flavorful feature name instead of `std_misc`.
Closes#24874
An automated script was run against the `.rs` and `.md` files,
subsituting every occurrence of `task` with `thread`. In the `.rs`
files, only the texts in the comment blocks were affected.
These commits build on [some great work on reddit](http://www.reddit.com/r/rust/comments/33boew/weekend_experiment_link_rust_programs_against/) for adding MUSL support to the compiler. This goal of this PR is to enable a `--target x86_64-unknown-linux-musl` argument to the compiler to work A-OK. The outcome here is that there are 0 compile-time dependencies for a MUSL-targeting build *except for a linker*. Currently this also assumes that MUSL is being used for statically linked binaries so there is no support for dynamically linked binaries with MUSL.
MUSL support largely just entailed munging around with the linker and where libs are located, and the major highlights are:
* The entirety of `libc.a` is included in `liblibc.rlib` (statically included as an archive).
* The entirety of `libunwind.a` is included in `libstd.rlib` (like with liblibc).
* The target specification for MUSL passes a number of ... flavorful options! Each option is documented in the relevant commit.
* The entire test suite currently passes with MUSL as a target, except for:
* Dynamic linking tests are all ignored as it's not supported with MUSL
* Stack overflow detection is not working MUSL yet (I'm not sure why)
* There is a language change included in this PR to add a `target_env` `#[cfg]` directive. This is used to conditionally build code for only MUSL (or for linux distros not MUSL). I highly suspect that this will also be used by Windows to target MSVC instead of a MinGW-based toolchain.
To build a compiler targeting MUSL you need to follow these steps:
1. Clone the current MUSL repo from `git://git.musl-libc.org/musl`. Build this as usual and install it.
2. Clone and build LLVM's [libcxxabi](http://libcxxabi.llvm.org/) library. Only the `libunwind.a` artifact is needed. I have tried using upstream libunwind's source repo but I have not gotten unwinding to work with it unfortunately. Move `libunwind.a` adjacent to MUSL's `libc.a`
3. Configure a Rust checkout with `--target=x86_64-unknown-linux-musl --musl-root=$MUSL_ROOT` where `MUSL_ROOT` is where you installed MUSL in step 1.
I hope to improve building a copy of libunwind as it's still a little sketchy and difficult to do today, but other than that everything should "just work"! This PR is not intended to include 100% comprehensive support for MUSL, as future modifications will probably be necessary.
Inspecting the current thread's info may not always work due to the TLS value
having been destroyed (or is actively being destroyed). The code for printing
a panic message assumed, however, that it could acquire the thread's name
through this method.
Instead this commit propagates the `Option` outwards to allow the
`std::panicking` module to handle the case where the current thread isn't
present.
While it solves the immediate issue of #24313, there is still another underlying
issue of panicking destructors in thread locals will abort the process.
Closes#24313
This commit modifies the standard library and its dependencies to link correctly
when built against MUSL. This primarily ensures that the right libraries are
linked against and when they're linked against they're linked against
statically.
Changes the style guidelines regarding unit tests to recommend using a
sub-module named "tests" instead of "test" for unit tests as "test"
might clash with imports of libtest.
Much of this code hasn't been updated in quite some time and this commit does a
small audit of the functionality:
* Implementation functions now centralize all functionality on a locally defined
`Thread` type.
* The `detach` method has been removed in favor of a `Drop` implementation. This
notably fixes leaking thread handles on Windows.
* The `Thread` structure is now appropriately annotated with `Send` and `Sync`
automatically on Windows and in a custom fashion on Unix.
* The unsafety of creating a thread has been pushed out to the right boundaries
now.
Closes#24442
This commit removes all the old casting/generic traits from `std::num` that are
no longer in use by the standard library. This additionally removes the old
`strconv` module which has not seen much use in quite a long time. All generic
functionality has been supplanted with traits in the `num` crate and the
`strconv` module is supplanted with the [rust-strconv crate][rust-strconv].
[rust-strconv]: https://github.com/lifthrasiir/rust-strconv
This is a breaking change due to the removal of these deprecated crates, and the
alternative crates are listed above.
[breaking-change]
One of the parameters to the magical "register a thread-local destructor"
function is called `__dso_handle` and largely just passed along (this seems to
be what other implementations do). Currently we pass the *value* of this symbol,
but apparently the correct piece of information to pass is the *address* of the
symbol.
In a PIE binary the symbol actually contains an address to itself which is why
we've gotten away with what we're doing as long as we have. In a non-PIE binary
the symbol contains the address `NULL`, causing a segfault in the runtime
library if it keeps going.
Closes#24445
`thread::spawn` was previously restricted to closures that return `()`,
which limited the utility of joining on a spawned thread. However, there
is no reason for this restriction, and this commit allows arbitrary
return types.
Since it introduces a type parameter to `JoinHandle`, it's technically
a:
[breaking-change]
However, no code is actually expected to break.
Issue #24292 demonstrates that the `scoped` API as currently offered can
be memory-unsafe: the `JoinGuard` can be moved into a context that will
fail to execute destructors prior to the stack frame being popped (for
example, by creating an `Rc` cycle).
This commit reverts the APIs to `unstable` status while a long-term
solution is worked out.
(There are several possible ways to address this issue; it's not a
fundamental problem with the `scoped` idea, but rather an indication
that Rust doesn't currently provide a good way to ensure that
destructors are run within a particular stack frame.)
[breaking-change]
Make the structure more amenable to what rustdoc is expecting to ensure that
everything renders all nice and pretty in the output.
Closes#23705Closes#23910
This commit renames and stabilizes:
* `Condvar::wait_timeout_ms` (renamed from `wait_timeout`)
* `thread::park_timeout_ms` (renamed from `park_timeout`)
* `thread::sleep_ms` (renamed from `sleep`)
In each case, the timeout is taken as a `u32` number of milliseconds,
rather than a `Duration`.
These functions are likely to be deprecated once a stable form of
`Duration` is available, but there is little cost to having these named
variants around, and it's crucial functionality for 1.0.
[breaking-change]
r? @alexcrichton
cc @sfackler @carllerche
This commit renames and stabilizes:
* `Condvar::wait_timeout_ms` (renamed from `wait_timeout`)
* `thread::park_timeout_ms` (renamed from `park_timeout`)
* `thread::sleep_ms` (renamed from `sleep`)
In each case, the timeout is taken as a `u32` number of milliseconds,
rather than a `Duration`.
These functions are likely to be deprecated once a stable form of
`Duration` is available, but there is little cost to having these named
variants around, and it's crucial functionality for 1.0.
[breaking-change]
for `Box<FnBox()>`. I found the alias was still handy because it is
shorter than the fully written type.
This is a [breaking-change]: convert code using `Invoke` to use `FnBox`,
which is usually pretty straight-forward. Code using thunk mostly works
if you change `Thunk::new => Box::new` and `foo.invoke(arg)` to
`foo(arg)`.
This commit cleans out a large amount of deprecated APIs from the standard
library and some of the facade crates as well, updating all users in the
compiler and in tests as it goes along.
This commit stabilizes the following APIs:
* `TypeId::of` - now that it has an `Any` bound it's ready to be stable.
* `Box<Any>::downcast` - now that an inherent impl on `Box<Any>` as well as
`Box<Any+Send>` is allowed the `BoxAny` trait is removed in favor of these
inherent methods.
This is a breaking change due to the removal of the `BoxAny` trait, but
consumers can simply remove imports to fix crates.
[breaking-change]
This attribute has been deprecated in favor of #[should_panic]. This also
updates rustdoc to no longer accept the `should_fail` directive and instead
renames it to `should_panic`.
This commit provides a safe, but unstable interface for the `try` functionality
of running a closure and determining whether it panicked or not.
There are two primary reasons that this function was previously marked `unsafe`:
1. A vanilla version of this function exposes the problem of exception safety by
allowing a bare try/catch in the language. It is not clear whether this
concern should be directly tied to `unsafe` in Rust at the API level. At this
time, however, the bounds on `ffi::try` require the closure to be both
`'static` and `Send` (mirroring those of `thread::spawn`). It may be possible
to relax the bounds in the future, but for now it's the level of safety that
we're willing to commit to.
2. Panicking while panicking will leak resources by not running destructors.
Because panicking is still controlled by the standard library, safeguards
remain in place to prevent this from happening.
The new API is now called `catch_panic` and is marked as `#[unstable]` for now.
This attribute has been deprecated in favor of #[should_panic]. This also
updates rustdoc to no longer accept the `should_fail` directive and instead
renames it to `should_panic`.
