Everywhere there's a `SWIFT_MODULE_DEPENDS_LINUX Glibc`, there should be
a corresponding `SWIFT_MODULE_DEPENDS_LINUX_STATIC Musl`.
This usually won't bite us, depending on build order and parallelism, but
I hit one of these yesterday so went looking to see if there were any
others.
rdar://136208589
The Concurrency runtime calculates deadlines for scheduling itself
using `swift_get_time()`; unfortunately, on Windows that was using
`QueryPerformanceCounter()`, while Dispatch uses
`QueryInterruptTimePrecise()`. The problem with that is that the two do
not necessarily correspond *at all*. In general
`QueryPerformanceCounter()` may be using any of a number of hardware
timers depending on the machine on which we're running.
In the VM I was testing on, the two differed by 20ms, but the worst case
is that they are completely unrelated, in which case `Task.sleep()` will
wait essentially a random amount of time.
rdar://135413803
Annotate all of the `Unsafe*` types and `unsafe` functions in the standard
library (including concurrency, synchronization, etc.) as `@unsafe`. Add a
few tests to ensure that we detect uses of these types in clients that
have disabled unsafe code.
Supported older compilers don't enable this feature by default, so it can't be
omitted from the `_Concurrency` module's flags (regression from
https://github.com/swiftlang/swift/pull/74543).
Additionally, remove `@_allowFeatureSuppression(IsolatedAny)` from all
declarations. We no longer need to support compilers that don't have the
`IsolatedAny` feature, so the suppression is superfluous and the alternative
branches didn't actually build anyways. _Additionally_, the suppressible
feature logic could not handle suppressing `IsolatedAny` simultaneously with
`SendingArgsAndResults`, resulting in a broken interface because `sending` was
used outside `#if $SendingArgsAndResults` guards.
To preserve compatibility with older compilers that do not allow `IsolatedAny`
to be enabled in production compilers, use an alias experimental feature when
building the stdlib (`IsolatedAny2`).
Also, add `@_allowFeatureSuppression(IsolatedAny)` in a couple spots it was
forgotten.
Partially resolves rdar://125138945
Don't delete the OS declaration of `exit` because the concurrency shims aren't always imported, and so the shim declaration might not always be available.
Don't override the OS declaration of `exit` in the concurrency shims since we can't just delete the OS one. Instead, set up internal shims just for building Concurrency that forward declares `exit`.
This entails passing a linker flags to Apple linkers when the standard
library is not meant for inclusion in such cache.
For this to have effect on every library, propagate link flags when
building _Concurrency and Observation.
This is needed for Apple internal configurations.
Addresses rdar://120653968
Use an optional isolated parameter to this new `next(_:)` overload to
keep it on the same actor as the caller, and pass `#isolation` when
desugaring the async for..in loop. This keeps async iteration loops on
the same actor, allowing non-Sendable values to be used with many
async sequences.
Introduce a new associated type `Failure` into the two protocols involved
in async sequences, which represents the type thrown when the sequence
fails. Introduce a defaulted `_nextElement()` operations that throws
`Failure` or produces the next element of the sequence. Provide a
default implementation of `_nextElement()` in terms of `next()` that
force-cases the thrown error to the `Failure` type.
Introduce special associated type inference logic for the `Failure`
type of an `AsyncIteratorProtocol` conformance when there is no
specific _nextElement()` witness. This inference logic looks at the
witness for `next()`:
* If `next()` throws nothing, `Failure` is inferred to `Never`.
* If `next()` throws, `Failure` is inferred to `any Error`.
* If `next()` rethrows, `Failure` is inferred to `T.Failure`, where
`T` is the first type parameter with a conformance to either
`AsyncSequence` or `AsyncIteratorProtocol`.
The default implementation and the inference rule, together, allow
existing async sequences to continue working as before, and set us up
for changing the contract of the `async for` loop to use
`_nextIterator()` rather than `next()`.
Give `AsyncSequence` and `AsyncIteratorProtocol` primary associated
types for the element and failure types, which will allow them to be
used more generally with existential and opaque types.
which executor for which type of setting, is consolidated and we have a
single knob we use to determine when to use dispatch as our global
executor.
Radar-Id: rdar://problem/119416196
# Motivation
For AsyncStream we created a custom internal Deque implementation to use for the buffering. This implementation was relatively bare-bones compared to the one in swift-collections. Furthermore, it lacked some methods that we need to implement the new `AsyncStream` APIs that support producer backpressure.
# Modification
This PR copies over the Deque implementation of swift-collections and makes it internal and non-inlinable.
# Result
We now have a fully functional Deque.
As of CMake 3.25, there are now global variables `LINUX=1`, `ANDROID=1`,
etc. These conflict with expressions that used these names as unquoted
strings in positions where CMake accepts 'variable|string', for example:
- `if(sdk STREQUAL LINUX)` would fail, because `LINUX` is now defined and
expands to 1, where it would previously coerce to a string.
- `if(${sdk} STREQUAL "LINUX")` would fail if `sdk=LINUX`, because the
left-hand side expands twice.
In this patch, I looked for a number of patterns to fix up, sometimes a
little defensively:
- Quoted right-hand side of `STREQUAL` where I was confident it was
intended to be a string literal.
- Removed manual variable expansion on left-hand side of `STREQUAL`,
`MATCHES` and `IN_LIST` where I was confident it was unintended.
Fixes#65028.
Move the TSan functionality from Concurrency into Threading. Use it
in the Linux `ulock` implementation so that TSan knows about `ulock`
and will tolerate the newer `swift_once` implementation that uses it.
rdar://110665213
Have the Concurrency runtime register a struct containing pointers to all of its type descriptors with standard manglings. Then have the runtime use that struct to quickly look up those types, instead of using the standard, slower type lookup code.
rdar://109783861
Revert a few unnecessary changes have been made to the _Concurrency module
recently that make its swiftinterface un-parseable by older compilers that we
need to support.
- The `consume` keyword is not understood by older compilers, so including it
in inlinable code is a problem. It has no actual effect on lifetimes where it
was used, so just omit it.
- Don't build the _Concurrency module with -enable-experimental-feature
MoveOnly. The MoveOnly feature is now enabled by default in recent compilers,
so the flag is superfluous when building the dylib. When emitting the
interface, having the feature enabled explicitly exposes code guarded by
`$MoveOnly` to older compilers that do not accept all of the code.
Resolves rdar://108793606
We ran into an issue on minimal stdlib build configurations, because they
pull slightly older toolchains to build and test the stdlib. The noncopyable
`Job` type was adopted immediately after such types were enabled by default,
before the toolchain builds could catch up. Adding this flag to the build of
the `_Concurrency` lib should make things more robust when somewhat older
compilers are building this part of the stdlib.
part of resolving rdar://106849189
