`ExecutorHooks.h` is now nothing to do with hooks, so rename it. Also
there are some additional functions it should declare, and a couple of
places where we've slightly messed up the boundary, for instance
`swift_task_asyncMainDrainQueue` was defined in `Task.cpp` rather than
in the executor implementations, which is wrong, so fix that too.
`CooperativeGlobalExecutor.cpp` now builds against the interface from
`ExecutorImpl.h`, rather than including the all the concurrency headers.
rdar://135380149
C++ executor implementations were `#include`ed into `GlobalExecutor.cpp`,
which makes it difficult to replace the global executor when using the
Embedded Concurrency library. Refactor things so that they build into
separate objects, which means replacing them is just a matter of writing
the relevant functions yourself.
rdar://135380149
`stdint.h` should be picked up from `clang/lib/Headers`, but is instead
being picked up by the system libc++. There's no C++ stdlib being used
here, so just exclude it.
Resolves rdar://134584685.
After we've enqueued a job, another thread may run it and destroy it. Don't try to get the job's task executor preference when we try to schedule it. Instead, get the task executor preference before enqueueing the job, then use that preference when scheduling if necessary. Since getting the executor preference is potentially somewhat expensive (we need to search the status records for an executor preference record), only do this if the pre-compare-and-swap states look like they'll need it.
rdar://136281920
This makes ManagedBuffer available and usable in Embedded Swift, by:
- Removing an internal consistency check from ManagedBuffer that relies on metatypes.
- Making the .create() API transparent (to hoist the metatype to the callee).
- Adding a AllocRefDynamicInst simplification to convert `alloc_ref_dynamic` to `alloc_ref`, which removes a metatype use.
- Adding tests for the above.
This plugs a hole where we failed to recognize a CF type when it
appeared as the payload of an enum stored as a property. Curiously,
RemoteMirror is able to reflect this when the enum appears by itself,
just not when it's stored as a property.
The simplest fix is to hook into the TypeInfo calculation which
computes a TypeInfo (basically, the tree of fields) from a TypeRef
(basically, the name of the type, including generic context).
Specifically, we sometimes end up here with a "type alias" that
none of the lookup support seems to be able to handle. Fortunately,
these aliases demangle into a pretty predictable shape, so this
just pattern-matches the specific demangle tree shape to recognize
these as "type aliases in the `__C` module whose name starts with `CF`
and ends with `Ref`".
Resolves rdar://82465109
- Use the spelling "canceled" per Apple Style Guide
- Use code voice for symbol name
- Use contractions
- Avoid parenthesis for asides
- Change "it" to "that function" to reduce ambiguity
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
These changes fix minor issues that cause compilation errors with our
C++ toolchain's stricter default settings:
* `Exception.cpp`: The return type of `__cxa_begin_catch` should be
`void *`, not `void` (see https://libcxxabi.llvm.org/spec.html).
* `Leaks.mm`: `''` is not a valid C char literal; use `'\0'` instead.
