* [Concurrency] Fix task excutor handling of default actor isolation
The task executor API did not properly account for taking the default
actor locking into account when running code on it, we just took the job
and ran it without checking with the serial executor at all, which
resulted in potential concurrent executions inside the actor --
violating actor isolation.
Here we change the TaskExecutor enqueue API to accept the "target"
serial executor, which in practice will be either generic or a specific
default actor, and coordinate with it when we perform a
runSynchronously.
The SE proposal needs to be amended to showcase this new API, however
without this change we are introducing races so we must do this before
the API is stable.
* Remove _swift_task_enqueueOnTaskExecutor as we don't use it anymore
* no need for the new protocol requirement
* remove the enqueue(_ job: UnownedJob, isolatedTo unownedSerialExecutor: UnownedSerialExecutor)
Thankfully we dont need it after all
* Don't add swift_defaultActor_enqueue_withTaskExecutor and centralize the task executor getting to enqueue()
* move around extern definitions
These APIs remained to preserve source compatibility during bringup. We do not
want them used in swift 6 mode... so deprecate in swift 5 and obsolete in swift
6.
This reverts c07aa9c425 which was done to
avoid a crash in optimnized caused by this PR:
https://github.com/apple/swift/pull/41088
Since this was almost 2 years ago, we probably don't have this issue
anymore as far as I can see other resolved issues, so try to remove the
workaround.
Resolves rdar://88711954
Setting this to Self.AsyncIterator.Failure doesn't work because
we replace it with its reduced type Self.Failure, and then if
you run a binary compiled against an older version of the stdlib
that does not declare the Failure associated type, we hit
infinite recursion while instantiating the witness table and
trying to populate the default type witness.
Instead, let's set the default to `any Error`. Associated type
inference is already smart enough to infer the type witness from
the AsyncIterator's Failure type; we don't need to rely on the
default mechanism for that.
Fixes rdar://129605725.
It could be confusing to adopters who were led to believe by the types
that they should "just" implement the sharedUnownedExecutor property,
but insead they have to implement the unownedExecutor on the specific
actor type.
Adding documentation clarify this as well as a simple test that
exercises this explicitly; We seem to have much coverage of main actor,
but not so much of custom executor global actors.
Specifically:
1. CheckedContinuation.resume now takes a sending parameter.
2. Async{Throwing,}Stream.yield takes a sending parameter.
3. withCheckedContinuation returns a transferring parameter.
Importantly due to the previous changes around mangling, this is a mangling
neutral change.
rdar://120420024
