MagicStack / uvloop

0.29.24 is needed to compile properly under Python 3.10

Don't start the idle handler in other threads or signal handlers,
leaving the job to `_on_wake()`.

Co-authored-by: hexin02 <hexin02@megvii.com>

test_shutdown_timeout_handler_not_set() might have a race between the
SHUT_WR message and `eof` asyncio event. This fix changes to use the
eof_received() callback triggered by SHUT_WR. Refs #412 2nd issue.

The waiter given to SSLProtocol should be woke up before the first
data callback, especially for `start_tls()` where the user protocol's
`connection_made()` won't be called and the waiter wakeup is the only
time the user have access to the new SSLTransport for the first time.
The user may want to e.g. check ALPN before handling the first data,
it's better that uvloop doesn't force the user to check this by
themselves.

Because `context.run()` doesn't hold reference to the callable, when
e.g. the protocol is written in Cython, the callbacks were not
guaranteed to hold the protocol reference. This PR fixes the issue by
explicitly add a reference before `context.run()` calls.

Refs edgedb/edgedb#2222

Refs actions/runner#781, add LD_LIBRARY_PATH manually here for now
before an upstream fix is added.

* Add typings and mypy test case
* Update flake8
* Rename test class

Co-authored-by: Fantix King <fantix.king@gmail.com>

Use _SyncSocketWriterFuture for cancellation, so that we could remove
the writer directly in the callback.

Fixes #378
Closes #389

This aligns uvloop with the same behavior in asyncio - when stdin,
stdout or stderr is None, the subprocess will use FD 0, 1 or 2 now
instead of sys.stdin, sys.stdout or sys.stderr.

Fixes #136

…yboardInterrupt` exceptions (#337)

* Don't stop the loop in run_until_complete on SystemExit & KeyboardInterrupt
* Add unit test (based on CPython's equivalent)
* Fix the test to resume its coverage.

Co-authored-by: Victor Stinner <victor.stinner@gmail.com>
Co-authored-by: Fantix King <fantix.king@gmail.com>

Regression test for
#363

This is a combined fix to correct contexts from which protocal callbacks
are invoked. In short, callbacks like data_received() should always be
invoked from consistent contexts which are copied from the context where
the underlying UVHandle is created or started.

The new test case covers also asyncio, but skipping the failing ones.

Co-authored-by: Yury Selivanov <yury@edgedb.com>

Fixes #381

This is mostly a copy of asyncpg/release workflow

To make it visible in https://pypi.org/project/uvloop/

Fixes: #318.

…ython 3.9

* Adds loop.shutdown_default_executor(), a new coroutine method added to
  asyncio in 3.9 for safe finalization of the event loop's ThreadPoolExecutor.