In some cases we might cause a StreamWriter to stay alive even when the
application has dropped all references to it. This prevents us from
doing automatical cleanup, and complaining that the StreamWriter wasn't
properly closed.
Fortunately, the extra reference was never actually used for anything so
we can just drop it.
Add test annotations required to run the test suite on iOS (PEP 730).
The majority of the change involve annotating tests that use subprocess,
but are skipped on Emscripten/WASI for other reasons, and including
iOS/tvOS/watchOS under the same umbrella as macOS/darwin checks.
`is_apple` and `is_apple_mobile` test helpers have been added to
identify *any* Apple platform, and "any Apple platform except macOS",
respectively.
If other exception was raised during exiting an expired
asyncio.timeout() block, insert TimeoutError in the exception context
just above the CancelledError.
When an `StopIteration` raises into `asyncio.Future`, this will cause
a thread to hang. This commit address this by not raising an exception
and silently transforming the `StopIteration` with a `RuntimeError`,
which the caller can reconstruct from `fut.exception().__cause__`
When wrapped, `_SSLProtocolTransport._force_close(exc)` is called just like in the unwrapped scenario `_SelectorTransport._force_close(exc)` or `_ProactorBasePipeTransport._force_close(exc)` would be called, except here the exception needs to be passed through the `SSLProtocol._abort()` method, which didn't accept an exception object.
This commit ensures that this path works, in the same way that the uvloop implementation of SSLProto passes on the exception (on which the current implementation of SSLProto is based).
Fix test_unhandled_exceptions() of test_asyncio.test_streams: break
explicitly a reference cycle.
Fix also StreamTests.tearDown(): the loop must not be closed
explicitly, but using set_event_loop() which takes care of shutting
down the executor with executor.shutdown(wait=True).
BaseEventLoop.close() calls executor.shutdown(wait=False).
* Try to fix asyncio.Server.wait_closed() again
I identified the condition that `wait_closed()` is intended
to wait for: the server is closed *and* there are no more
active connections.
When this condition first becomes true, `_wakeup()` is called
(either from `close()` or from `_detach()`) and it sets `_waiters`
to `None`. So we just check for `self._waiters is None`; if it's
not `None`, we know we have to wait, and do so.
A problem was that the new test introduced in 3.12 explicitly
tested that `wait_closed()` returns immediately when the server
is *not* closed but there are currently no active connections.
This was a mistake (probably a misunderstanding of the intended
semantics). I've fixed the test, and added a separate test that
checks exactly for this scenario.
I also fixed an oddity where in `_wakeup()` the result of the
waiter was set to the waiter itself. This result is not used
anywhere and I changed this to `None`, to avoid a GC cycle.
* Update Lib/asyncio/base_events.py
---------
Co-authored-by: Carol Willing <carolcode@willingconsulting.com>
- `ThreadedChildWatcher.close()` is now *officially* a no-op; `_join_threads()` never did anything.
- Threads created by that class are now named `asyncio-waitpid-NNN`.
- `test.test_asyncio.utils.TestCase.close_loop()` now waits for the child watcher's threads, but not forever; if a thread hangs, it raises `RuntimeError`.
asyncio.TaskGroup and asyncio.Timeout classes now raise proper RuntimeError
if they are improperly used.
* When they are used without entering the context manager.
* When they are used after finishing.
* When the context manager is entered more than once (simultaneously or
sequentially).
* If there is no current task when entering the context manager.
They now remain in a consistent state after an exception is thrown,
so subsequent operations can be performed correctly (if they are allowed).
Co-authored-by: James Hilton-Balfe <gobot1234yt@gmail.com>
Fix test_asyncio timeouts: don't measure the maximum duration, a test
should not measure a CI performance. Only measure the minimum
duration when a task has a timeout or delay. Add CLOCK_RES to
test_asyncio.utils.
Expect the test to be "short" but don't measure the exact performance
of the CI. SHORT_TIMEOUT is about 30 seconds whereas the cancelled
coroutine takes around 1 hour.
Replace harcoded sleep of 500 ms with synchronization using a pipe.
Fix also Process._feed_stdin(): catch also BrokenPipeError on
stdin.write(input), not only on stdin.drain().
Sometimes the child_handled event was missing because either
the child quits before it gets a chance to handle the signal,
or the parent asserts before the event notification is
delivered via IPC. Synchronize explicitly to avoid this.
SubprocessProtocol process_exited() method can be called before
pipe_data_received() and pipe_connection_lost() methods. Document it
and adapt the test for that.
Revert commit 282edd7b2a.
_child_watcher_callback() calls immediately _process_exited(): don't
add an additional delay with call_soon(). The reverted change didn't
make _process_exited() more determistic: it can still be called
before pipe_connection_lost() for example.
Co-authored-by: Davide Rizzo <sorcio@gmail.com>
test_asyncio and test_compileall now clean up multiprocessing by
calling multiprocessing _cleanup_tests(): explicitly clean up
resources and stop background processes like the resource tracker.
The default task name is "Task-<counter>" (if no name is passed in during Task creation).
This is initialized in `Task.__init__` (C impl) using string formatting, which can be quite slow.
Actually using the task name in real world code is not very common, so this is wasted init.
Let's defer this string formatting to the first time the name is read (in `get_name` impl),
so we don't need to pay the string formatting cost if the task name is never read.
We don't change the order in which tasks are assigned numbers (if they are) --
the number is set on task creation, as a PyLong instead of a formatted string.
Co-authored-by: Łukasz Langa <lukasz@langa.pl>
subprocess's communicate(None) closes stdin of the child process, after
sending no (extra) data. Make asyncio variant do the same.
This fixes issues with processes that waits for EOF on stdin before
continuing.
Also use `raise TimeOut from <CancelledError instance>` so that the CancelledError is set
in the `__cause__` field rather than in the `__context__` field.
Co-authored-by: Guido van Rossum <gvanrossum@gmail.com>
Co-authored-by: Alex Waygood <Alex.Waygood@Gmail.com>
This starts the process. Users who don't specify their own start method
and use the default on platforms where it is 'fork' will see a
DeprecationWarning upon multiprocessing.Pool() construction or upon
multiprocessing.Process.start() or concurrent.futures.ProcessPool use.
See the related issue and documentation within this change for details.
Partially revert changes made in GH-93453.
asyncio.DefaultEventLoopPolicy.get_event_loop() now emits a
DeprecationWarning and creates and sets a new event loop instead of
raising a RuntimeError if there is no current event loop set.
Co-authored-by: Guido van Rossum <gvanrossum@gmail.com>
asyncio.get_event_loop() now always return either running event loop or
the result of get_event_loop_policy().get_event_loop() call. The latter
should now raise an RuntimeError if no current event loop was set
instead of creating and setting a new event loop.
It affects also a number of asyncio functions and constructors which
call get_event_loop() implicitly: ensure_future(), shield(), gather(),
etc.
DeprecationWarning is no longer emitted if there is no running event loop but
the current event loop was set.
Co-authored-by: Łukasz Langa <lukasz@langa.pl>
Such buildbots (at the time of writing, only "AMD64 RHEL8 FIPS Only Blake2 Builtin Hash 3.x") cannot use multiprocessing with a fork server, so just skip the test there.
It was a no-op when used as recommended (after close()).
I had to debug one test (test__sock_sendfile_native_failure) --
the cleanup sequence for the test fixture was botched.
Hopefully that's not a portend of problems in user code --
this has never worked so people may well be doing this wrong. :-(
Co-authored-by: kumar aditya
Alas, warnings.catch_warnings() has global scope, not thread scope, so this is still not perfect, but it reduces the time during which warnings are ignored. Better solution welcome.
This is the next step for deprecating child watchers.
Until we've removed the API completely we have to use it, so this PR is mostly suppressing a lot of warnings when using the API internally.
Once the child watcher API is totally removed, the two child watcher implementations we actually use and need (Pidfd and Thread) will be turned into internal helpers.
There is no reason for this watcher to be attached to any particular loop.
This should make it safe to use regardless of the lifetime of the event loop running in the main thread
(relative to other loops).
Co-authored-by: Yury Selivanov <yury@edgedb.com>
Co-authored-by: Jelle Zijlstra <jelle.zijlstra@gmail.com>
This PR reverts gh-93369 and gh-97896 because they've made asyncio tests unstable. After these PRs were merged, random GitHub action jobs of random commits started to fail unrelated tests and test framework methods.
The reverting is necessary because such shrapnel failures are a symptom of some underlying bug that must be found and fixed first.
I had a hope that it's a server overload because we already have extremely rare disc access errors. However, one and a half day passed, and the failures continue to emerge both in PRs and commits.
Affected issue: gh-93357.
First reported in https://github.com/python/cpython/pull/97940#issuecomment-1270004134.
* Revert "gh-93357: Port test cases to IsolatedAsyncioTestCase, part 2 (#97896)"
This reverts commit 09aea94d29.
* Revert "gh-93357: Start porting asyncio server test cases to IsolatedAsyncioTestCase (#93369)"
This reverts commit ce8fc186ac.
The main problem was that an unluckily timed task cancellation could cause
the semaphore to be stuck. There were also doubts about strict FIFO ordering
of tasks allowed to pass.
The Semaphore implementation was rewritten to be more similar to Lock.
Many tests for edge cases (including cancellation) were added.
This reverts commit 0587810698.
Reason: This broke buildbots (some warnings added by that commit are turned to errors in the SSL buildbot).
Repro: ./python Lib/test/ssltests.py
Warn on loop initialization, when setting the wakeup fd disturbs a previously set wakeup fd, and on loop closing, when upon resetting the wakeup fd, we find it has been changed by someone else.
When a task catches CancelledError and raises some other error,
the other error should not silently be suppressed.
Any scenario where a task crashes in cleanup upon cancellation
will now result in an ExceptionGroup wrapping the crash(es)
instead of propagating CancelledError and ignoring the side errors.
NOTE: This represents a change in behavior (hence the need to
change several tests). But it is only an edge case.
Co-authored-by: Thomas Grainger <tagrain@gmail.com>
Once the task group is shutting down, it should not be possible to create a new task.
Here "shutting down" means `self._aborting` is set, indicating that at least one task
has failed and we have cancelled all others.
Co-authored-by: Łukasz Langa <lukasz@langa.pl>
The inspect version was not working with unittest.mock.AsyncMock.
The fix introduces special-casing of AsyncMock in
`inspect.iscoroutinefunction` equivalent to the one
performed in `asyncio.iscoroutinefunction`.
Co-authored-by: Łukasz Langa <lukasz@langa.pl>
test_asyncio, test_logging, test_socket and test_socketserver now
create AF_UNIX domains in the current directory to no longer fail
with OSError("AF_UNIX path too long") if the temporary directory (the
TMPDIR environment variable) is too long.
Modify the following tests to use create_unix_domain_name():
* test_asyncio
* test_logging
* test_socket
* test_socketserver
test_asyncio.utils: remove unused time import.
run_until() of test.test_asyncio.utils now uses an exponential sleep
delay (max: 1 second), rather than a fixed delay of 1 ms. Similar
design than support.sleeping_retry() wait strategy that applies
exponential backoff.
* Replace time.sleep(0.010) with sleeping_retry() to
use an exponential sleep.
* support.wait_process(): reuse sleeping_retry().
* _test_eintr: remove unused variables.
On slow buildbot workers, some test_ssl tests fail randomly because
of short timeout (30 seconds). Use support.LONG_TIMEOUT instead which
is longer and also adjusted (by regrtest --timeout option) on
buildbot workers known to be slow.
One more thing that can help prevent people from using `preexec_fn`.
Also adds conditional skips to two tests exposing ASAN flakiness on the Ubuntu 20.04 Address Sanitizer Github CI system. When that build is run on more modern systems the "problem" does not show up. It seems ASAN implementation related.
Co-authored-by: Zackery Spytz <zspytz@gmail.com>
Co-authored-by: Serhiy Storchaka <storchaka@gmail.com>
The existing event loop `start_tls()` method is not sufficient for
connections using the streams API. The existing StreamReader works
because the new transport passes received data to the original protocol.
The StreamWriter must then write data to the new transport, and the
StreamReaderProtocol must be updated to close the new transport
correctly.
The new StreamWriter `start_tls()` updates itself and the reader
protocol to the new SSL transport.
Co-authored-by: Ian Good <icgood@gmail.com>
Pierre Ossman (ThinLinc team)