The multiprocessing Server class now explicitly catchs SystemExit and
closes the client connection in this case. It happens when the
Server.serve_client() method reachs the end of file (EOF).
In the case of multiprocessing.synchronize() being missing, the
test_concurrent_futures test suite now skips only the tests that
require multiprocessing.synchronize().
Validate that multiprocessing.synchronize exists as part of
_check_system_limits(), allowing ProcessPoolExecutor to raise
NotImplementedError during __init__, rather than crashing with
ImportError during __init__ when creating a lock imported from
multiprocessing.synchronize.
Use _check_system_limits() to disable tests of
ProcessPoolExecutor on systems without multiprocessing.synchronize.
Running the test suite without multiprocessing.synchronize reveals
that Lib/compileall.py crashes when it uses a ProcessPoolExecutor.
Therefore, change Lib/compileall.py to call _check_system_limits()
before creating the ProcessPoolExecutor.
Note that both Lib/compileall.py and Lib/test/test_compileall.py
were attempting to sanity-check ProcessPoolExecutor by expecting
ImportError. In multiprocessing.resource_tracker, sem_unlink() is also absent
on platforms where POSIX semaphores aren't available. Avoid using
sem_unlink() if it, too, does not exist.
Co-authored-by: Pablo Galindo <Pablogsal@gmail.com>
Avoid linear runtime of ShareableList.__getitem__ and
ShareableList.__setitem__ by storing running allocated bytes in
ShareableList._allocated_bytes instead of the number of bytes for
a particular stored item.
Co-authored-by: Antoine Pitrou <antoine@python.org>
Add os.waitstatus_to_exitcode() function to convert a wait status to an
exitcode.
Suggest waitstatus_to_exitcode() usage in the documentation when
appropriate.
Use waitstatus_to_exitcode() in:
* multiprocessing, os, subprocess and _bootsubprocess modules;
* test.support.wait_process();
* setup.py: run_command();
* and many tests.
When the pull is not used via the context manager or terminate() is called, there is a system in multiprocessing.util that handles finalization of all pools via an atexit handler (the Finalize) class. This class registers the _terminate_pool handler in the registry of finalizers of the module, and that registry is called on interpreter exit via _exit_function. The problem is that the "happy" path with the context manager or manual call to finalize() does some extra steps that _terminate_pool does not. The step that is not executed when the atexit() handler calls _terminate_pool is pinging the _change_notifier queue to unblock the maintenance threads.
This commit moves the notification to the _terminate_pool function so is called from both code paths.
Co-authored-by: Pablo Galindo <Pablogsal@gmail.com>
Multiprocessing and concurrent.futures tests now stop the resource
tracker process when tests complete.
Add ResourceTracker._stop() method to
multiprocessing.resource_tracker.
Add _cleanup_tests() helper function to multiprocessing.util: share
code between multiprocessing and concurrent.futures tests.
Capturing exceptions into names can lead to reference cycles though the __traceback__ attribute of the exceptions in some obscure cases that have been reported previously and fixed individually. As these variables are not used anyway, we can remove the binding to reduce the chances of creating reference cycles.
See for example GH-13135
This PR implements a fix for `multiprocessing.Process` objects; the error occurs when Processes are created using either `fork` or `forkserver` as the `start_method`.
In these instances, the `MainThread` of the newly created `Process` object retains all attributes from its parent's `MainThread` object, including the `native_id` attribute. The resulting behavior is such that the new process' `MainThread` captures an incorrect/outdated `native_id` (the parent's instead of its own).
This change forces the Process object to update its `native_id` attribute during the bootstrap process.
cc @vstinner
https://bugs.python.org/issue38707
Automerge-Triggered-By: @pitrou
* subprocess: Add user, group and extra_groups paremeters to subprocess.Popen
This adds a `user` parameter to the Popen constructor that will call
setreuid() in the child before calling exec(). This allows processes
running as root to safely drop privileges before running the subprocess
without having to use a preexec_fn.
This also adds a `group` parameter that will call setregid() in
the child process before calling exec().
Finally an `extra_groups` parameter was added that will call
setgroups() to set the supplimental groups.
multiprocessing tests now stop the ForkServer instance if it's
running: close the "alive" file descriptor to ask the server to stop
and then remove its UNIX address.
Fix multiprocessing.util.get_temp_dir() finalizer: clear also the
'tempdir' configuration of the current process, so next call to
get_temp_dir() will create a new temporary directory, rather than
reusing the removed temporary directory.
sys._base_executable is now always defined on all platforms, and can be overridden through configuration.
Also adds test.support.PythonSymlink to encapsulate platform-specific logic for symlinking sys.executable
The multiprocessing.resource_tracker replaces the multiprocessing.semaphore_tracker module. Other than semaphores, resource_tracker also tracks shared_memory segments. Patch by Pierre Glaser.
Deprecated passing the following arguments as keyword arguments:
- "func" in functools.partialmethod(), weakref.finalize(),
profile.Profile.runcall(), cProfile.Profile.runcall(),
bdb.Bdb.runcall(), trace.Trace.runfunc() and
curses.wrapper().
- "function" in unittest.addModuleCleanup() and
unittest.TestCase.addCleanup().
- "fn" in the submit() method of concurrent.futures.ThreadPoolExecutor
and concurrent.futures.ProcessPoolExecutor.
- "callback" in contextlib.ExitStack.callback(),
contextlib.AsyncExitStack.callback() and
contextlib.AsyncExitStack.push_async_callback().
- "c" and "typeid" in the create() method of multiprocessing.managers.Server
and multiprocessing.managers.SharedMemoryServer.
- "obj" in weakref.finalize().
Also allowed to pass arbitrary keyword arguments (even "self" and "func")
if the above arguments are passed as positional argument.
* bpo-35493: Use Process.sentinel instead of sleeping for polling worker status in multiprocessing.Pool
* Use self-pipe pattern to avoid polling for changes
* Refactor some variable names and add comments
* Restore timeout and poll
* Use reader object only on wait()
* Recompute worker sentinels every time
* Remove timeout and use change notifier
* Refactor some methods to be overloaded by the ThreadPool, document the cache class and fix typos
* Added tests for shared_memory submodule.
* Added tests for ShareableList.
* Fix bug in allocationn size during creation of empty ShareableList illuminated by existing test run on Linux.
* Initial set of docs for shared_memory module.
* Added docs for ShareableList, added doctree entry for shared_memory submodule, name refactoring for greater clarity.
* Added examples to SharedMemoryManager docs, for ease of documentation switched away from exclusively registered functions to some explicit methods on SharedMemoryManager.
* Wording tweaks to docs.
* Fix test failures on Windows.
* Added tests around SharedMemoryManager.
* Documentation tweaks.
* Fix inappropriate test on Windows.
* Further documentation tweaks.
* Fix bare exception.
* Removed __copyright__.
* Fixed typo in doc, removed comment.
* Updated SharedMemoryManager preliminary tests to reflect change of not supporting all registered functions on SyncManager.
* Added Sphinx doctest run controls.
* CloseHandle should be in a finally block in case MapViewOfFile fails.
* Missed opportunity to use with statement.
* Switch to self.addCleanup to spare long try/finally blocks and save one indentation, change to use decorator to skip test instead.
* Simplify the posixshmem extension module.
Provide shm_open() and shm_unlink() functions. Move other
functionality into the shared_memory.py module.
* Added to doc around size parameter of SharedMemory.
* Changed PosixSharedMemory.size to use os.fstat.
* Change SharedMemory.buf to a read-only property as well as NamedSharedMemory.size.
* Marked as provisional per PEP411 in docstring.
* Changed SharedMemoryTracker to be private.
* Removed registered Proxy Objects from SharedMemoryManager.
* Removed shareable_wrap().
* Removed shareable_wrap() and dangling references to it.
* For consistency added __reduce__ to key classes.
* Fix for potential race condition on Windows for O_CREX.
* Remove unused imports.
* Update access to kernel32 on Windows per feedback from eryksun.
* Moved kernel32 calls to _winapi.
* Removed ShareableList.copy as redundant.
* Changes to _winapi use from eryksun feedback.
* Adopt simpler SharedMemory API, collapsing PosixSharedMemory and WindowsNamedSharedMemory into one.
* Fix missing docstring on class, add test for ignoring size when attaching.
* Moved SharedMemoryManager to managers module, tweak to fragile test.
* Tweak to exception in OpenFileMapping suggested by eryksun.
* Mark a few dangling bits as private as suggested by Giampaolo.
Changes in this commit:
1. Use a _strong_ reference between the Pool and associated iterators
2. Rework PR #8450 to eliminate a cycle in the Pool.
There is no test in this commit because any test that automatically tests this behaviour needs to eliminate the pool before joining the pool to check that the pool object is garbaged collected/does not hang. But doing this will potentially leak threads and processes (see https://bugs.python.org/issue35413).
Fix a reference issue inside multiprocessing.Pool that caused the pool to remain alive if it was deleted without being closed or terminated explicitly.
After some failures in AMD64 FreeBSD CURRENT Debug 3.x buildbots
regarding tests in test_multiprocessing_spawn and after examining
similar failures in test_socket, some errors in the calculation of
ancillary data buffers were found in multiprocessing.reduction.
CMSG_LEN() can often be used as the buffer size for recvmsg() to
receive a single item of ancillary data, but RFC 3542 requires portable
applications to use CMSG_SPACE() and thus include space for padding,
even when the item will be the last in the buffer.
The failures we experience are due to the usage of CMSG_LEN() instead of
CMSG_SPACE().
Fail `test_semaphore_tracker_sigint` if no warnings are expected and one is received.
Fix race condition when the child receives SIGINT before it can register signal handlers for it.
The race condition occurs when the parent calls
`_semaphore_tracker.ensure_running()` (which in turn spawns the
semaphore_tracker using `_posixsubprocess.fork_exec`), the child
registers the signal handlers and the parent tries to kill the child.
What seem to happen is that in some slow systems, the parent sends the
signal to kill the child before the child protects against the signal.
Fix a race condition in Popen of
multiprocessing.popen_spawn_win32. The child process now duplicates
the read end of pipe instead of "stealing" it.
Previously, the read end of pipe was "stolen" by the child process,
but it leaked a handle if the child process had been terminated
before it could steal the handle from the parent process.
In some conditions the standard streams will be None or closed in the child process (for example if using "pythonw" instead of "python" on Windows). Avoid failing with a non-0 exit code in those conditions.
Report and initial patch by poxthegreat.
Fix deadlocks in :class:`concurrent.futures.ProcessPoolExecutor` when task arguments or results cause pickling or unpickling errors.
This should make sure that calls to the :class:`ProcessPoolExecutor` API always eventually return.
* bpo-31310: multiprocessing's semaphore tracker should be launched again if crashed
* Avoid mucking with process state in test.
Add a warning if the semaphore process died, as semaphores may then be leaked.
* Add NEWS entry
* bpo-31308: If multiprocessing's forkserver dies, launch it again when necessary.
* Fix test on Windows
* Add NEWS entry
* Adopt a different approach: ignore SIGINT and SIGTERM, as in semaphore tracker.
* Fix comment
* Make sure the test doesn't muck with process state
* Also test previously-started processes
* Update 2017-08-30-17-59-36.bpo-31308.KbexyC.rst
* Avoid masking SIGTERM in forkserver. It's not necessary and causes a race condition in test_many_processes.
* Make error message more informative
Replace assertions in error-reporting code with more-informative version that doesn't cause confusion over where and what the error is.
* Additional clarification + get travis to check
* Change from SystemError to TypeError
As suggested in PR comment by @pitrou, changing from SystemError; TypeError appears appropriate.
* NEWS file installation; ACKS addition (will do my best to justify it by additional work)
* Making current AssertionErrors in multiprocessing more informative
* Blurb added re multiprocessing managers.py, queues.py cleanup
* Further multiprocessing cleanup - went through pool.py
* Fix two asserts in multiprocessing/util.py
* Most asserts in multiprocessing more informative
* Didn't save right version
* Further work on multiprocessing error messages
* Correct typo
* Correct typo v2
* Blasted colon... serves me right for trying to work on two things at once
* Simplify NEWS entry
* Update 2017-08-18-17-16-38.bpo-5001.gwnthq.rst
* Update 2017-08-18-17-16-38.bpo-5001.gwnthq.rst
OK, never mind.
* Corrected (thanks to pitrou) error messages for notify
* Remove extraneous backslash in docstring.
* Make error message more informative
Replace assertions in error-reporting code with more-informative version that doesn't cause confusion over where and what the error is.
* Additional clarification + get travis to check
* Change from SystemError to TypeError
As suggested in PR comment by @pitrou, changing from SystemError; TypeError appears appropriate.
* NEWS file installation; ACKS addition (will do my best to justify it by additional work)
multiprocessing.Process.is_alive() now removes the process from the
_children set if the process completed.
The change prevents leaking "dangling" processes.
* bpo-26732: fix too many fds in processes started with the "forkserver" method
A child process would inherit as many fds as the number of still-running children.
* Add blurb and test comment
* bpo-19896: Add typcodes 'q' and 'Q' to multiprocessing.sharedctypes. Patch by Antony Lee.
* Add NEWS entry.
* Slightly tweak NEWS entry
Make it clear this is more of a fix rather than a new feature.
* bpo-30794: added kill() method to multiprocessing.Process
* Added entries to documentation and NEWS
* Refactored test_terminate and test_kill
* Fix SIGTERM and SIGKILL being used on Windows for the tests
* Added "versionadded" marker to the documentation
* Fix trailing whitespace in doc
multiprocessing.Queue.join_thread() now waits until the thread
completes, even if the thread was started by the same process which
created the queue.
Fix the following warning which occurs randomly when running
test_handle_called_with_mp_queue of test_logging.QueueListenerTest:
Warning -- threading_cleanup() failed to cleanup -1 threads after 4 sec (count: 0, dangling: 1)
* Clear potential ref cycle between Process and Process target
Besides Process.join() not being called, this was an indirect cause of bpo-30775.
The threading module already does this.
* Add issue reference
* Fix bpo-30596: Add close() method to multiprocessing.Process
* Raise ValueError if close() is called before the Process is finished running
* Add docs
* Add NEWS blurb
* bpo-24484: Avoid race condition in multiprocessing cleanup
The finalizer registry can be mutated while inspected by multiprocessing
at process exit.
* Use test.support.start_threads()
* Add Misc/NEWS
* Fix race condition in signal wakeup in forkserver (followup to PR #1989)
There's an admittedly well-known race condition where ECHILD can arrive
just before the C function epoll_wait() and the latter wouldn't therefore
return EINTR. The solution is to use set_wakeup_fd(), which was designed
to avoid such race conditions.
* Reset wakeup fd in child
multiprocessing.Queue.get() with a timeout now polls its reader in
non-blocking mode if it succeeded to aquire the lock but the acquire
took longer than the timeout.
Co-Authored-By: Grzegorz Grzywacz <grzgrzgrz3@gmail.com>
* bpo-16500: Allow registering at-fork handlers
* Address Serhiy's comments
* Add doc for new C API
* Add doc for new Python-facing function
* Add NEWS entry + doc nit
* bpo-30414: multiprocesing.Queue._feed do not break from main loop on exc
Queue background running thread was not handling exceptions correctly.
Any exception occurred inside thread (putting unpickable object) cause
feeder to finish running. After that every message put into queue is
silently ignored.
* bpo-30414: multiprocesing.Queue._feed do not break from main loop on exc
Queue background running thread was not handling exceptions correctly.
Any exception occurred inside thread (putting unpickable object) cause
feeder to finish running. After that every message put into queue is
silently ignored.
Elisha Hollander
finefoot
Binbin
Victor Stinner
Inada Naoki
Asheesh Laroia
Vinay Sharma
Allen
idomic
Antoine Pitrou
Thomas Krennwallner
Galden
Dong-hee Na
Batuhan Taşkaya
Batuhan Taşkaya
Pablo Galindo
Christopher Hunt
Adam Meily
Jake Tesler
Gregory P. Smith
Steve Dower
Patrick McLean
Min ho Kim
Victor Stinner
Serhiy Storchaka
Thomas Moreau
Pierre Glaser
penguindustin
Davin Potts
Rémi Lapeyre
Steve Dower
Srinivas Thatiparthy (శ్రీనివాస్ తాటిపర్తి)
Alexander Buchkovsky
Julien Palard
Zackery Spytz
tzickel
Derek B. Kim
Antoine Pitrou
Allen W. Smith, Ph.D
Victor Stinner
Gareth Rees
Vitor Pereira
grzgrzgrz3
Jon Dufresne