STINNER Victor added the comment:
I didn't see this bug recently, so I close the issue.
--
resolution: - fixed
status: open - closed
___
Python tracker rep...@bugs.python.org
http://bugs.python.org/issue19652
___
STINNER Victor added the comment:
asyncio_subprocess_watchdog.patch:
* add test_asyncio.watchdog.setup_watchdog(): use
faulthandler.dump_traceback_later() to kill the process if it hangs more than
'timeout' seconds (do nothing if faulthandler is not present, eg. on other
Python VMs than
Guido van Rossum added the comment:
Hmm... When did you last see the hang? We've had a variety of checkins that
might have affected this code.
If you've seen at least one traceback later than the commit below, feel free to
commit. Otherwise, let's wait.
changeset: 87088:eb42adc53923
New submission from STINNER Victor:
I didn't understand the purpose of the issue #19334: it doesn't mention the OS
nor the test. I prefer to create a more specific issue, so here you have one
specific hang.
Guido van Rossum added the comment:
The traceback is pretty useless. So is the title of the other bug report hangs
for 1 hour (which just means it hangs forever but the test runner kills it
after one hour).
We would need to run the tests with -v so at least we can pinpoint which of the
600
STINNER Victor added the comment:
Ah, it looks like the test uses a child process. The following issue has a
similar problem, we don't know that status of the child process:
http://bugs.python.org/issue19564
I proposed a patch in this issue for multiprocessing, to enable also
faulthandler in
Guido van Rossum added the comment:
I cannot help you unless you tell me which specific test is failing. But once
you have a proposed fix I will review it. Thanks for advocating for the
minority platforms!
--
___
Python tracker
STINNER Victor added the comment:
I cannot help you unless you tell me which specific test is failing.
Oh, I didn't notice that test_subprocess_send_signal() is part of a mixin:
SubprocessTestsMixin. Configure output:
checking sys/devpoll.h usability... no
checking sys/devpoll.h presence...
Guido van Rossum added the comment:
Oh, sorry, I didn't realize the name of the failing test was in the issue
title. But even that's no excuse, because it's also in the log. :-(
Fortunately the line where we're hanging is also in the log: line 291 in
selectors.py is in PollSelector. So that's