Xavier de Gaye added the comment: On a heavily loaded system, there is no guarantee that the date obtained from a call to gettimeofday() in PyCOND_TIMEDWAIT(), is not in the past when pthread_cond_timedwait() is called, as there may be a context switch in between. When this happens, a Python thread cannot take the gil even if it is available.
This tentative patch monitors the load of the system and dynamically increases the value used for the switch interval when needed. With this patch on the android-21-armv7 emulator: python -m test -m test_lru_cache_threaded -v test_functools Ok python -m test -m test_deadlock -v test_importlib Ok python -m test -m test_is_alive_after_fork -v test_threading Ok The measured elapsed time is not correct in this patch when it is above 1 second. ---------- Added file: http://bugs.python.org/file42874/pycond_timedwait.patch _______________________________________ Python tracker <rep...@bugs.python.org> <http://bugs.python.org/issue26939> _______________________________________ _______________________________________________ Python-bugs-list mailing list Unsubscribe: https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com
