Message361801
Victor, sorry.
I spent about 4 days in total trying to figure out what is the reason for hanging.
I see that the signal handler is installed; but, in a rare situation, it is not called when the process is killed.
If I reduce an amount of code executed in a signal handler by spawning a auxiliary thread, setting threading. Event in sighandler, and doing all dirty work in this secondary thread -- the problem goes away.
I'm not happy with this "patch" because it looks like reducing a change of the race without the elimination of the root cause.
I've mastered a quick hack to skip the problematic test on a hang, it may unblock other developers.
I want to spend more time on the issue detection. Unfortunately, it is not easy because of non-determenistic problem source. |
|
Date |
User |
Action |
Args |
2020-02-11 11:41:21 | asvetlov | set | recipients:
+ asvetlov, vstinner, yselivanov, aeros |
2020-02-11 11:41:21 | asvetlov | set | messageid: <1581421281.18.0.800907757907.issue38323@roundup.psfhosted.org> |
2020-02-11 11:41:21 | asvetlov | link | issue38323 messages |
2020-02-11 11:41:20 | asvetlov | create | |
|