Processing of SIGSTOP could result in a bogus stop notification from NPL if there was a pending
notification waiting for a thread to stop. The notification would fire even though the thread
would be resumed immediately. I fix this by using a lower-level resume operation which does not
check for stop notifications. The same approach was used for other occurences of this problem
(e.g. D9696).
Details
Details
Diff Detail
Diff Detail
Event Timeline
Comment Actions
It seems that SIGSTOP processing was broken to begin with https://llvm.org/bugs/show_bug.cgi?id=23264 so this will not help. I will take a step back and see what is the underlying problem.