#241 Plug potential epoll race
Merged by rharwood. Opened by simo.
simo/gssproxy epollrace  into  master

Download 241.patch

There is a slight chance that the code will race and close the epoll descriptor when another thread has already grabbed the socket and initialized it.
This seem to cause threads to get stuck waiting forever.
Not confirmed but the most likely working hypothesis in a bug related to rpc.gssd which now makes heavy use of threads.

Ignore the jenkins failure, it was transient, I restarted the build from jenkins and it succeeded, but it seem there is no report back when you restart it from there ...

There's a comment a few lines down that needs updated for this change.

I might reorder to remove these before merging.

Commit d55be9fa fixes this pull-request

Pull-Request has been merged by rharwood

Commit 0ccfd32f fixes this pull-request

Pull-Request has been merged by rharwood

Metadata