From: <ch...@su...> - 2014-01-21 13:21:25
|
Hi! > > The timeout should better be handled, if handled at all, in the > > checkpoint code. I would not care much about this case because when the > > parent is dead does not talk to the child the test will fail anyway... > > > > I didn't care about the test failure in such case. I cared about that > without alarm() the child will keep running (i.e. blocking on IO) even > after the parent is dead. I've noticed that as well, the child will continue to run and because it called setsid() the test driver (pan) will not kill it because it left it's parent process group. > But I agree that it's not the best place to handle this inside the test > case. I was thinking of adding the alarm() code into the checkpoint child code. This should fix the problem once for all. But even then this codepath happens only when something went wrong so this has not big priority... -- Cyril Hrubis ch...@su... |