On Thu, Apr 1, 2010 at 7:19 PM, Dave Keck <davek...@gmail.com> wrote: >>> Look like a race condition. The man page says getppid() will not fail, but >>> not what happens after the parent dies. I will test this. >> >> If the parent dies, the process becomes the child of launchd. For a quick >> fix I checked for the parent PID greater than 1. >> >> A better fix, the parent process passed its PID to the child process as an >> argument. In the child process, if getppid() does not return the PID from >> the parent process, exit. > > Another problem is if the parent dies after the getppid(), and another > process starts with the same PID as the old parent, all before the > call to kevent(). In this scenario you'll now be waiting for an > unrelated process to die. This situation is likely impossible > currently with incrementing PIDs, but there's been mention of PID > randomization on darwin-dev: > > http://lists.apple.com/archives/darwin-dev/2009/Oct/msg00056.html > > I only mention this because I use this kevent method to detect a > parent's death, and have been looking for a better technique that > also doesn't require the parent's cooperation...
Call getppid, set up the kevent listener, then before you actually block in kevent, call getppid *again*. If the answer matches the first call, then you can go ahead and call kevent. If the answer doesn't match, you've hit the race condition you outline here, your parent is dead, and you can exit immediately. If the parent quits after the second getppid but before the call to kevent, you should still get the exit notification because you registered it before the parent quit. Mike _______________________________________________ Cocoa-dev mailing list (Cocoa-dev@lists.apple.com) Please do not post admin requests or moderator comments to the list. Contact the moderators at cocoa-dev-admins(at)lists.apple.com Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/cocoa-dev/archive%40mail-archive.com This email sent to arch...@mail-archive.com
