SIGSTOP, job control, and etc.

[Jerre Bowen]

spencer at eecs.umich.edu (Spencer W. Thomas) writes:

> Actually, the rules are not nearly as complicated as you imply.  If
> you are not root, you can send signals to any process with the same
> uid.  If you are root, you can send signals to any process.  You can
> send SIGCONT to any process which is a direct descendent of your
> process (in addition to the above rule).

This is correct and stated more clearly than my explanation.

> Process groups have nothing to do with it (unless you use the killpg
> function).

Yep.  You're right.  I'm eating crow.

> If an orphan (a process whose parent is init (process 1)) receives a
> SIGTSTP, it will be killed instead.  However, you can send SIGSTOP to
> an orphan and restart it later with SIGCONT [[Oops -- not true on the
> Iris.  It is true on Suns and 4BSD systems.]]

This is a bug in IRIX and will be fixed in a future release.  Then receipt
of SIG{TSTP,TTIN,TTOU} by a process whose parent is init will kill the
process, but receipt of SIGSTOP by the process will not.

		Jerre Bowen
		bowen at sgi.com