Simon Burge wrote:

> The current behaviour is to return EPERM.  It sounds like from the rest
> of your message you meant that instead of ESRCH, or did I misunderstand
> you?

You're talking about killing a process the user don't own, with the
context of the discussion being user killing a process he don't own
while not knowing if it exists or not.

In the first case EPERM is natural. In the latter, some would say
EPERM would serve as indication as to if there is a process with that
PID running or not -- kill(<pid-user-cant-see>, sig) returning EPERM
means that there *is* such a process, somewhat defeating the privacy
restriction.

While this is true, I just noted that you can find out if a PID is
running or not in other ways than sending it signals; so it doesn't
matter what you return.... however, I think we should be returning ESRCH
to be consistent with the rest of the output.

-e.

-- 
Elad Efrat
PGP Key ID: 0x666EB914