Re: Signal handling after SIGCONT, PR/39814 related

[Mindaugas Rasiukevicius <rmind%netbsd.org@localhost>]

SODA Noriyuki <soda%yuruyuru.net@localhost> wrote:
> >> >However, unlike NetBSD, both Linux and (Open)Solaris do not process
> >> >SIGKILL until SIGCONT is sent.
> >> 
> >> Why do you believe so?
> 
> > Have you tried to write a program (or take one from PR) and test?
> 
> Yes, the result was:
> - the SIGKILL signal was immediately delivered to the process
> and
> - the process was immediately killed by the signal
> on both Linux and Solaris.
> 
> But the output of the (modified) program in the PR/39814 was
> "PID XXX still exists." on both Linux on Solaris.
> This is not because the process was not killed, but because
> kill(pid, 0) succeeds against zombie (i.e. defunct) processes
> on both Linux on Solaris.
> On NetBSD, kill(pid, 0) fails against zombie processes, and that
> makes the difference of the output.

OK, I did not expected kill() success on zombie processes, that explains
my confusion. On the other hand, it would be odd, if process would be
non-killable with SIGTERM.

Thanks.

-- 
Best regards,
Mindaugas