Fix kill -0 please

[ale5000-git]

It would be nice to have kill -0 working even if process_architecture_matches_current fails.
Currently it say: can't kill pid 20808: Permission denied

kill -0 doesn't send any signal, isn't possible to make it working even with bitness mismatch?

[rmyorston]

I don't think any fix is necessary. POSIX says of the kill utility:

For each pid operand, the kill utility shall perform actions equivalent to the kill() function...

And of the kill() function it says:

If sig is 0 (the null signal), error checking is performed but no signal is actually sent.

So if you don't have permission to signal a process kill -0 should report the error and fail, even though no signal is sent.

And indeed, that's what I see on Linux:

$ /usr/bin/kill -0 1; echo $?
kill: sending signal to 1 failed: Operation not permitted
1
$ 

[ale5000-git]

kill -9 can kill even if there is a bitness mismatch.
I think it is more useful to use kill -0 to check if a process is active (like as in #470) as you can NOT use ps -p $pid or test -d /proc/$pid in BusyBox for Windows (and wait is blocking).

Or maybe what about -0 to check if the process is active and -00 to check if the normal kill would work?

[ale5000-git]

These are the updated POSIX pages:
https://pubs.opengroup.org/onlinepubs/9799919799/utilities/kill.html
https://pubs.opengroup.org/onlinepubs/9799919799/functions/kill.html

Quote:

The kill() function shall send a signal to a process or a group of processes specified by pid.
The signal to be sent is specified by sig and is either one from the list given in [<signal.h>](https://pubs.opengroup.org/onlinepubs/9799919799/basedefs/signal.h.html) or 0.
If sig is 0 (the null signal), error checking is performed but no signal is actually sent.
The null signal can be used to check the validity of pid.

---

The most relevant part is: The null signal can be used to check the validity of pid.