`shell-command-on-region' behaviour changed

[Leo]

Hello all,

I have noticed blank *Shell Command Output* buffer appears frequently
for quite a while. I keep deleting it since it appears in almost every
buffer switching in ido.

This is due to shell-command-on-region no longer respects its
well-documented and established behaviour of deleting that output buffer
when there's no output. The document still says so by the way.

I think the documented behaviour is better.

`shell-command-on-region' and `shell-command' have more noninteractive
uses than interactive ones. Many emacs packages authors rely on this
behaviour by ignoring the output-buffer arg to shell-command. Due to
this change packages like dired (its dired-do-shell-command) starts
leaving a blank buffer. And the doc string already says *Shell Command
Output* is shared if output buffer is unspecified.

I am unclear about what use case the changed behaviour tries to cover.
But the change will be better if it only affects non-programming use of
shell-command/shell-command-on-region, i.e. by using a different
output-buffer for interactive use.

I only use M-! or M-| for simple shell scripting and I am only
interested in the output for that moment at least for 99% of the cases.

Leo