bug#51177: 29.0.50; stop-process on pipes

[Helmut Eller]

On Wed, Oct 13 2021, Eli Zaretskii wrote:

>> From: Helmut Eller <eller.helmut@gmail.com>
>> Date: Wed, 13 Oct 2021 11:20:01 +0200
>> 
>> 
>> I would like to request this feature: stop-process, when called with a
>> sub-process that is connected via pipes, should use delete_read_fd.  It
>> should basically do the same as for sockets.
>
> Unlike with sockets, we are talking about a real sub-process on the
> other end of the pipe, and it will now get SIGPIPE.  Are we sure this
> is OK? perhaps it will interfere with the process's cleanup when it
> receives a signal?

A valid concern, yes.

> If there's any real possibility this could change behavior, I think we
> should make such behavior optional.

Maybe we could add a pair of functions that exposes delete_read_fd and
add_process_read_fd more directly.

Helmut