bug#44743: native-comp: confirm-exit-emacs warns about active processes when compiling

[Eli Zaretskii <eliz@gnu.org>]

> From: Stefan Kangas <stefan@marxist.se>
> Date: Sat, 21 Nov 2020 11:22:58 -0800
> Cc: 44743@debbugs.gnu.org
> 
> > FWIW, I think this should be controlled by a user option.  It is not
> > at all obvious that everyone would like compilation processes to be
> > killed automatically, people might want to wait for them to complete.
> 
> Compiling in the background should in my opinion work as transparently
> as possible.  The fact that we cache compilation results should be
> considered an implementation detail.  We don't need to shape our
> outwardly behavior by such implementation details.

I understand your opinion, but I don't think that's the only opinion
that could exist.  Caching the compiled modules can hardly be regarded
as an implementation detail when compilation takes a tangible amount
of time -- which is why we cache the results in the first place.  IOW,
if compilation is interrupted, Emacs will try to compile it again the
next time, and the code will run slower than expected.  So if this is
an implementation detail, it will be acutely obvious to users, and
they may wish to wait a bit with exiting Emacs to let the compilation
run to the end.  It is not unlike the case where you sent an email
message and want to exit Emacs before the message transmission has
ended.  Users will appreciate a degree of control in these cases.

> We could of course support what you suggest.  I'm not against it as an
> option.  But I don't think it is very important, and it would take some
> time and effort to implement and maintain.  I'm not sure that effort is
> well-spent at this point, and would rather leave it for the future.

I think interrupting compilation also comes with maintenance
head-aches, such as the temporary files left behind, incomplete .eln
files we'd need to clean up, etc.

> IOW, I think we should work on reasonable defaults first, and only add
> options in later once we are sure that we really need them.

I think the argument is about what is "reasonable" here, all the rest
is agreed upon.