bug#21953: Eliminate warnings in the emacs-25 release branch

[John Wiegley]

>>>>> Glenn Morris <rgm@gnu.org> writes:

> Sorry, but I have to say that I think this goal is firstly unachievable, and
> secondly won't make the release any better. In fact probably the opposite,
> by distracting from more important issues.

I've never worked on a project where getting to warning free was not both
achievable, and a good idea.

For example, at one point the Borland C++ compiler had over 4,000 warnings. As
such, every one of them was completely ignored. Later on, after much effort,
we reduced that number to zero. While doing so, we discovered that a crash bug
we'd been working for _days_ was loudly pointed out to us by the compiler in
that same set of warnings. Had the warnings been "real" to us, it would have
saved days of developer time, on several occasions. It's worth the effort.

Having enough warnings that they no longer warn of anything is pretty much the
same as turning them off. That's why I don't accept the "it's too hard"
argument. If it adds another month or two to our release cycle, so be it.
Emacs 25.1 is the time to get them under control.

Some warnings are not wise to fix, and these should be silenced; but such
decisions should be made either by class within a package, or individually.

John