bug#21132: 25.0.50; Emacs on Windows crashes evaluating x-frame-geometry in batch mode

[Eli Zaretskii <eliz@gnu.org>]

> Date: Sat, 25 Jul 2015 20:17:39 +0200
> From: martin rudalics <rudalics@gmx.at>
> CC: 21132@debbugs.gnu.org
> 
>  >> Emacs built from the latest source on Windows crashes when invoked
>  >> as follows:
>  >>
>  >>    emacs.exe -Q -batch --eval="(x-frame-geometry (selected-frame))"
>  >
>  > Thanks, fixed.
> 
> Do you have any simple guideline how to find similar instances of this
> bug?

Look for unconditional uses of FRAME_W32_WINDOW in code that is not
guaranteed to be invoked in a GUI session.  Functions exposed to Lisp
are an obvious candidate, as in this case.

> And what is the difference between !FRAME_INITIAL_P (f) and
> FRAME_W32_WINDOW (f)?

FRAME_W32_WINDOW is not a predicate.  You probably meant
FRAME_INITIAL_P and FRAME_W32_P.  The difference is the method used to
output text.

> I suppose that !FRAME_INITIAL_P (f) implies
> FRAME_W32_WINDOW (f).

No, it implies (on w32) FRAME_W32_P or FRAME_TERMCAP_P.

> So there are probably cases when we use
> FRAME_W32_WINDOW (f) and FRAME_INITIAL_P (f) holds.  Right?

If there are, we will crash.

> And do we prefer (FRAME_W32_WINDOW (f) != 0) to (!FRAME_W32_WINDOW (f))?

We prefer !FRAME_W32_P (f)

> Also I believe that I should replace FRAME_X_WINDOW by FRAME_W32_WINDOW.

Yes.

>  > I believe similar changes are needed in xfns.c and in nsfns.m, could
>  > someone with access to the relevant OSes please check and apply?
> 
> I'll look into these.

Thanks.