Re: [Emacs-diffs] master 6cd5678: Clarify compiler-pacifier in frame.c

[Paul Eggert]

Eli Zaretskii wrote:
> I think you didn't read the code of XParseGeometry (the one that is
> executed on MS-Windows) well enough, if you think GCC gave a false
> alarm.

I just now reread that code in current master and still see only a false alarm. 
If that function returns a mask where (mask & XValue) != 0, then *x must have 
been set by this statement:

  5298    if (mask & XValue)
  5299      *x = clip_to_bounds (INT_MIN, tempX, INT_MAX);

So, if GCC warns about the use of x in the calling code:

  5342    int geometry = XParseGeometry (SSDATA (string),
  5343                                   &x, &y, &width, &height);
  5344    Lisp_Object result = Qnil;
  5345    if (geometry & XValue)
  5346      {
  5347        Lisp_Object element;
  5348
  5349        if (x >= 0 && (geometry & XNegative))

... then GCC is giving a false alarm: x must be initialized in line 5349 if 
(geometry & XValue) is nonzero in line 5345.

If my reasoning is wrong, can you give the path through the function and its 
caller where x is used as an uninitialized variable? Because if there is such a 
path, we shouldn't be marking x with UNINIT; UNINIT is only for pacifying false 
alarms.

Similarly for y.