> Just to check I've understood what you want here, I did a first run with
 > the latest gtkutil-frame-2.diff applied, and with breakpoints on those two
 > lines, of simply "emacs -Q", and I get, for just opening the initial
 > window, 4 hits on each breakpoint.
 >
 > The first hit is w = 0, h = 0.
 >
 > The other three are w = 64, h = 167.

OK.

 > I wasn't exactly sure what you meant by a "good" or "bad" scenario; but I
 > guess you meant "good" == "with patch applied", and "bad" == "without patch
 > applied"?

With "good" scenario I meant one where you don't see any shrinking.  A
"bad" scenario is one where you see it.  Sorry for being unclear.

Please apply the attached diff to an unpatched Emacs and tell me whether
the frame sizes for a bad scenario are in any way different from those
that you see for the same bad scenario in an unpatched Emacs.  Here the
patch has no effect.

Thanks, martin