[A21.5] Re: commit: Fix window-configuration problem.

[Stephen J. Turnbull]

APPROVE 21.5

Does 21.4 need a similar patch?

Michael Sperber writes:
 > "Stephen J. Turnbull" <stephen at xemacs.org> writes:

 > > First, why is *any* permanent motion OK?
 > 
 > Anyway, I've figured out what caused this.  (It has nothing to do with
 > window configurations.)

OK, thanks for confirming this.

 > The problem is that bumping the minibuffer size
 > up and then down again isn't symmetic.  You can try this out by opening
 > up a frame on *scratch*, splitting it, and then doing this there:
 > 
 > (enlarge-window 1 nil (minibuffer-window (selected-frame)))
 > (enlarge-window -1 nil (minibuffer-window (selected-frame)))
 > 
 > The attached patch fixes this for me.  You still (potentially) get a
 > small bit of permanent motion due to the rounding to the default text
 > height, but it stabilizes after the first roundtrip.

I agree this is a fix; enlarge-window should not check if it's
restoring a window configuration.  resize-minibuffer-mode may be buggy
or need a feature to make its behavior more predictable, but that's
not a problem with the window configuration code (including
enlarge-window).




  If there are no
 > objections, I'll push this on Tuesday or so.
 > 
 > 2008-04-26  Mike Sperber  <mike at xemacs.org>
 > 
 > 	* window.c (set_window_pixsize): Round up when we're shrinking,
 > 	down when we're growing * to make sure that pairs of grow / shrink
 > 	meant to * cancel out actually do cancel out.  This fixes a
 > 	problem with `resize-minibuffer-mode' where successive grow/shrink
 > 	actions are meant to cancel out, but previously didn't.
 > 
 > -- 
 > Cheers =8-} Mike
 > Friede, Völkerverständigung und überhaupt blabla