Re: Standardizing ConfigureRequest vs. size_inc hints

[Matthias Clasen <maclas gmx de>]

On Sat, 2002-08-17 at 04:47, Havoc Pennington wrote:
> 
> Matthias Clasen <maclas gmx de> writes:
> > Window Managers should not react to WM_NORMAL_HINTS changes by changing
> > the size of the window, since that leads to race conditions.
> > 
> 
> FWIW metacity does resize the window to the closest valid size,
> because the way it's implemented is that the size hints impose a
> constraint and the invariant is imposed that the constraint is never
> violated, but I don't think that breaks anything (since if the client
> immediately sends a configure request things will end up right in the
> end).
 
No that won't break anything. 

> 
> I don't understand doubling window size if size_inc goes 5->10, I
> would never have expected that behavior.

This is fvwm trying to keep the same "cell size" ie make sure the xterm
is still 80x25 instead of 40x12.5 after the _inc change. This is broken
and leads to the race condition Dominik explained, since it can only
work if the wm gets the chance to do this "cell size tracking" on *each*
_inc change separately, when in fact it may loose some _inc changes due
to request batching on the client side.

Matthias