[AC21.5R21.4] Draft fix for metacity maximization bug

[Vin Shelton]

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

> APPROVE COMMIT 21.5 RECOMMEND 21.4
>
> Vin, there's a bunch of "future work" cruft here, so I don't
> necessarily recommend this patch as-is.  Also, I haven't actually
> checked if it applies, and the backwards-compatibility variable might
> be a little bogus in the stable line.  Let me know what, if anything
> of this you want.
>
> Thanks to Glynn Clements, without whose discussion I never would have
> figured this out, to the good folks at Sun, HP, etc, who funded Real
> Documentation that Actually Helps, and to the various users who kept
> reporting the issue even after we thought we (mostly) had it licked.
>
> I provided an earlier patch, about which I wrote:
>
>     sjt> It is possible to inhibit this, with no huge ill-effect on
>     sjt> XEmacs functionality.  I'm supplying a patch, with this
>     sjt> action controlled from Lisp by the variable `wedge-metacity',
>     sjt> defaulting to nil.
>
> This patch is *superseded* by the following patch, which may not apply
> to 21.4 (sorry), and if you applied the earlier patch you'll have to
> revert it (sorry again).  If you can apply it and are willing to test
> it, it should "just work", no need to fiddle with anything.  I will
> work up a 21.4 patch per Vin's comments (if any) within a day or so.
> And it's being committed to 21.5 immediately.
>
> If you run with the patch and have minor problems (but enough that
> you're willing to avoid maximizing under metacity et al), you can
> restore the old buggy behavior with (setq wedge-metacity t).
>
> I'm pretty sure this patch is "right", although I haven't been able to
> test with metacity.
>
> (1) The Xt conventions say "don't call geometry request functions from
> the resize callback".  This patch moves the request from a resize
> callback into a function explicitly intended to change size.  The
> theory is right.
>
> (2) With the old behavior, the resize callback called itself
> recursively, at least once.  Now it doesn't.
>
> (3) The fluxbox window manager now gives "true maximization" where it
> used to flash to the true maximum, then drop back to the largest
> cell-oriented window.
>

Stephen,

Are you interested in generating a 21.4 version of this patch?  If
not, I'd be happy to sift through it if you told me which parts are
relevant to the actual maximization bug.

Thank you,
  Vin