As far as I understand Hendriks analysis, there is indeed nothing upstream FVWM
can do about this. He included the fvwm mailing list before his analysis showed
that the bug is introduced in Debian.
Hendriks analysis indeed suggests to me that, until all X servers shipped with
Debian support XKEYBOARD, the packaged fvwm should not silently assume that this
extension is present. Optimally, it should dynamically detect at startup which
functions are offered by X and then use them accordingly. More simply - indeed -
it could just stick to the deprecated functions for now, which seem to be
(still) supported by all X servers.
Thanks for pointing out that one can effectively disable the patch also by an
undefine. While this may be technically interesting for the package maintainers,
suggesting that Debian useres should routinely recompile their software on need
is really beyond the scope of a binary distribution.
@Hendrik: Many thanks for your detailed analysis.
Dr. Claude Krantz
Max-Planck-Institut fuer Kernphysik