[Xcb] Tasks required in order to fix XCB-based Xlib

[Barton C Massey]

In message <41F8A3AD.7020200@verizon.net> you wrote:
> First, there was a minor change to the pkg-config scripts of xcb-proto,
> involving the method used to find the protocol descriptions.  The old
> method parsed the CFLAGS and looked for -I flags; the new method just
> supplies the path to the protocol descriptions as "xcbincludedir".  (For
> consistency with automake, perhaps this should be called
> "pkgincludedir"?)  This would be a five-minute makefile/configure.ac change.

Excellent.  Please do.

> The far bigger issue is that of the M4 macros and M4 wrapper definitions
> used in the Xlib build in order to autogenerate wrappers around many XCB
> protocol functions.
[...]
> Another possible solution would be to grab the last M4 definitions, run
> them through the existing Xlib M4 macros, and commit that to the Xlib
> repository.  This would be the fastest fix to get things unbroken, and
> might still be useful even if we go with the above solution, but it may
> be less than ideal in the long run.

Yes, please do this.  Hopefully there isn't a "long run" for
Xlib/XCB: if there is, then we can always backport using
XSLT on the existing XML descriptions or something.

Any investigation on passing the test suite?  Or are we too
far from a once-again working version yet?

    Bart