["Olivier Biot" <ethereal@xxxxxxxxxx>]

From: Gerald Combs

| Joerg Mayer wrote:
|
| > On Fri, Feb 20, 2004 at 10:04:21AM -0600, Gerald Combs wrote:
| >
| >>  "Makefile fragments included this way are always distributed
because
| >>   there are needed to rebuild Makefile.in."
| >>
| >>It appears that gtk/Makefile.common should have been included in
the
| >>distribution automatically.
| >
| >
| > Hmm, that is cool. Did you create the tarball via make distcheck
or via
| > make dist (and did the make distcheck on some other system)?
|
| The tarball was created by running "make distclean", "autogen.sh",
| "configure", and "make distcheck" (which depends on "dist") in that
order.
|
| BTW, I had to rename the CVS directory for the release so that
| make-version.pl wouldn't define CVSVERSION.  Is there a cleaner way
to
| undefine CVSVERSION?

AFAIK make dist should not include the CVS directories and the
.cvsignore files, however this seems to be the case. As a result, the
make-version.pl script sees that we're building from CVS and it will
generate a CVS version in cvsversion.h.

Maybe we want to create a beacon file in the top-level directory which
would then inform make-version.pl that we're creating a real official
release hence creating a cvsversion.h header *without* defining the
CVS version.

Regards,

Olivier