Hi, after thinking a bit about the stuff with version.texi and other places where we use version strings, I am of the opinion that this sort of version string generation, now scattered over several files and sometimes automatic and sometimes not, should maybe done by autogen.sh exclusively.
The disadvantage would be that we would get the unchanged version numbers even from people downstream. However, since downstream is not likely to add anything to our own ChangeLog file, that's the situation we are likely going to have to deal with, anyway. The release/snapshot/package creation process would run autogen.sh, so that should get us the right results. As I see it, there are a few issues remaining for that automation: a) configure.ac needs version info. We would likely read it in from a generated file. Or let autoconf run on standard input and pipe the version information into it before configure.ac. b) the RPM spec file. I don't think it is possible to read info from other files, so we would likely have to splice it into the file itself using sed. If we do that, we actually could do the same with configure.ac. It does not seem to make much sense in having some of the information updated by configure rather than autoconf: it will likely only buy us inconsistencies. So actually autogen.sh should probably be the only utility actually taking a look at ChangeLog, and the information should trickle down from there. -- David Kastrup, Kriemhildstr. 15, 44793 Bochum _______________________________________________ auctex-devel mailing list [email protected] http://lists.gnu.org/mailman/listinfo/auctex-devel
