release naming (was: GNU Autoconf test version 2.59c available)

[Ralf Wildenhues]

Hi Eric,

* Eric Blake wrote on Thu, Apr 13, 2006 at 02:38:47PM CEST:
> 
> Should this have been numbered 2.59d?  I thought the odd-letter extensions
> were for CVS versions, and the even-letter extensions are for alpha
> releases.  Or is that only libtool following that convention?

As far as I know, Libtool is following that convention, but Autoconf
is not.  The nonexistent 2.59a was because Paul found some claims of
a 2.59a version in the wild[1].

To tell you the truth, I've thought quite long about naming it 2.59d
last night; IMHO it would have been clearer.  I decided against it
because I wasn't totally sure about all consequences of changing
configure.ac right before the release; it would have required me to
checkin all Makefile.in's with `cvs ci -f', for example, in order to
make sure the CVS contains an exact copy of what goes in the tarball.

BTW, I did such a `cvs ci -f' round after the change to 2.59d after
the release, so that potential contributors do not experience spurious
automake reruns.

I believe it may be a good idea to follow a similar release numbering
as Libtool[2] also in Autoconf.  That would be to have the odd last
part of the version number/character as CVS versions, and the even
ones as release versions.  But it would clearly be inconsistent with
what Autoconf has had up until now: 2.59 is stable, and 2.59c is now
a released (alpha) version.


Apologies BTW for the messed-up Mail-Followup-To: header of the
release announcement.  This and some other nits that have come up
last night will be sorted out in due course, soon..

Cheers,
Ralf

[1] http://lists.gnu.org/archive/html/autoconf/2004-08/msg00082.html
[2] http://www.gnu.org/software/libtool/contribute.html under the
    headline "Release Numbering"