[mythtv] Re: LIBVERSION = 0.15.0.99?

[Axel Thimm]

On Sat, May 29, 2004 at 03:51:22PM -0400, Isaac Richards wrote:
> On Saturday 29 May 2004 02:25 pm, Axel Thimm wrote:
> > On Sat, May 29, 2004 at 11:31:58AM -0400, Isaac Richards wrote:
> > > Why don't you just put 'cvs' in the package name, say they conflict with
> > > the normally named packages, and use the date you generated the packages
> > > as the version number?
> >
> > I would have to obsolete "mythtv" with "mythtv-cvs" and vice
> > versa. While technically possible it is error prone (all
> > dependencies need to be rewritten in the specfiles from "-cvs" to
> > "" and back at each switch), and at the end I could be doing more
> > damage than good. These obsoletions could be versioned, but then
> > again the issue about what version to attach to the cvs builds
> > rises again, so we are where we started.
> 
> Where we started?  The package name would be:
> 
> mythtv-cvs-20040529-1.rhfc1.at
> 
> No versioning issues whatsoever.

IMHO it is even worse. You propose two sets of packages with
completely different and uncomparable versioning. So for switching
from mythsomething-cvs = 20040529 to mythsomething = 0.16 you need
special versioned obsoletes all around the specfiles. Even worse you
need to change dependencies (e.g. Requires: libmyth-cvs = 20040529 vs
libmyth = 0.16).

This leaves too much manuall fidling and _will_ break packages at
every second switch. Renaming packages is in general something very
fragile, so I wouldn't follow that path.

mythtv isn't the first project to be packaged in cvs snapshots and
many models have already been probed and failed. The surviving two
models are using the next version with cvs/pre/rc appended to the
release tag, or crafting an intermediate versioning scheme.

> It's immediately clear when you pulled the CVS checkout, and much
> more useful than some randomly generated version string.  The user
> must install 'mythtv-cvs' instead of just switching to use a
> different repository.  Makes much more sense than meaningless
> libversion increases in the source that needlessly break binary
> compatability.

LIBVERSION was probably the wrong entity to quote. Forget about
LIBVERSION, simply stating that cvs between 0.15 and 0.16 is to be
called 0.15.99cvs would be enough, no code changes neccessary.

> > The best thing would be to maintain an interim version in the
> > sources themselves, perhaps 0.14.99cvs.
> 
> And that's just not going to happen.  I'm not going to do something
> the wrong way just to make making CVS packages that I already
> disagree with easier.

No, problem, I was just proposing something that I considered would
improve things, perhaps it does not, so let's drop it, and continue
with business as usual ;)

Why do you disagree with CVS packages, I thought you had given your go
some months back, and I believe it did pay off, or not? If you want
the CVS packaging to stop, let me know before I start pushing out the
next builds.
-- 
Axel.Thimm at ATrpms.net
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://mythtv.org/pipermail/mythtv-dev/attachments/20040529/8a69be53/attachment.pgp