cvs vs svn...

[Paweł Sakowski]

On Fri, 2005-09-09 at 09:40 +0100, wrobell wrote:
> [svn log] -> few days please.

While we're at revolutionary ideas...

Currently %changelog doesn't really correspond to the CVS commit log. It
is subject to truncation, editing (adding CAN entries, fixing typos,
rewording). Furthermore, at some points it (here both changelog and
commit log) simply restates the change made ("added BR foo-devel"), at
others it has nothing to do with the change made ("rel 3. Nothing really
changed in this spec, but that other patch got really improved"). Some
places it just adds useless noise to the spec ("trying to do foo",
"another attempt to do foo", "foo won't work anyway, reverting last 2
commits") or just document flame wars ("last commit was issued by an
idiot").

So, the crazy idea is to give up automatic %changelog generation and
treat it more like a NEWS file (i.e. document important changes, not
each invidual commit). I would imagine that a new %changelog entry would
be added for each new release and state the reason that the new release
is a Good Thing, rather than document each change (nobody really cares
about "spaces->tabs"). Note: it's just a gut feeling that this is the
kind of information that people reading the %changelog are actually
looking for. I myself am not sure what exactly I need from the
changelog, and it's important that a changelog isn't manually writter
Ars Gratia Artis, but it actually brings useful information to the
developers. If it turns out that the commit logs bear close resemblance
to the information useful to the developers, the above idea is obviously
pointless (I'm not that evil to make people write the same thing twice
in the {commit ,%change}log).

Yes, it's no coincidence that the above changelogging scheme works
better with svn than the current one ;)

-- 
Paweł Sakowski <saq at pld-linux.org>
PLD Linux Distribution