changelog format

[David Cantrell]

On 04/20/2009 11:18 AM, Adam Williamson wrote:
> On Mon, 2009-04-20 at 10:26 -1000, David Cantrell wrote:
>> On 04/20/2009 10:21 AM, Adam Williamson wrote:
>>> On Sun, 2009-04-19 at 23:32 +0200, Björn Persson wrote:
>>>> I gather I'm expected to write the changelog in a spec file by hand.
>>> Why is this, BTW? Is there a reason we don't just generate it from CVS
>>> commit messages, beyond "no-one's had time / inclination to implement
>>> it"?
>> Because there's no forced standard for CVS commit messages.  Plus, each
>> CVS commit in pkgcvs does not necessarily equal a new release or version
>> increment in the package.  At least that's not how I use pkgcvs.
>
> It doesn't have to - you can generate the changelog as part of the
> package submission process, containing all the commit messages since the
> last package build.

But that's the thing I don't really want in the RPM changelog.  All of 
the commit messages since the last package build are mostly noise to the 
average user.

>> I view the RPM changelog as mostly fluff for end-user consumption.  We
>> [the packagers] can summarize the main points of that release, note bug
>> numbers addressed, and other major points for that iteration of the package.
>
> You can have a keyword that you put in CVS commit messages that
> suppresses them from being added to the RPM changelog, if they wouldn't
> be useful in that context (like "rebuilt with no changes for some
> procedural reason").
>
> (The win of doing things this way, btw, is that it saves maintainers
> time and effort, reduces errors introduced by the manual creation of
> what should be boilerplate content, and makes it less likely that
> important information will be left out of RPM changelogs, since you have
> to have a commit message and most developers habitually write useful
> ones).

You don't have to sell me on the idea, I do like it.  But it should have 
been implemented when pkgcvs was created.  As it stands, there are far 
too many garbage commit messages in pkgcvs now and far too much useful 
information in the rpm changelogs.

I think this problem would be better solved in a larger 'moving pkgcvs 
to some other vcs' discussion.

-- 
David Cantrell <dcantrell at redhat.com>
Red Hat / Honolulu, HI