More comments on commit msgs

[simo]

On Tue, 2008-01-15 at 07:53 -0600, Gerald (Jerry) Carter wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Not picking on Michael here.  Just a good example of a bad commit msg:
> 
>   commit 3972deb90c4b645fb4d207a7e132cd7e180e78bb
>   Author: Michael Adam <obnox at samba.org>
>   Date:   Thu Jan 3 00:51:36 2008 +0100
> 
>     Use a better parameter name.
> 
>     Michael
> 
> As we are preparing to make the release management less
> technical and more management, these types of commits msgs are
> impossible to decode.  Please take the time to be more
> descriptive.  Thanks.

Not to put on Michael,
but in general I too would like to ask people to be more descriptive.
And explain why the change was important.
I am parsing the commits between release-3-0-28 and v3-0-test and at
least in one or 2 cases even looking at the code is not helpful in
understanding if the change is important or not, or, worse, what does it
affect.

As another side note, given that git let you commit locally and later
push, it would be *very useful for release mgmt if you could compound
and do a single commit push when you have 2 patches completely
interdependent (like modification of the header and of a function, etc).

Keeping separate commits for different changes (even on related matters)
is perfectly fine (and preferable when the change is big and can be
split), but it would be very nice if in the log message you could
reference the commit you depend on.

Simo.

-- 
Simo Sorce
Samba Team GPL Compliance Officer <simo at samba.org>
Senior Software Engineer at Red Hat Inc. <ssorce at redhat.com>