Re: [ANN] Please use Gerrit from now on for Patch Review

On Wed, Jun 20, 2012 at 12:26:23PM +0200, Lionel Elie Mamane wrote:
> Ah, I understand. When the gerrit repos are the "true one source" and
> gerrit will do the "push" automatically once someone validates the
> patch in the web interface, what will "Committer" be? The one that
> uploaded the patch or the one that validated it in the web interface?
> IMHO it would be nice if it would be the one that validated in the web
> interface.

Author will be the whoever was the original author. Commiter will be be the
guy/gal pressing the submit button. Everyone giving a codereview+1 will be a
signoff in the commit.

> With "diff --recursive" or quilt or debdiff (after building custom
> Debian packages) or ... It is not inconceivable for people to hack on
> the sources distributed by their distros rather than upstream sources;

IMHO yes, distro build are a) patched b) usually an older major release and
thus patching created against them are possible trouble. That said, we still
need the patch maildrop, but for exact those reasons its not totally trivial.
Once there though, it will be better than now, as the conflicts are noted early
on (on submittal) and not late when the reviewer is trying to do something with
it.

> I often do that for software where I just want to fix this one bug,
> because "apt-get source package" is so much easier than "find where
> the hell upstream sources are, use specific tool I might not be
> familiar with (git, darcs, svn, mercurial, ... there are so many)".
>
> Anyway, I meant "setup git for gerrit with SSH keys and all that".

Maildrop will be of some help, but in general people should be encouraged to
get to upstream as the conflict changes between majors are too high. (Consider
someone on Ubuntu LTS sending his 3.5 patch in 2 years).

> (although I *do* *not* understand *at* *all* why we have to require an
> OpenID, instead of offering it as an alternative)

http://xkcd.com/927/
OpenId is currently the most common authentication esp. for non-nerds (and we
should try to consolidate on one authenication method, OpenId is currently the
best guess).

Best,

Bjoern