On 10/04/2013 09:00 PM, Ian Bull wrote:
I'm not sure how the more experienced Gerrit users
deal with this.
I don't consider myself as an expert, but I generally try to keep
discussion related to use-case, test scenario and possible designs
on the Bugzilla, and put discussions about code itself of Gerrit as
it allows inline comments in contributions, versioning and
comparison of patches, easy fetch, automatic CI check.
When the discussion on Bugzilla has come to a point where it becomes
possible to write code, I put a link to a Gerrit contrib on
Bugzilla. Then most of the discussion happens on the Gerrit patch,
and when it is done, it gets merged and then we can close the
Bugzilla entry.a
Bugzilla tracks ideas, Gerrit tracks code changes.
Not sure it's optimal, but I find it more comfortable than dealing
with patches to merge and test, mark obsolete and put comments
without a scope in Bugzilla.
I also think it makes it easier to review a contribution when we see
it does not introduce regression before even we look at the code.
And I also find it easy for a contributor to learn to take care
about regression tests when pushing a change on Gerrit: if he broke
something, a mail automatically warns the contributor. It tends to
give more sense of responsibility and then to provide better quality
in patches.
SWTBot enabled Gerrit last year. Since then, project had 8 new
contributors; for a total of 20 contributors in 5 years of
existence. I'm not sure it is directly related, but I do feel Gerrit
has been helpful to increase project activity, diversity and
openness.
My 2c.
|