Re: Rietveld workflow problems

[David Kastrup]

Reinhold Kainhofer <address@hidden> writes:

> Am Wednesday, 21. September 2011, 15:04:05 schrieb David Kastrup:
>> Reinhold Kainhofer <address@hidden> writes:
>> > Am Wednesday, 21. September 2011, 10:52:37 schrieb David Kastrup:
>> >> Perhaps it would be nice if we found a way to play with "Gerrit",
>> >> supposedly a git-based system similar to Rietveld.
>> > 
>> > I looked at gerrit a while ago. If you want to take a look at it:
>> > http://server.kainhofer.com:8088/
>> > 
>> > Here is a quick summary:
>> > 
>> > -) Each review is basically one commit in a branch on the gerrit server.
>> 
>> Well, that's pretty much what we already have.
>
> No. In gerrit you really need to clean up your patches before you
> submit them for review. I typically have lots of small commits in a
> branch when I upload a patch to rietveld. git-cl will simply take the
> diff to origin/master (i.e. all patches in the branch combined into
> one large patch) and show the combined diff. gerrit, on the other
> hand, will show each small patch as one review.
>
> So, I would have to clean up my local branch before submitting for
> review (i.e. rebase -i and squash and reorder the patches). That's a
> very fundamental difference in the approach. Rietveld works on the
> diff level, gerrit on the git commit level.

Which is what actually will get committed and thus permanently recorded.
So it makes sense to get the structure as well as the result of a patch
series open to review.

If you feel you want to review a single humongous patch, make a separate
review branch and pick the whole of the detailed branch into it with git
checkout -p.

As an example, how would the dev/syntax branch look when submitted
uncompressed for review?

That's pretty much a real-life example.  And what happens with existing
comments when one restructures/rebases a branch?

-- 
David Kastrup