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

[Bjoern Michaelsen]

Hi,

On Tue, Jun 19, 2012 at 05:04:58PM +0200, Lubos Lunak wrote:
>  Which is the problem. Besides asking just to do that it should have also said 
> why one should do that. I'm one of the people who haven't done that, 
> because 'Do it because.' scores pretty low with my motivation. 

I hoped "as you otherwise will be completely locked out of commiting, once we
switched over!" (as from that other mail) scores better in your motivation.

> I wasn't joking when I mentioned in some other mail that it would be nice if
> you finally told us something about this Gerrit stuff that wasn't just a
> random remark. I still have yet to see a good description of what Gerrit is
> actually supposed to do for us exactly.

http://nabble.documentfoundation.org/ANN-Please-use-Gerrit-from-now-on-for-Patch-Review-td3990754.html#a3991006

for starters, nothing of that being new, actually most of it being as old as:

LibreOffice Conference 2011, Friday, October 14, 2011 - Presentation "Crowdsourcing Code Reviews (Bjoern Michaelsen, Canonical Ltd)". 
http://wiki.documentfoundation.org/images/8/85/Crowdsourcing_code_reviews.odp

>  And while I myself decided I wouldn't bother until it comes, right now it 
> appears to me like nobody has been really told, and that does bother me.

Thats hardly helpful. There is a lot to room to shape, improve and form the way
we are using gerrit. "I dont know what it is, but I dont like its color" is not
a valid argument.

>  And I hope that this 'completely' is the daydreaming part.

No its not. The last thing we need is two repositories and admining them both
with fragile syncs etc.

> I'm a bit hazy on the details (since, you know, I've just picked them up
> randomly here and there), but if I understand it correctly, we are asked to
> move exclusively to a different workflow with an unproven tool.

No you are not. You will be able to bypass review on gerrit and commit directly
to master as you did before, _if_ you set up an account and allow us to get you
the ACL in gerrit set up right for you. Everyone who hasnt done that since a
month ago, is only hurting himself and costing the gerrit team a lot of nerves.

> In a month.  And I thought it was a running joke that this codebase breaks
> every single tool that's use on it.

Hmm, so how about getting used to the new possiblities _now_ so that we are not
doing hop over to reviewing patches on gerrit with 400 developers on _one_ day
with limited people available to spread the knowledge? Dont you think it might
help doing that over a month, rather than on one day? It might smooth things
out a bit, you know ...

So again, what we are doing now is:
- trying to get everyone get an account on a TDF-controlled git host, which in
  addition will enable us to continue to do all of the above
- Nothing but the remote host to push to changes by doing this
- Putting things under review in gerrit is currently not planned to ever be
  required -- but it is assumed to be soon the only/canonical way to do reviews
  for the benefit of all (and without major drawback as reviews will be doable
  from the commandline, from the web, possibly from IRC or via mail later too)

So this enables us to push things forward without ever taking anything from
you. You will be able to continue to push to master directly under gerrit.

However, I and I guess the rest of the gerrit team are rather confident, that
at one point peer pressure will make you wisen up: It is a lot more comfortable
to get one mail in private from a tinderbox that you broke the build, than one
that goes out to 50 people. If you made fragile changes and the the other 49
people did do the pre-master tinderbox builds, they have every right to be
annoyed by you.

So: There havent been and wont be any policy changes with regard to code
review. Once we have the tinderboxes lined up for this, I assume this to solve
itself and people doing the sane thing.

> Tools are there for people, not people for tools.

Which is exactly why we are constantly in trouble finding enough reviewers.
Because the current tooling -- or lack thereof -- sucks and I am confident I am
not alone with that opinion.

So dont complain against some daydreamed phantom of what you imagine gerrit to
be, but help us make it something doing what you need.

Best,

Bjoern