Re: Considering Gerrit for CFs

On Fri, Feb 8, 2013 at 4:20 PM, Peter Eisentraut <peter_e@gmx.net> wrote:
> On 2/8/13 5:23 AM, Magnus Hagander wrote:
>> But do you have any actual proof that the problem is in "we
>> loose reviewers because we're relying on email"?
>
> Here is one: Me.
>
> Just yesterday I downloaded a piece of software that was previously
> unknown to me from GitHub and found a bug.  Within 15 minutes or so I
> had fixed the bug, made a fork, sent a pull request.  Today I read, the
> fix was merged last night, and I'm happy.
>
> How would this go with PostgreSQL?  You can use the bug form on the web
> site, but you can't attach any code, so the bug will just linger and
> ultimately put more burden on a core contributor to deal with the
> minutiae of developing, testing, and committing a trivial fix and
> sending feedback to the submitter.  Or the user could take the high road
> and develop and patch and submit it.  Just make sure it's in context
> diff format!  Search the wiki if you don't know how to do that!  Send it
> to -hackers, your email will be held for moderation.  We won't actually
> do anything with your patch, but we will tell you to add it to that
> commitfest app over there.  You need to sign up for an account to use
> that.  We will deal with your patch in one or two months.  But only if
> you review another patch.  And you should sign up for that other mailing
> list, to make sure you're doing it right.  Chances are, the first review
> you're going to get is that your patch doesn't apply anymore, but which
> time you will have lost interest in the patch anyway.
>
> So, I don't have any further evidence that we are losing reviewers, but
> in light of the above and the options out there were interested
> developers can contribute much more easily, I'm amazed that we are
> getting any new contributors or reviewers at all.
>
> Of course, Gerrit doesn't actually address most of the issues above, but
> it could be part of a step forward.

You're outlining an issue for submitters. Berkus was complaining about
issues for reviewers. These are clearly different issues. And I don't
think gerrit helps at all with the submitters process that you've
outlined above - it's a tool to help the reviewing. That doesn't, of
course, mean that we shouldn't try to solve both things - but they are
completely different.

Basically, what you're saying above, is we should start accepting pull
requests from github. There's nothing preventing us from doing that
(other than the wish to do so), no need to change tooling for review
for that. It just means that committers need to use git and add
peoples repositories as remotes instead of applying patches. Probably
not a huge burden today since most pg developers are used to git by
now.

However, it's not going to change the requirement to help review other
things, that's a pure policy issue. Which I'm pretty sure we don't
enforce for 10-minute-trivial-fixup-patches. And it's not going to
change the fact that it takes time before someone gets around to your
patch, that's a resource issue. And it's not going to change the fact
that a patch migt not apply after 2 months, that's a consequence of
the second problem. It doesn't change the fact that you have to sign
up - it just makes it more likely that you're already signed up, since
so many people are on github already, but you *do* have to sign up for
a service, wherever it's hosted.

But it does change the fact that you don't have to deal with email,
and can use web instead.

Personally, I find it much easier to just "git clone, make changes,
git diff, attach to email", than "fork on github. git clone, make
changes, push to github, create pull request on github. repeatedly
check status of said pull request since email notifications aren't
usable". But that's me, personally, and I realize many people today
prefer web interfaces for as much as possible. There's nothing
stopping us from supporting both, of course.

--Magnus HaganderMe: http://www.hagander.net/Work: http://www.redpill-linpro.com/