cordova-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Patrick Mueller <>
Subject Re: who wants to play with me on a commit approval process for weinre?
Date Mon, 16 Jan 2012 17:06:07 GMT
On Mon, Jan 16, 2012 at 11:32, Patrick Mueller <> wrote:

> I'm wondering if someone would like to experiment with me on getting an
> approval process in place for commits.

BTW, I'd be interested in finding out about other Apache project's commit

Probably the most interesting commit process for a non-Apache project I've
seen is WebKit's.  It goes like this:

- open a bug
- when you have code ready to commit, add the patch as an attachment to the
bug; there's a command-line script that will do this for you
- WebKit's Bugzilla knows about these attachments, and provides a mechanism
for you to signal that a commit needs a review, that the commit has passed
review, and that the commit has failed review.
- the reviewer signals pass/fail on the review
- if the review passes, the code is automagically comitted and a build is

Since I've been involved, they've taken this one step further, in that you
can have a build run against your patch without having it committed to the
SCM.  This is a substitute for the presumed step of the reviewer applying
the patch and making sure things till work.  Again, there's an indicator in
Bugzilla to show you the status of this build.

But wait, there's more!

- all this stuff is logged to IRC as well as other obvious channels (bug
emails, scm emails, etc)
- there's an IRC bot that can be used to automagically "back out" a
committed patch by creating a new bug and a "reverse patch", then applying
it and committing it.

It's a pretty nice system, but obviously relies on:

- some automation code running in various places
- extending the bug system to take all this stuff into account

All in all, it's quite nice - everything revolves around Bugzilla.

Anything else in this vein available at Apache?

Patrick Mueller

  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message