[Buildbot-devel] Re: Branches in changes

[Chris Leishman]

On Fri, Nov 25, 2005 at 02:37:00PM -0800, Brian Warner wrote:
<snip>
> > I'm wondering, however, if the 'branch' should not actually be
> > associated with the file - not with the change.  This would, of course,
> > mean having a list of tuples in the change, of the form:
> > 
> >   ((file, branch), (file, branch), ...)
> 
> Well, I'm really keen on sticking to the "one Change / one branch" approach.
> That makes the rest of the Buildbot work-flow much much more sensible. Each
> build is of exactly one branch, you can assign blame based upon branch
> (everyone gets blamed for breaking TRUNK, nobody gets blamed for breaking
> personal branches, etc), you can allocate build resources on a per-branch
> basis (do the big tests on TRUNK but only some smaller tests on personal
> branches, etc).
> 
> It does mean that perhaps tools like svn_buildbot.py should be allowed to
> send multiple Changes when faced with files that don't all end up inside the
> same branch.

I agree with the above.  I think I was just suggesting that the 'change'
as sent to buildbot could contain sets of branches and associated files
in that branch - then the master could split them into different change
objects internally if that made more sense.  It would be nice if the
status displays also kept the changes concatenated so that it is obvious
which files where committed together (even if they cross branches).

But all that is kind of optional - it's probably easier to get the
change injection tool (svn_buildbot in this case) can split them into
different changes.

chris