SVNPoller fails to recognise a branch deletion

[nhemingway]

If a SVN branch is deleted it is detected (correctly) as a change by SVNPoller. However, it then blindly creates a Change without taking into account the fact that the Change will attempt to checkout and build something that is no longer there.

[nhemingway]

Patch to resolve #261

[nhemingway]

The patch I've just attached (buildbot_branch_deletions.diff) has svnpoller checking whether the change reported by SVN is a branch deletion. If it is, a buildbot Change is not created.

[dustin]

I can't really see why someone would want buildbot to handle this sort of situation, but this does seem a better way to handle it.

In the dev tree as #261:buildbot_branch_deletions.diff

[nhemingway]

The "standard" or "documented" use for buildbot is to build the same branch on multiple architectures. Our app is perl running in house on a single target platform. But we do have multiple project and ticket branches separate from trunk. So we use bb to have a builder perl branch. All fine until someone deletes their ticket branch (having merged it back to trunk) at which point bb fires off a (pointless) build for a branch that's no longer there!

[dustin]

got it -- apparently I wasn't thinking too hard. Thanks!

[dustin]

pushed to brian's tree