Base being kept as a reference can cause spurious tests to be triggered

[daveh86]

I've been chasing a spate of spurious builds on our Jenkins instance with your plugin.

With extended logging I was able to gather the following:

Dec 22, 2016 5:46:39 AM FINEST org.jenkinsci.plugins.ghprb.GhprbPullRequest
Comment was made after last update time, Trying to trigger a spurious build.
Dec 22, 2016 5:46:39 AM FINEST org.jenkinsci.plugins.ghprb.GhprbPullRequest
[David Hows] Added comment: Trying to trigger a spurious build.
Dec 22, 2016 5:46:39 AM FINE org.jenkinsci.plugins.ghprb.GhprbPullRequest
New commit. Sha: Head[62fb90280d2b79af09d6ef2117d065540a09c0f8 => 62fb90280d2b79af09d6ef2117d065540a09c0f8] Base[fc2a64f154d94072c89171f94166d81ebfe5a77d => 49813035a92dd2eb792ff331f535a8e34c14e5b0]

This was in response to my adding a comment to a pull-request that had not been updated in 2~ weeks to be merged into our main development branch.

It appears that the reference used as the "Base" has moved forward in time, despite being the same branch.

I checked both commit hashes and they don't correspond to any "particular" commits (forks for other branches, point that the first commit was made on the PR's branch, initial commits, etc).

I suspect that the underlying github plugin used (or github itself) simply returns an "arbitrarily" old commit, which then changes as new commits are added to the branch. This in turn leads to the bad behaviour.