its clearing its changes cache after doing work. so if it fails half way through, the next time around, it has crap in it.

there is an implied assumption in the code that the cache will be empty when it starts. in git and hg, we dont have this assumption - we clear it at the start rather than the end.

i think the proper fix here is that we shouldnt be using the same object across multiple changes / branches / etc. we should create a new one each time, so we dont have to rely on clearing it at all.

specificaly, this is at ClearCaseRepositoryScanner.processBranch, and the clearing call is changeParser.resetChangeSets(). this call should at least be the first line in the method.