[Svnmerge] [PATCH] Fix for bidirectional merging corner cases

[Raman Gupta]

Raman Gupta wrote:
> A refactorization of the code in revision 18696 that checks for
> reflected revisions broke it for a corner case. The current code
> may incorrectly consider the first revision to be reflected even
> when it is not. This can occur if the first revision included a
> change to the merge property, but one that did not change the
> merged revision list for the source branch.
> 
> See my patch below. Since old_revs is not initialized in the loop
> that checks the merge_metadata_changed revs, the first rev to be
> checked is ALWAYS considered reflected.

Is anyone going to review (and commit) this patch? It's pretty simple,
shouldn't take too long.

Cheers,
Raman