Vendor import and funky branch merge

[Steve Madsen]

We have a project here that is basically a set of vendor sources with a small 
number of local changes.  The project started life as an initial import, had 
some local changes committed, and recently another import.

Between the local changes and the second import, some branches were made for 
more intensive work.

After the second import, it became obvious to me that some of those local 
changes were actually patches from the vendor and should have been imported 
so they ended up on the vendor branch.  To fix the cases were a file was 
entirely vendor patches, I used "cvs admin -b<vendor> <file>" to revert to 
the vendor version and ignore the local changes entirely.

This is creating an interesting situation now that we're trying to merge one 
of those branches back to the HEAD.  As an example, one file was at revision 
1.4 when the branch was made.  No changes have been made on the branch.  When 
I merge, cvs wants to merge changes between 1.1 and 1.4 into HEAD, while the 
HEAD is revision 1.1.1.2.

As it turns out, in every case where I used admin -b, CVS is using the wrong 
range of revisions to perform the merge.  Anyone else seen something like this?

--
Steve Madsen <address@hidden>
Tadpole Computer, Inc.  http://www.tadpole.com