[cvsnt] Mergepoint issues on 2.5.0.3 b2382

[Tony Hoyle]

Richard Wirth wrote:
> Hello Arthur, hello Andreas,
> 
> May be Tony has tested a different solution than provided in the patch
> by Andreras?
> 
> The patch works as follows:

It's not a patch really it's a reversion to the code that was removed 
because it was broken (because it verifiably caused data loss).  The 
overriding rule when considering functions in cvsnt is that it must 
*not* lose data.  Anything that does is either fixed or removed. 
Convenience is very low down the list compared to that.

The problem is if you merge A->B you cannot assume that A==B.

Merge A->B means that B contains A (possibly changed a bit due to 
conflict resolitution, so B might not contain A at all in the most 
pathalogical case).  It does not mean that A contains B.

Let's go through this again.  For the final time.

Create file - Revision 1.1
Commit a revision to A - Revision 1.1.2.1
Make changes on A - Revision 1.1.2.2
Make changes on HEAD - Revision 1.2
Merge HEAD->A - diff(1.1:1.2)->1.1.2.3
Merge A->HEAD

With one directional mergepoints:
diff(1.1:1.1.2.3)->1.2
With bidirectional mergepoints:
diff(1.1.2.3:1.1.2.3)->1.2 (ie. a null diff, losing the changes in 
1.1.2.1 and 1.1.2.2).

In the bidirectional case the revision 1.1.2.1 and 1.1.2.2 changes are 
lost - and that's quite a trivial example.  Expand it to multiple 
branches and revisions and it gets hard to track.

Tony