RE: Preserving annotations across merges?

[Arthur Barrett]

Chris,

> Now
> the obvious answer to this problem is that we should make 
> sure all Jtest
> issues are resolved before the merge

Yes - it occurs to me that if the patch was tested on the "original"
before it was merged, that if a "bug" was discovered in the "merged"
version (but not the original) that that would indeed be the
responsibility of the "merger" to fix.  I can't see how Jtest can
determine that one thing needs to be done in one case and another in
another.

> I was hoping that 
> there would be
> some clever way to have the merge also merge the annotations, so
> ownership would move across merges like the changes. 

That change would require the "original" author to be stored in
Entries.Extra along with the Mergepoint - it's possible in CVSNT -
probably also possible in CVS, though it'd be a little harder I suspect
because there is no Entries.Extra.  Then the commit could record this
author in the RCS file for the change (a little dangerous I think) or
store it in a mergeauthor attribute in the RCS file that annotate can
read later on.  Either you could make the change to the source or you
could get someone to do it for you (hint - the company I work for does
CVS professional support that can also include custom features).

I personally think the "bug" is in Jtest.  It is the one doing things
this (incorrect) way - it should look to see if this was a "merge" (by
looking for a mergepoint in the log) and if it is (based on some
preference) then finding the author of the original change.

A change to the CVSNT annotate code to output something to indicate if a
change was part of a merge or not could be done quite easily I think.
CVS doesn't have mergepoints so I'll assume we are talking about CVSNT
from here on.

A third alternative is to modify the annotate code with an option so
that the author is displayed in the author field - unless there is a
mergepoint - in which case it displays MERGE or something.  

Regards,


Arthur Barrett