RCS Keyword Conflicts During Merge (was RE: [cvsnt] Re:minorBugin update in Build 1927)

David Hauck davidh at netacquire.com
Tue Apr 26 23:59:57 BST 2005


Hi, one other note: I do see that it is specifically recommended (ala Open
Source Development with CVS 3e, pg 164) that -kk *not be used* during merges
when binary files are involved. It just seems unfortunate given that -kk
would seem to be orthogonal to -kb (i.e., could just be ignored) in this
context.
Comments? Thanks, -David

> Hi Tony,
>
> > Hi Tony,
> >
> > > David Hauck wrote:
> > > > anyways. I've often wondered if CVS merge could be optimized
> > > somehow in this
> > > > regard to eliminate/ignore RCS keyword conflicts; they're a
> > pain, in my
> > > > mind, to deal with currently.
> > > >
> > > cvs does try to minimise this, and is reasonably successful a
> lot of the
> > > time.  There are some keywords like $Log$, which simply can't be done
> > > automatically though.
> > >
> > > If you're 3-way merging one of the files (the modified file from the
> > > user) is going to have the keywords in whatever you do, so it's nearly
> > > impossible to handle that.  A branch merge to a clean sandbox
> can avoid
> > > a lot of conflists though.
> >
> > I only ever do branch merges (i.e., -j <tag> -j <tag> representing one
> > branches changes merged onto an active sandbox representing
> > another branch).
> > Whenever a file being merged has changed in both branches I see the RCS
> > keyword conflicts.
>
> I wanted to report something that I just encountered with my latest branch
> merge. I recently moved over to using -kk with the update merge
> command and
> this (seemed) to work great. At least all the keyword conflicts
> disappeared
> and I no longer had to manually correct these. *However*, I found out that
> handling of binary files is less than optimal in this situation. In
> particular the update merge command ends up performing non-binary
> operations
> on binary files (line endings are converted) as a result and this
> messes up
> subsequent commits of these files if they're marked as modified in the
> merge.
>
> I would have thought that the -kk option was orthogonal to
> repository files
> tagged as binary (i.e., doesn't the binary setting take precedence, or,
> alternatively, shouldn't the -kk option be meaningless for binary files)?
> I'm using 2.0.58d (client/server) - perhaps this has been fixed in later
> versions?
>
> Regards,
> -David
>
> > -David
> >
> > > Tony
> > > _______________________________________________
> > > cvsnt mailing list
> > > cvsnt at cvsnt.org
> > > http://www.cvsnt.org/cgi-bin/mailman/listinfo/cvsnt
> >
> > _______________________________________________
> > cvsnt mailing list
> > cvsnt at cvsnt.org
> > http://www.cvsnt.org/cgi-bin/mailman/listinfo/cvsnt
>
> _______________________________________________
> cvsnt mailing list
> cvsnt at cvsnt.org
> http://www.cvsnt.org/cgi-bin/mailman/listinfo/cvsnt




More information about the cvsnt mailing list