Re: Easy way to flush history/versions of repository?

[Rob Helmer]

On Thu, Feb 14, 2002 at 11:27:22PM +0000, Stephan Jou wrote:
> Please excuse me if this is a simple question. I am a relative CVS newbie,
> and I read the CVS manual and FAQ, but could not find an obvious answer.
> 
> We have been using CVS for a year now, and have accumulated quite a large
> repository.  We have a major release coming up, and I would like to archive
> the current repository and start a new one, without the history/version
> baggage of the old repository.
> 
> As far as I can tell, the only way of doing this is to initialize a brand
> new repository, and then perform a CVS import.  I think this may be a bit
> time consuming (please correct me if I'm wrong) because we have a mixture of
> binary and text files, and the last time I did a CVS import, I found I had
> to explicitly tell CVS which ones were binary.

Ok, if you really want to do this, first use "cvs export" to get
the current sources without CVS metadata directories.

Backup your old repository, verify the backup against the original.
Remove your current repo.

Create a new repository using "cvs init".

Checkout the CVSROOT module of your new repository. Add lines
to the "cvswrappers" file for any binaries you intend
on keeping in CVS ( for example, this is what I use for *.gif files ) :

*.[Gg][Ii][Ff]   -k 'b' -m 'COPY'

Check "cvswrappers" back in. 

Now you can add your module, and any files with extensions mentioned
in CVSROOT/cvswrappers will be treated as binaries by CVS.

Does anyone recommend "cvs add" for this vs. "cvs import" if vendor
branches are not being used? Reasons?



Thanks,
Rob Helmer
Namodn