[cvsnt] How to deal with too large CVS database?

[Gerhard Fiedler]

Bo Berglund wrote:

> Pedja:
>> We had an idea to lock currnet repository, make it read only and create
>> new one with packed database which contains only the last versions of
>> each file in source. That would allow us to access previous versions
>> but continue development with fresh, and small source database. 
>> 
>> We cannot find way to achieve this but preserve file version numbers
>> within source (if we create new repository and put new source in it,
>> file version numbers are reset). Is it possible at all to preserv
>> version numbers or pack database by keeping only the last version of
>> each file within source?

> There is an cvs admin command to delete specific revisions of a file:
> 
> cvs admin -o <range of revisions to delete> filename

I think you can (and have to) do this with custom shell scripts or some
other enhancement to cvs. 

For example, you could create a copy of your repository and run a script on
it that goes through every single file and removes all revisions that don't
have a tag on it. Wincvs has already a number of Python scripts and
functionality that perform at least sub-tasks of this.

Gerhard