Re: cvs and new distributions

[YONETANI Tomokazu <qhwt+dragonfly-kernel@xxxxxxxxxx>]

On Mon, Sep 20, 2004 at 11:45:20AM -0700, Matthew Dillon wrote:
> 
> :> On Mon, Aug 30, 2004 at 07:26:13AM +0200, Jeroen Ruigrok/asmodai wrote:
> :> > -On [20040830 06:12], YONETANI Tomokazu (qhwt+dragonfly-kernel@xxxxxxxxxx) wrote:
> :> > >Actually there's no simple way in CVS to add unversioned tree into
> :> > >a repository other than by using `cvs import'.
> :> > 
> :> > So the question is, are we going to use cvs import for this again?
> :> 
> :> Why not? I don't see any problem in using `cvs import' to actually
> :> import an external codes to our tree. Or did I miss an earlier discussion
> :> against using `cvs import' at all?
> :
> :Did I receive any response to this message?
> 
>     cvs import can be used, though what will happen is that we will
>     get a mess of random vendor tags and release tags.  That's why
>     I prefer cvs add over cvs import.  But I don't think it's a big
>     deal.

If that matters, you can get rid of the mess by issuing `cvs admin'
from the check out of the imported directory:

  cvs admin -b -o release_tag -n release_tag -n vendor_tag

This is possible if
A) the committer is permitted to use `cvs admin' command,
B) multiple vendor imports over the same directory is not needed, and
C) no one wants to edit files in the directory during the race
   between `cvs import' and `cvs admin'

This method relies on `cvs admin' command, but it still has advantages
of not having to hack up our cvs command or commit mail script.