On Wed, 30 Mar 2005 08:04:39 -0500, Aaron Bentley
<[EMAIL PROTECTED]> wrote:
> The library-add algorithm prefers to use library revisions, and will
> only use cacherevs when
> - there's no ancestor revision in the library or
> - building from the ancestor would invove crossing an archive boundary

Thanks, that was the hint I was missing. 

The library was "almost" empty. I nuked it completely and now it's
behaving as expected. The strategy you describe for baz 1.3 (and that
I've seen mentioned in other threads) seems a  more scalable approach
for large projects. I have to confess that multi-GB libraries for a
project that is ~50MB of untarred source seemed a but much.

I think I now understand why the emphasis on "rolling" your archive
periodically (yearly), as each of those archive switches limits how
far back library reconstruction goes. I rather use cacherevs.

cheers,

martin


_______________________________________________
Gnu-arch-users mailing list
[email protected]
http://lists.gnu.org/mailman/listinfo/gnu-arch-users

GNU arch home page:
http://savannah.gnu.org/projects/gnu-arch/

Reply via email to