On Saturday 03 December 2005 09:14, Erik Huelsmann wrote:
> > > In short, I'm a bit stumped. I think we need more real data (the
> > > output from your tool is useful, but it only goes so far). What I'd
> > > really like is a complete un-recovered repository, and possibly an
> > > uncorrupted copy of the corrupt revision's file if possible. And a
> > > pony.
> >
> > I'd like that as well. I can work with one guy to get you access to a
> > complete, un-recovered repository, but he doesn't have an uncorrupted
> > copy of the file in question. :-( I'm not sure it's worth the trouble
> > unless we can tickle the failure ourselves, and I don't believe we'll be
> > able to do that without an uncorrupted copy of the file.
>
> Have either of you thought about the fact that the cause may not be
> fsfs, but the (new) xdelta code generating svndiffs? Maybe even looked
> into that already? The fact that we don't get any reports for BDB
> might be because you can't easily look into a bdb-rev.
I can't speak for Malcolm, but looking at the xdelta is something I've
definitely been meaning to check... I just haven't had time to get around to
it yet. However, If a BDB repo exhibited a problem with the svndiffs, I'd
expect that it would at error out with some sort of svndiff error. I haven't
really seen anything of the sort, which is why I've been focusing my
attention on FSFS.
-John
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sun Dec 4 02:20:31 2005