Re: svnadmin dump generates (hopefully) spurious warnings - Loading this dump into an empty repository will fail

Charles Butterfield wrote:
> Clearly the dump does a lot of processing to create a valid
> self-contained image for the first rev it emits. I assumed it kept
> track of that processing and suitability modified the remaining revs
> too, so that the resulting dump file could be loaded into a virgin repo.
> It sounds like that may not true.

Well, it's like this (AIUI):

A non-incremental dump starts with a full-text representation of all files at
the initial rev, throwing away all history from before that rev. This latter
issue is the problem for you, since some file or directory in r2573 is a copy of
something from r1926. But you threw away the history before r2000, so there is
no way to figure out what file/revision could be safely substituted.

In order to do what you want, svnadmin dump would have to walk through the
entire history twice:

1) to see what files depend on revisions prior to the lower cutoff rev and
calculate which rev could be substituted;

and

2) emit a modified dumpfile with the appropriate substitutions.

Currently, 'svnadmin dump' just walks the repository from lowest to highest rev
in a strictly linear fashion and just emits the sparse copy-from information
that exists in the repo itself.

Hope this explains things...

John

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: users-help_at_subversion.tigris.org
Received on 2008-02-21 02:23:38 CET