Blair Zajac <blair@orcaware.com> writes:
> > Actually, a really simple script that wrapped `svnlook log old-repos' and
> > `svnadmin setlog new-repos' could do the trick, I imagine. Both
> > operations are pretty darn fast.
>
> True, but on the current svn repository, that would mean forking at
> least 3000 processes, which is too many.
Nonono. I mean, don't sweat the log stuff during your incremental
dumps. Just, at the time you go to *load* your dumps, keep your old
repos around so you can fly over it, reading the logs, and updating
them on the newly-loaded repos. Obviously, this is no good if your
reason for re-loading your repos is because your old one got blown to
a cloud of pink mist by a photon torpedo.
> How about adding a new table to the repository that just stores a
> binary flag indicating for a particular revision if the log message
> has been modified. Then we can have several new svnadmin commands
> (these aren't the best names)
>
> svnadmin logs_get_set just prints those revision numbers
> with set bits
> svnadmin logs_clear_change takes a list of revisions to clear
> the bit
>
> This bit gets set whenever `svnadmin setlog' or a `svn commit' runs
> on the repos. Then the hot backup script runs `svnadmin logs_get_set'
> and only runs `svnlook log' on the indicated revisions.
/me revokes Blair's Developer role on tigris.org...
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Aug 22 17:37:19 2002