Re: [Issue 2992] fsfs svnadmin recover may leave repository in a broken state

The issue isn't crashes, it's backups. current was written and the
user responded to, but the backup was written with an older version of
current.

-Dan C

On Mon, Feb 4, 2008 at 9:22 AM, David Glasser <glasser_at_davidglasser.net> wrote:
> I haven't had a chance to read the rest of this issue/thread, but:
>
> FSFS is pretty clear on this point. Crashing between moving
> revprops/FOO into place and moving current into place is quite
> possible. However, the semantics there are also very well defined:
> the commit *didn't happen*. If current hasn't been updated, then the
> commit hasn't happened yet; no user should have been told that that
> revision was committed successfully; etc. I don't see why we should
> add "recovery" code that makes it look like a commit has happened when
> it never did!

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-02-04 19:13:22 CET