Evan Driscoll wrote on Fri, Jan 27, 2012 at 10:31:32 -0600:
> On 1/27/2012 3:41, Daniel Shahaf wrote:
> > Evan Driscoll wrote on Thu, Jan 26, 2012 at 15:04:43 -0600:
> >> 1. Did removing rep-cache.db fix it, or is there still a potential for
> >> some latent repository corruption?
> >
> > Yes. rep-cache.db is used by the commit process. If that file was
> > silently corrupted, it's possible that the text: and props:
> > lines on node-revisions point to invalid data. (In English: versioned
> > properties, file contents, and symlinks targets may be affected.)
> >
> > I think the list archives and the developer docs explain this, so I'm
> > not going into detail here. But feel free to follow up (on list) if you
> > have questions that those don't cover.
>
> I ran 'svnadmin verify' (as Mark Cooke suggested) and it did not
> indicate any problems. Does that mean there aren't any (or that the
> probability is vanishingly small)?
No. If SQLite returned wrong answers when queried (as opposed to
declaring itself corrupt), it's possible that a commit process would
have set the contents of a committed revision to the contents of some
old file rather than to the newly-submitted contents.
'svnadmin verify' cannot identify such a corruption.
How likely it is --- judge for yourself.
>
> Evan
>
>
Received on 2012-01-27 17:44:10 CET