> From: Karl Fogel [mailto:kfogel@newton.ch.collab.net]
> Sent: 24 September 2002 15:17
> Philip Martin <philip@codematters.co.uk> writes:
>> I don't really like this approach, I'd prefer a command that contacts
>> the new repository, to ensure that we don't get invalid text-bases.
Me too.
And having a simple GUID on the repos would help a great deal. Contact
the new repos, compare GUIDs and abort if they're not equal.
> > The current patch would be the only client command that, by design,
> > will allow the user to produce a corrupt working copy.
>
> Philip, if it makes you feel any better...
>
> 1) When issue #689 is finished (repos<->wc checksums), such
> corruption will at least be instantly noticed, and not spread
> beyond that working copy.
>
> 2) If we don't supply this command, people *will*, I guarantee it,
> start writing their own scripts to modify the entries file :-).
> We can either pretend it's not going to happen, or we can supply
> the functionality ourselves and thus leave the door open to add
> more safety checks later.
That's true. But please file an issue on this the moment this code is
committed. Without the security checks, but with the functionality
committed it is a showstopper for 1.0 IMO.
Sander
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Sep 24 15:52:59 2002