--On Saturday, March 12, 2005 12:11 PM -0600 kfogel@collab.net wrote:
> I think r13366 was seductive, but ultimately not the best way to go.
> As Mike Pilato points out, some of the svn_fs functions still
> (unavoidably) bypass the repository format check, so it's not
> foolproof. Furthermore, the locking tables/directories and the
> locking hooks are created unconditionally the first time 1.2 uses a
> legacy repository. If we don't also upgrade the format number, the
> repository would then be in a weird in-between state, at least to
> anyone who bothered to look closely.
Oh, that's an interesting point. So, we weren't updating the format even
though we added the tables and hooks at open time? Ok. That's different and
a reason why the 'compromise' of delaying the format upgrade isn't viable.
> My preference right now is to leave r13384 as it is, and put a
> prominent warning in the release notes, which admins ought to be
> reading anyway. We can, if it ever becomes an issue, easily provide a
> 'downgrade' script, but I doubt there will be much demand.
I would prefer keeping the status quo as well. Additionally, as we run the
release candidates for 1.2, we can judge how much of a real-world issue this
will become. -- justin
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Mar 12 20:57:07 2005