On Thu, 06 Dec 2007, David Glasser wrote:
...
> This bug does appear to exist on trunk. I think the basic problem is
> that we attempt to use the requested depth of the operation to tell us
> how deep we have to lock the working copy (ie, only lock the top if
> it's a shallow update), but if we're deleting a directory we need to
> have the whole directory locked.
>
> We could solve this by always locking all the way down, or by
> extending the lock when it is time to delete. (Man, I want a WC where
> all locks are deep because there's only one place where you need to
> lock them.) Karl, opinions?
>
> Repro recipe attached to issue 3039.
If we know it's a delete when taking out the lock, let's lock all the way
down. If we don't know until later, extend the lock when we figure out
we're doing a delete.
--
Daniel Rall
- application/pgp-signature attachment: stored
Received on Thu Dec 6 23:20:22 2007