[RFC] Dirstate blowing up on None as a minikind fingerprint.

[Wouter van Heyst]

On Fri, Jul 20, 2007 at 01:19:51PM -0500, John Arbash Meinel wrote:
> Wouter van Heyst wrote:
> > On Sat, Jul 14, 2007 at 10:41:04AM +0200, Wouter van Heyst wrote:
> >> Hey all,
> >>
> >> on the nested-trees branch a test was failing with funny dirstate
> >> errors, it turns out because the reference_revision for the entry is
> >> None and that got fed into DirState.save() which proceeded to blow up.
> >>
> >> The attached fix gets rid of that, but it's not directly tested
> >> (TestCompositeTree.test_apply_inventory_delta does indirectly).
> > 
> > Bweh, only now do I notice the missing attachement.
> > 
> > Wouter van Heyst
> > 
> 
> +1 if you had a test case which exposed this.

Right, at the time I just wanted to get it out in the open (and stupidly
didn't attach anything). Yesterday I sent it mostly for vila's benefit.
At the moment I'm working on this bug #127115, so will hopefully have
tests today.

> It seems to only happen if you have a nested subtree. So only people using
> --dirstate-with-subtree would ever see it.

Aye.

> It shouldn't be too difficult to create a tree which has an inventory
> referencing a subtree with no 'reference_revision', and then just calling
> DirState.set_state_from_inventory() with that.

Will do.

Wouter van Heyst