update on stack merging

[Martin Pool]

On Wed, Jul 9, 2008 at 2:18 AM, John Arbash Meinel
<john at arbash-meinel.com> wrote:
> Your "questionable" change seems to be changing the meaning of the test
> quite a bit. Specifically, instead of testing that the target repository
> doesn't have the old revisions, you are testing that the source
> repository doesn't have the new revisions.
>
> While I can agree that it might be good to test for that, it means that
> a regular branch will work just fine, and doesn't need any stacking support.

You're right, that is a bit questionable. :_)

> I thought there was a way to ask a stacking repository for only the data
> that it has locally. If there isn't we should probably add one and use
> it here.

It's like this: stacking is configured by the branch.  If you open the
repository through the branch you get an object configured to look at
the fallback repository.  If you open the repository of that directory
itself then you can see what's stored in it.

Apparently the reason why this is failing is that there is no whitebox
test that sprout(stacked=True) does not copy across the data, and in
fact it looks like currently it does copy it.  Robert says I've merged
his latest code, so I will try to fix this myself.  There may be a fix
in Aaron's stacking-policy branch though I think not.

-- 
Martin <http://launchpad.net/~mbp/>