[MERGE] Fix error construction in KnitVersionedFiles

[Martin Pool]

On Tue, Sep 30, 2008 at 12:43 AM, John Arbash Meinel
<john at arbash-meinel.com> wrote:
> BB:approve
>
> Though I'll comment we often now have useless information when we get
> stuff like RevisionNotPresent. We don't know what index is missing it,
> or what file_id, etc we are looking for.
> This particular case might be okay, as the 'keys' parameter might give
> us enough information. (If it is a simple revision_id it is either an
> inventory or a revision text, if it is a tuple it is in the .tix, etc.)
>
> I don't have a good answer for it, as it happened when we factored out
> the "KnitAccess" information.

Yes, it does need to go deeper - with this you at least get a message,
but not all the data that would be useful.  I'm going to keep looking
in to it.  Thanks for the review.

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