Re: [PATCH 0/3] sha256 fixes for filter-branch

> It actually does get covered. Dropping that case-arm means we'll fall
> through to the one below, which does _roughly_ the same thing with a
> bunch of extra warnings. But none of the tests actually check the error
> messages, so they don't notice.
> 
> Here's a series which fixes it, plus extra tests to notice the
> distinction.
> 
>   [1/3]: t7003: test ref rewriting explicitly
>   [2/3]: filter-branch: drop multiple-ancestor warning
>   [3/3]: filter-branch: drop $_x40 glob