One case where patience diff does much better

[Martin Pool]

On 26 Feb 2006, John Arbash Meinel <john at arbash-meinel.com> wrote:

> Wow, resurrecting an *old* thread. :) I saw someone accessed the branch,
> and wondered why. :)

I think I was away or too busy when it originally happened.  There's a
bug open about diffutils getting tied up on pathological input, and I
was talking to Bram about it.

> If it isn't on by default, I don't think it will be used. I think a
> reasonable test case would be to upgrade an existing bzr branch, and
> make sure all of the sha1 sums match before and after (and haven't changed).
> 
> That will ensure that even if the algorithm is performing sub-optimally,
> it is at least performing *correctly*.

That makes sense.  Should we put it in now before 0.8 or hold it off?
The diff algorithm by itself seems fairly low risk.

-- 
Martin