[PATCH 1 of 3 V2] ancestor: a new algorithm for ancestors of a set of revs not in another set that is faster when the sets are close

[Greg Ward]

On 21 November 2012, Siddharth Agarwal said:
> # HG changeset patch
> # User Siddharth Agarwal <sid0 at fb.com>
> # Date 1353546962 28800
> # Node ID 562ed7c273988f91645be6189b51ab3119752c09
> # Parent  536f5df951bece49b760392fd543e8082489b7d0
> ancestor: a new algorithm for ancestors of a set of revs not in another set that is faster when the sets are close

To my surprise, http://mercurial.selenic.com/wiki/ContributingChanges
says nothing about how long the first line of a commit message should
be. But I'm pretty sure 80 chars is the absolute maximum beyond which
thou shalt not go, and 75 chars is much better.

How about:

  ancestor: add faster algorithm for computing difference of ancestor sets

Then immediately qualify in the next paragraph that it's geared
towards the "sets are close" case that is common in rebase. Nice that
it also speeds up incoming.

       Greg