[MERGE] iter_changes-based merge

[Aaron Bentley]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Martin Pool wrote:
> Could you please document interesting_ids and interesting_files?

Sure.

> Could you please say something there about why - if I understand
> correctly, it's that this is meant to enable future performance
> improvements.

That's half of it.  The other half is greater consistency, and less
reliance on Tree.inventory.

> +    def _three_way(base, other, this):
> +        #if base == other, either they all agree, or only THIS has
> changed.
> +        if base == other:
> +            return 'this'
> +        if this not in (base, other):
> +            return 'conflict'
> +        # "Ambiguous clean merge"
> +        elif this == other:
> +            return "this"
> +        else:
> +            assert this == base
> +            return "other"
> 
> I can't see how that assertion could ever fail at runtime, so you might as
> well remove it and avoid the extra comparison.

If I thought it was likely to fail in normal use, I'd have raised a
proper error instead of asserting.  It was partly there for clarity, so
I'll remove it and stick a comment in.

> This looks reasonable to me but because I'm not intimately familiar with
> this code a review from someone else might be good.

Okay, I'll hold off merging for a bit.

Aaron
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFGk6ib0F+nu1YWqI0RAl1ZAJ9IcU9jdR75SK3BJuUlKNcHXxImKQCfRGjK
z3pD46h02ZEBFHIiYGaq12Y=
=dSLI
-----END PGP SIGNATURE-----