0.8 release plans

[Denys Duchier]

Robert Collins <robertc at robertcollins.net> writes:

> One low hanging fruit in particular is to teach the hash cache's 'scan'
> routine to stat in inode order [on linux].

it already does, no?

what can easily be done is to modify compare_trees to iterate in path order
instead of over ids (I played with that a bit and it helps a lot in some
cases). path order is typically closely related to inode order.  If your
worktree is really scrambled inode-wise, no trick can help you since you need to
go through dominating dentries in order to reach the one for a file.

Cheers,

--Denys