On Wed, 2008-11-05 at 05:23 +0100, Neels J. Hofmeyr wrote:
>
> Julian Foad wrote:
> > On Mon, 2008-11-03 at 11:30 +0000, Julian Foad wrote:
> >> On Sun, 2008-11-02 at 20:40 -0800, neels_at_tigris.org wrote:
> >>> Author: neels
> >>> Date: Sun Nov 2 20:40:11 2008
> >>> New Revision: 34018
> >>>
> >>> Log:
> >>> Remove the dir_closed() diff callback, it is now unused everywhere.
> >> Hi Neels.
> >>
> >> Reasons for removing the USES of this callback: because there's nothing
> >> we want them to do at the moment. Agreed. Thanks for pruning that dead
> >> code.
> >>
> >> Reasons for removing the API and DRIVER of this callback: because we
> >> don't believe it's a sensible and supportable feature, or we don't
> >> believe that there can be any good use for it in the 1.6.x time frame.
> >> I'm not aware of any such reasons. I think the dir_closed() callback is
> >> a sensible complement to dir_opened(), and removing it may cause us
> >> problems when we want to fix tree conflict handling bugs in 1.6.x time
> >> frame.
> >>
> >> What do you think?
>
> Well, using the TREE_CONFLICTED parameter is actually moving away from
> having to have a dir_closed() callback. That's how I perceive the nature of
> the diff callbacks: they don't have a child_baton passed around, so they are
> supposed to get all the information on a node at the same time (this is
> related to my intention to change added_dir() to also include prop changes,
> which is the single exception as far as I've seen).
Ah, now there you have a very good point. That sounds like a good change
and then I will agree (er, maybe, at least more so) that "dir_closed" is
redundant.
> So I'm pretty confident we won't need the dir_closed() callback, at least
> not for tree-conflicts. (Still, the dir_opened() is needed because it checks
> an existing directory before going through the child nodes.)
OK.
There's one other little "gotcha": as I noticed a few days ago, there
are cases where one of the diff producers opens a directory and a file
and then doesn't send any changes for the file. We haven't defined
whether that's an allowable mode of operation of the diff callbacks, but
given that it happens I think we might have to make the diff consumers
aware of it. But it's not a particularly ocmmon case, I think, so we're
probably OK to ignore it for now.
- Julian
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe_at_subversion.tigris.org
For additional commands, e-mail: dev-help_at_subversion.tigris.org
Received on 2008-11-05 13:36:46 CET