"Bert Huijben" <b...@qqmail.nl> writes: > Please add a note that the anchor_abspath argument should be removed > after we move to a single database. In the single database there is no > information on a path stored in its parent, so you never have to lock > an anchor for updating/deleting/adding a child.
Done. > I didn't review this issue, but maybe this code relying on a lock of > the parent should just lock the parent instead of requiring all the > callers of this new functions to think about an ancestor. The calling code cannot simply lock the parent because the parent might not be part of the working copy. I suppose the function svn_wc__acquire_write_lock could always lock the anchor and hide that from the caller, but we have to ensure that the release code releases the same set of locks, and that is difficult when the original target path may no longer be versioned. It might involve caching the anchor path soemwhere (in the PDH?). I think it's simpler to have the calling code hold it because 1.6 already knows when we need to use anchors. -- Philip
