> The trick is to stop thinking of "replace_directory" as "I'm going
> to make a change to this directory". Sometimes, it just means "I
> need the baton for this directory, so I can change some children."
No, no, I already have that.
But let's say that I type "svn update foo/a/b foo/c/d foo/e" and there
are no changes in foo/c/d (which the filesystem can trivially
determine, because the node-revision of foo/c/d is the same in the
base and target revisions). Will the resulting edit do replace_dir()
calls on c and d just to convince the client to update its revision
numbers under c/d?
I also don't understand why it's extra work for the client to do the
version bumping itself. It shouldn't have to keep track of "committed
targets" (whatever those are in the context of an update); it just
needs the list of non-overlapping targets. If I "svn update foo/c/d"
and absolutely nothing has changed there, I still need to set the
version number for every directory and file under foo/c/d.
Received on Sat Oct 21 14:36:21 2006