philip@tigris.org writes:
> + if (anchor)
> + {
> + /* Closing and reopening the first session, forcing the anchor to
> + null, is required. Otherwise the ra_dav server will send
> + differences between the working copy and end_revision instead
> + of between start_revision and end_revision. This in itself is
> + not a problem, the editor could be modified to handle it, but
> + ra_local doesn't do this and there is no way to distinguish
> + which we are using. */
> + SVN_ERR (ra_lib->close (session));
> + SVN_ERR (svn_client__open_ra_session (&session, ra_lib, URL, NULL,
> + FALSE, FALSE, TRUE,
> + auth_baton, pool));
This kind of worries me. Here we have the two RA layers behaving in a
way that a caller of the RA interface can distinguish. That seems
like badness -- a sign that something deeper is wrong here. If the RA
layer is behaving inconsistently, then people can't just use it based
on the docs in svn_ra.h, they have to know more.
I realize that fixing this was beyond the scope of your change,
Philip, but if you have any analysis to add here, that would help. We
should get this inconsistency resolved soon.
-K
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:36:58 2006