Re: svn commit: r27587 - in branches/svnsync_ra_serf/subversion: include libsvn_ra libsvn_ra_serf svnsync

On Nov 4, 2007 1:14 PM, David Glasser <glasser@davidglasser.net> wrote:
> On 11/4/07, lgo@tigris.org <lgo@tigris.org> wrote:
> > Author: lgo
> > Date: Sun Nov 4 12:33:37 2007
> > New Revision: 27587
> >
> > Log:
> > Expand the functionality of svn_ra_replay_range: make it provide the revision
> > properties together with the actual content.
>
> Good idea!
>
> > - /* The actual copy is just a replay hooked up to a commit. */
> > -
> > - SVN_ERR(svn_ra_get_commit_editor2(rb->to_session, &commit_editor,
> > + /* The actual copy is just a replay hooked up to a commit.
> > + Include all the revision properties from the source repositories, except
> > + those starting with 'svn:', those are not guaranteed to get through
> > + the editor anyway. */
> > + filtered = filter_props(&filtered_count, rev_props, NULL, SVN_PROP_PREFIX,
> > + pool);
> > + SVN_ERR(svn_ra_get_commit_editor3(rb->to_session, &commit_editor,
> > &commit_baton,
> > - "", /* empty log */
> > + filtered,
> > commit_callback, rb->sb,
> > NULL, FALSE, pool));
>
> Is it really OK to give no value (as opposed to "") for svn:log?
> That's not what svn_ra_get_commit_editor3's doc implies. I know you
> clean it up later, but still...

In fact, it's *not* OK: if the target is ra_svn, this causes a SEGV
(libsvn_ra_svn/client.c(ra_svn_commit) in the first
svn_ra_svn_write_tuple).

Maybe you should just explicitly filter out svn:sync:*, svn:author,
and svn:date.

--dave

-- 
David Glasser | glasser_at_davidglasser.net | http://www.davidglasser.net/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org