On Tue, 10 Apr 2007, C. Michael Pilato wrote:
> While working on this:
>
> > Author: cmpilato
> > Date: Tue Apr 10 11:27:26 2007
> > New Revision: 24515
> >
> > Log:
> > Fix some merge-tracking OBOEs.
>
> I noticed this:
>
> > @@ -1979,7 +1981,8 @@
> > continue;
> > SVN_ERR(svn_wc__entry_versioned(&child_entry, child_wcpath, adm_access,
> > FALSE, pool));
> > - child_repos_path = child_wcpath + strlen(merge_cmd_baton->target) + 1;
> > + child_repos_path = child_wcpath + \
> > + (merge_target_len ? merge_target_len + 1 : 0);
> > child_url = svn_path_join(parent_wc_url, child_repos_path, pool);
>
> And said to myself, "Self, should that svn_path_join() also be URI-encoding
> the child_repos_path portion? And if so, how many other places in the new
> merge tracking code contain similar mistakes?"
>
> Self couldn't answer quickly, but suggested I pose the question to Dan and
> the dev-list. So there ya go.
Both code sections you tweaked here were introduced to handle sub-tree
merge info. Peter Lundblad is currently working on re-working the
algorithm which deals with sub-tree merge info. Both here, and IIRC,
in the copy/move code, we do URL manipulation.
- application/pgp-signature attachment: stored
Received on Tue Apr 10 21:12:32 2007