On Tue, Mar 27, 2012 at 13:12, Stefan Sperling <stsp_at_elego.de> wrote:
> On Tue, Mar 27, 2012 at 12:36:21PM -0400, Greg Stein wrote:
>> On Tue, Mar 27, 2012 at 06:06, <stsp_at_apache.org> wrote:
>> >...
>> > +++ subversion/trunk/subversion/libsvn_wc/wc_db.c Tue Mar 27 10:06:41 2012
>> >...
>> > @@ -3657,7 +3657,7 @@ db_op_copy(svn_wc__db_wcroot_t *src_wcro
>> > dst_op_depth,
>> > dst_parent_relpath,
>> > presence_map, dst_presence));
>> > - if (is_move)
>> > + if (is_move && status != svn_wc__db_status_added)
>> > SVN_ERR(svn_sqlite__bind_int64(stmt, 7, 1));
>>
>> Hrm. What happens if the status is svn_wc__db_status_copied? (another
>> possible result from scan_addition) Don't you want to specifically
>> test for status_moved_here?
>
> As far as I understand, this could be the first time the node is moved.
> So it might be status normal.
>
> I thought briefly about performing a switch on 'status' to make sure
> we consider all cases, but decided to defer that for later and just
> fix the 'add' case for now.
>
> Considering 'svn cp A B; svn mv B C' is something a user could do,
> and given that this sequence doesn't currently result in useful DB state,
> additional work is definitely needed here.
Maybe leave a ### marker in there with a comment on current problems
and initial thoughts on future work?
I'd hate to lose your mind-state and have that code get shipped half-done.
Cheers,
-g
Received on 2012-03-27 19:23:12 CEST