Re: svn commit: r1100169 - /subversion/trunk/subversion/libsvn_wc/adm_ops.c

rhuijben_at_apache.org writes:

> Author: rhuijben
> Date: Fri May 6 11:21:15 2011
> New Revision: 1100169
>
> URL: http://svn.apache.org/viewvc?rev=1100169&view=rev
> Log:
> Following up on r1100163, add a write lock check to the revert handling.
>
> * subversion/libsvn_wc/adm_ops.c
> (new_revert_internal): Expect a write lock on the directory containing
> this node, which might be the directory itself for the wcroot.
>
> Modified:
> subversion/trunk/subversion/libsvn_wc/adm_ops.c
>
> Modified: subversion/trunk/subversion/libsvn_wc/adm_ops.c
> URL: http://svn.apache.org/viewvc/subversion/trunk/subversion/libsvn_wc/adm_ops.c?rev=1100169&r1=1100168&r2=1100169&view=diff
> ==============================================================================
> --- subversion/trunk/subversion/libsvn_wc/adm_ops.c (original)
> +++ subversion/trunk/subversion/libsvn_wc/adm_ops.c Fri May 6 11:21:15 2011
> @@ -1590,6 +1590,20 @@ new_revert_internal(svn_wc__db_t *db,
> {
> SVN_ERR_ASSERT(depth == svn_depth_empty || depth == svn_depth_infinity);
>
> + /* We should have a write lock on the parent of local_abspath, except
> + when local_abspath is the working copy root. */
> + {
> + const char *dir_abspath;
> +
> + SVN_ERR(svn_wc__db_get_wcroot(&dir_abspath, db, local_abspath,
> + scratch_pool, scratch_pool));
> +
> + if (svn_dirent_is_child(dir_abspath, local_abspath, NULL))
> + dir_abspath = svn_dirent_dirname(local_abspath, scratch_pool);
> +
> + SVN_ERR(svn_wc__write_check(db, dir_abspath, scratch_pool));
> + }
> +
> SVN_ERR(svn_wc__db_op_revert(db, local_abspath, depth,
> scratch_pool, scratch_pool));

I suppose that is sort of how the checking worked in 1.6, but is it
right for 1.7? Perhaps 1.7 should be checking for the write lock inside
the SQLite transaction?

There is the question about depth as well, a recursive revert needs a
recursive lock.

-- 
Philip