On Tue, Apr 26, 2011 at 10:35 AM, Greg Stein <gst...@gmail.com> wrote: > On Tue, Apr 26, 2011 at 11:15, <hwri...@apache.org> wrote: >>... >> +++ subversion/trunk/subversion/libsvn_wc/props.c Tue Apr 26 15:15:24 2011 >> @@ -2018,6 +2018,7 @@ do_propset(svn_wc__db_t *db, >> svn_wc_notify_action_t notify_action; >> svn_wc__db_kind_t kind; >> svn_wc__db_status_t status; >> + svn_skel_t *work_item = NULL; > > Why is this initialized? It is set unconditionally below. > >>... >> @@ -2179,9 +2159,12 @@ do_propset(svn_wc__db_t *db, >> apr_hash_set(prophash, name, APR_HASH_KEY_STRING, value); >> >> /* Drop it right into the db.. */ >> - SVN_ERR(svn_wc__db_op_set_props(db, local_abspath, prophash, NULL, NULL, >> + SVN_ERR(svn_wc__db_op_set_props(db, local_abspath, prophash, NULL, >> work_item, >> scratch_pool)); >> >> + /* Run our workqueue item for sync'ing flags with props. */ >> + SVN_ERR(svn_wc__wq_run(db, local_abspath, NULL, NULL, scratch_pool)); > > This raises an interesting question: is it okay to run the workqueue > at any point in time? I believe the answer is "yes", but are there > situations where a caller may expect that the queue-run is under their > control, and the caller wants to wait a bit longer?
Good point: that is an interesting question. For this specific application, I don't think we need to worry about accidentally running a caller's existing queue prematurely, but I can certainly see places where that would be an issue. Unfortunately, we don't have a very good handle on where those might be. -Hyrum
