On Fri, Mar 25, 2005 at 12:50:51AM -0500, Nick Mathewson wrote:
> On Thu, Mar 24, 2005 at 09:03:22PM -0500, Geoffrey Goodell wrote:
> > see attached
>
> This patch is incorrect. It will break in two ways:
>
> > --- control.c.orig 2005-03-24 20:58:17.000000000 -0500
> > +++ control.c 2005-03-24 20:59:16.000000000 -0500
> > @@ -174,7 +174,6 @@
> > connection_t **conns;
> > int n_conns, i;
> >
> > - global_event_mask = 0;
> > get_connection_array(&conns, &n_conns);
> > for (i = 0; i < n_conns; ++i) {
> > if (conns[i]->type == CONN_TYPE_CONTROL &&
>
> This will make global_event_mask contain every event that every
> controller was ever interested in: no good!
>
> > @@ -267,7 +266,7 @@
> > for (i = 0; i < n_conns; ++i) {
> > if (conns[i]->type == CONN_TYPE_CONTROL &&
> > conns[i]->state == CONTROL_CONN_STATE_OPEN &&
> > - conns[i]->event_mask & (1<<event)) {
> > + global_event_mask & (1<<event)) {
>
> This will make every controller receive every event that any
> controller is interested in.
>
> I think you might be confused about the semantics of SETEVENTS: it
> takes a list of all the events that the controller is currently
> interested in: events not listed are cleared. I've updated
> control-spec.txt to reflect this.
Hm, you are right. I had no idea about the clearing, though I suppose
that that would explain the alarming lack of unsubscription messages!
Thanks,
Geoff
Attachment:
signature.asc
Description: Digital signature