* re-enable poll_recalc so that it triggers the recalc of the signals

properly.  This fixes the random hangs that we were seeing this morning
  on Linux that were a result of fixing the thread deadlock yesterday.
  (worked great on my OS X box, which uses select() instead of poll()).

This commit was SVN r6730.

opal/event/poll.c

@@ -82,11 +82,7 @@ static struct pollop {
 static void *poll_init	(void);
 static int poll_add		(void *, struct opal_event *);
 static int poll_del		(void *, struct opal_event *);
-#if 0
-/* Open MPI: JMS As far as I can tell, this function is not used
-   anywhere */
 static int poll_recalc	(void *, int);
-#endif
 static int poll_dispatch	(void *, struct timeval *);
 
 const struct opal_eventop opal_pollops = {
@@ -94,7 +90,7 @@ const struct opal_eventop opal_pollops = {
 	poll_init,
 	poll_add,
 	poll_del,
-	NULL,
+        poll_recalc,
 	poll_dispatch
 };
 
@@ -117,9 +113,6 @@ poll_init(void)
  * recalculate everything.
  */
 
-#if 0
-/* Open MPI: JMS As far as I can tell, this function is not used
-   anywhere. */
 static int
 poll_recalc(void *arg, int max)
 {
@@ -130,7 +123,6 @@ poll_recalc(void *arg, int max)
 	return (0);
 #endif
 }
-#endif
 
 static int
 poll_dispatch(void *arg, struct timeval *tv)