Fix Alfredo's bug3 with contexts and work stealing. A worker pushing a ready task to another context should also pick a new worker if he isn't part of that other context.

src/sched_policies/work_stealing_policy.c

@@ -582,7 +582,6 @@ int ws_push_task(struct starpu_task *task)
 	struct _starpu_work_stealing_data *ws = (struct _starpu_work_stealing_data*)starpu_sched_ctx_get_policy_data(sched_ctx_id);
 	int workerid = -1;
 
-pick_worker:
 #ifdef USE_LOCALITY
 	workerid = select_worker_locality(task, sched_ctx_id);
 #endif
@@ -590,20 +589,15 @@ pick_worker:
 		workerid = starpu_worker_get_id();
 
 	/* If the current thread is not a worker but
-	 * the main thread (-1), we find the better one to
-	 * put task on its queue */
-	if (workerid == -1)
+	 * the main thread (-1) or the current worker is not in the target
+	 * context, we find the better one to put task on its queue */
+	if (workerid == -1 || !starpu_sched_ctx_contains_worker(workerid, sched_ctx_id))
 		workerid = select_worker(sched_ctx_id);
 
 	starpu_pthread_mutex_t *sched_mutex;
 	starpu_pthread_cond_t *sched_cond;
 	starpu_worker_get_sched_condition(workerid, &sched_mutex, &sched_cond);
 	STARPU_PTHREAD_MUTEX_LOCK_SCHED(sched_mutex);
-
-	/* Maybe the worker we selected was removed before we picked the mutex */
-	if (ws->per_worker[workerid].queue_array == NULL)
-		goto pick_worker;
-
 	record_data_locality(task, workerid);
 
 #ifdef HAVE_AYUDAME_H
@@ -615,11 +609,12 @@ pick_worker:
 	}
 #endif
 
+	STARPU_PTHREAD_MUTEX_LOCK(&ws->per_worker[workerid].worker_mutex);
 	_starpu_fifo_push_task(ws->per_worker[workerid].queue_array, task);
+	STARPU_PTHREAD_MUTEX_UNLOCK(&ws->per_worker[workerid].worker_mutex);
 	locality_pushed_task(task, workerid, sched_ctx_id);
 
 	starpu_push_task_end(task);
-
 	STARPU_PTHREAD_MUTEX_UNLOCK_SCHED(sched_mutex);
 
 #if !defined(STARPU_NON_BLOCKING_DRIVERS) || defined(STARPU_SIMGRID)