Task should be not blocked if flag which are we waiting for has been already set

[lkasperowicz]

Without changes:

void TestTask(void *argument)
{
  /* Infinite loop */
  for(;;)
  {
    uint32_t flags = 0;
    osThreadFlagsSet(testTaskHandle, 1<<0);
    flags = osThreadFlagsWait(1<<0, osFlagsWaitAny, osWaitForever );
    osDelay(1);
    osThreadFlagsSet(testTaskHandle, 1<<0);
    osThreadFlagsSet(testTaskHandle, 1<<1);
    flags = osThreadFlagsWait(1<<0, osFlagsWaitAny, osWaitForever );
    osDelay(1);
    flags = osThreadFlagsWait(1<<1, osFlagsWaitAny, osWaitForever ); // task is blocked here even though 1<<1 flag has been 
    // already set
    osDelay(1);
  }
}

To be more consistent in the way of handling Flags (whether Thread Flags or Event Flags is used) osThreadFlagsWait should not block a task when flag which we are waiting for has been already set. For instance, the same code but using Event Flags

void TestTask(void *argument)
{
  osEventFlagsId_t evt_id = osEventFlagsNew(NULL);

  for(;;)
  {
    uint32_t flags = 0;
    osEventFlagsSet(evt_id, 1<<0);
    flags = osEventFlagsWait(evt_id, 1<<0, osFlagsWaitAny, osWaitForever );
    osDelay(1);
    osEventFlagsSet(evt_id, 1<<0);
    osEventFlagsSet(evt_id, 1<<1);
    flags = osEventFlagsWait(evt_id, 1<<0, osFlagsWaitAny, osWaitForever );
    osDelay(1);
    flags = osEventFlagsWait(evt_id, 1<<1, osFlagsWaitAny, osWaitForever ); // task is not blocked, value 2 is returned
    osDelay(1);
  }
}

When osThreadFlagsWait waits only for (1<<0) flag but (1<<1) flag is also set then inside FreeRTOS xTaskNotifyWait function notification state is marked as taskNOT_WAITING_NOTIFICATION.

When next osThreadFlagsWait waits for (1<<1) flag, task goes to BLOCKED state even though (1<<1) flag is actually set and task's notification value is 2 (1<<1) but task's notification state is taskNOT_WAITING_NOTIFICATION.

To resolve this issue task's notification state should be marked as taskNOTIFICATION_RECEIVED again after previous osThreadFlagsWait.

[VladimirUmek]

Hi Lukasz,

many thanks for your investigation, reproducer and the bugfix implementation! I've checked the behavior and it looks good.
I will perform some additional tests and with a bit of coding style alignment your implementation will be merged!