IRQ Handlers with LL drivers

[magene]

I generally prefer the STM Low Level drives and chose that driver family for most peripheral in CubeMX Project Manager->Advanced Settings->Driver Selector.  There seem to be one disadvantage: the IRQHandlers are not generated as weak which I think happens with the CubeMX generated HAL IRQHandlers.  So with the LL drivers, I have to add a line of code to the CubeMX generated stm32h7xx_it.c file.  Like this:

void USART1_IRQHandler(void)
{
  /* USER CODE BEGIN USART1_IRQn 0 */
        myUSART1_IRQHandler();
  /* USER CODE END USART1_IRQn 0 */
  /* USER CODE BEGIN USART1_IRQn 1 */

  /* USER CODE END USART1_IRQn 1 */
}

Is there a way to tell CubeMX to generate the IRQHandler stubs as weak?

And what does the CubeMX Project Manager->Advanced Settings->Register CallBack disable/enable setting do? I've generated code with this enabled and disabled and don't see any obvious difference.

Thanks

[TDK]

IRQ handlers are never generated as weak--not in HAL or LL.

> Is there a way to tell CubeMX to generate the IRQHandler stubs as weak?

There are weak definitions in startup_*.s file that would conflict with any other weak definitions, so this is not a possibility.

The callback stuff has to deal with how HAL handles callbacks within its machinery, not how IRQ handlers are defined. Not going to help you here.

[magene]

@TDK Thanks for the help. Given that information, it seems like I have 2 options.  Keep doing what I'm doing by adding a line of code to every IRQHandler in the stm32H7xx_it.c file that CubeMX generates. Or remove stm32H7xx_it.c from my project and write my own.

A better question is how would you handle this situation?

Thanks again

[TDK · Accepted solution]

Yes, those are effectively your options.

The HAL has its own callbacks for transfer complete, or when a character is received, or whatever. I would implement and use those if you don't need computational speed.

If you do need computational speed, I would add the user code to the IRQ function in stm32_*_it.c directly. Calling your own function in there with user code is also fine, but will be a little slower.

Really up to you, no wrong method.