blackpill-f4: #define PLATFORM_HAS_POWER_SWITCH leads to error: implicit declaration of function 'adc_start_conversion_direct'

[lenvm]

Using 9fe5889, adding #define PLATFORM_HAS_POWER_SWITCH to blackpill-f4.h and compiling using the command make PROBE_HOST=blackpill-f401cc leads to the following error.

  CC      platforms/common/stm32/timing_stm32.c
platforms/common/stm32/timing_stm32.c: In function 'sys_tick_handler':
platforms/common/stm32/timing_stm32.c:110:25: error: implicit declaration of function 'adc_start_conversion_direct'; did you mean 'adc_start_conversion_injected'? [-Werror=implicit-function-declaration]
  110 |                         adc_start_conversion_direct(ADC1);
      |                         ^~~~~~~~~~~~~~~~~~~~~~~~~~~
      |                         adc_start_conversion_injected
cc1: all warnings being treated as errors
make[1]: *** [timing_stm32.o] Error 1
make: *** [all] Error 2

This error was not present a couple months ago when compiling in the same way.

[Rachel Mant]

Ah, that's our bad. PR #1434 wasn't really careful enough to define the tpwr monitoring code only for native or make sure we'd done it in a way that's compatible with all MCUs that make use of that SysTick handler.

Looks like the F4's have different ADC routines and interfaces.. We've got two options here - one is to change the #ifdef guard to check not only that PLATFORM_HAS_POWER_SWITCH is defined but that the platform is at least targeting a STM32F1; the other is to refactor the ADC aspect of this into the platform code properly and then implement it for the F4 platforms too.

[lenvm]

Hi @dragonmux, thanks for your quick reply!

Of course, the ideal solution would be to refactor the code properly and to implement the feature for the F4 platforms as well. I would however already be happy if the #ifdef was changed as a remedial action, until there is time to do the refactoring.

[Rachel Mant]

Before we go and open a PR with an attempted fix then.. could you change the offending #ifdef to: #if defined(PLATFORM_HAS_POWER_SWITCH) && define(STM32F1) and confirm for us that it does in fact fix the issue?

[lenvm]

Hi @dragonmux, I have implemented the changes necessary to fix the immediate issue. This does indeed fix the issue. I have committed the changes and created pull request #1648.