For example:
#define IO_Pin GPIO_PIN_5
#define IO_Port GPIOA
In stm32fxxx_hal_gpio.h:
#define GPIO_PIN_5 ((uint16_t)0x0020) /* Pin 5 selected */
Why do that? Setting an absolute number instead defining the pin number itself (5), which could be later converted using bit shifting (1<<5 = 0x20).
This makes much harder to make automated code.
So, if I quickly wanted to change IO_Pin to input or output, I can't simply make a quick mask operation into MODER register. The only way is to hardcode the number.
// Change IO_pin to OUTPUT mode.
IO_Port->MODER = (IO_Port->MODER & ~(uint32_t)(MODE_ANALOG<<10)) | (uint32_t)(MODE_OUTPUT<<10);
// Change IO_pin to INPUT mode.
IO_Port->MODER = (IO_Port->MODER & ~(uint32_t)(MODE_ANALOG<<10)) | (uint32_t)(MODE_INPUT<<10);
// Change IO_pin to AF mode.
IO_Port->MODER = (IO_Port->MODER & ~(uint32_t)(MODE_ANALOG<<10)) | (uint32_t)(MODE_AF<<10);
This makes an issue for allowing portable code that could be reconfigured by only changing the GPIO label in Cube MX.
Or there's something I'm missing out?
The ideal would be this, allowing changing the pin label while not needing to modify the code:
#define GPIO_PIN_5 5 /* Pin 5 selected */
// Change IO_pin to OUTPUT mode.
IO_Port->MODER = (IO_Port->MODER & ~(uint32_t)(MODE_ANALOG<<(IO_Pin*2))) | (uint32_t)(MODE_OUTPUT<<(IO_Pin*2));
// Change IO_pin to INPUT mode.
IO_Port->MODER = (IO_Port->MODER & ~(uint32_t)(MODE_ANALOG<<(IO_Pin*2))) | (uint32_t)(MODE_INPUT<<(IO_Pin*2));
// Change IO_pin to AF mode.
IO_Port->MODER = (IO_Port->MODER & ~(uint32_t)(MODE_ANALOG<<(IO_Pin*2))) | (uint32_t)(MODE_AF<<(IO_Pin*2));
