Fixes UART detach. Fixes #3878

[lbernstone]

From what I can tell, there used to be a hidden mechanism to detach the uart pins in gpio_matrix_out. This no longer works, so we need to keep track of what pins we assign.

[aster94]

any news on this?

[drojf]

Are you sure that pinMatrixInDetach() takes the same first argument as pinMatrixOutDetach()? If you look at the definition/SPI code for the same thing, you'll find it takes a different argument to the ...OutDetach() function:

arduino-esp32/cores/esp32/esp32-hal-matrix.c

Lines 28 to 41 in 37a7fb3

	void IRAM_ATTR pinMatrixOutDetach(uint8_t pin, bool invertOut, bool invertEnable)
	{
	gpio_matrix_out(pin, MATRIX_DETACH_OUT_SIG, invertOut, invertEnable);
	}
	void IRAM_ATTR pinMatrixInAttach(uint8_t pin, uint8_t signal, bool inverted)
	{
	gpio_matrix_in(pin, signal, inverted);
	}
	void IRAM_ATTR pinMatrixInDetach(uint8_t signal, bool high, bool inverted)
	{
	gpio_matrix_in(high?MATRIX_DETACH_IN_LOW_HIGH:MATRIX_DETACH_IN_LOW_PIN, signal, inverted);
	}

arduino-esp32/cores/esp32/esp32-hal-spi.c

Lines 147 to 181 in 37a7fb3

	void spiDetachMISO(spi_t * spi, int8_t miso)
	{
	if(!spi) {
	return;
	}
	if(miso < 0) {
	if(spi->num == HSPI) {
	miso = 12;
	} else if(spi->num == VSPI) {
	miso = 19;
	} else {
	miso = 7;
	}
	}
	pinMatrixInDetach(SPI_MISO_IDX(spi->num), false, false);
	pinMode(miso, INPUT);
	}
	void spiDetachMOSI(spi_t * spi, int8_t mosi)
	{
	if(!spi) {
	return;
	}
	if(mosi < 0) {
	if(spi->num == HSPI) {
	mosi = 13;
	} else if(spi->num == VSPI) {
	mosi = 23;
	} else {
	mosi = 8;
	}
	}
	pinMatrixOutDetach(mosi, false, false);
	pinMode(mosi, INPUT);
	}

I guess there could also be an error in the esp32-hal-matrix.c file definition for pinMatrixInDetach().

If they really do take different arguments, you only need to change the pinMatrixOutDetach() call (so as to match the SPI example)

[lbernstone]

I didn't change the header definitions for either of the pinMatrix detach functions, so I'm pretty sure the arguments are correct. This PR is specifically to deal with the fact that the UART_TXD_IDX pins (which are in the 200s, IIRC) no longer work in the IDF functions. The SPI versions of those pins must still work, since nobody is complaining about SPI::end failing.

[dka23]

See also #3205