Enabling RMT/semaphores in complex code can cause analogread to return values 2x too big

[marcmerlin]

As first reported here MartyMacGyver/ESP32-Digital-RGB-LED-Drivers#7
using the RMT neopixel driver causes analogread to return 4096 for 50%-100% and proper values from 0 to 50%.
Sadly if you are sampling data that needs to go in the 50 to 100% range (joystick), it's impossible to get usable data.
If i comment out just these 3 lines of code, then analogread works properly again:
//RMT.conf_ch[RMTCHANNEL].conf1.mem_rd_rst = 1;
//RMT.conf_ch[RMTCHANNEL].conf1.tx_start = 1;
//xSemaphoreTake(ws2812_sem, portMAX_DELAY);

[projectgus]

Are you seeing this running with Arduino or on IDF?

[marcmerlin]

arduino, in my IoTuz code.
Sadly if I take the code way down to just turning the LEDs on with the RMT code, then analogread works.
Yet, if I take all the IoTuz code, and comment out the 3 RMT lines above, then all the analogreads that weren't working right, start working right again.

[projectgus]

Please open the issue here: https://github.com/espressif/arduino-esp32/issues/ as the ADC software implementations (along with various other subsystem parts) are different.

If you can reproduce the issue under IDF by itself then feel free to reopen here.

[marcmerlin]

Ooops, sorry I was under the wrong project, moved accordingly, thanks for the note.
espressif/arduino-esp32#290