Examples broken under esp-idf V4.4

[jetpax]

Thanks for the great work Mike.

I found that the significant changes in the esp-idf i2s driver since V4.3, mainly to support mclk it seems, cause some of these examples to break (eg record_mic_to_sdcard_blocking.py), with these errors

E (19344) I2S: i2s_calculate_common_clock(1215): sample rate is too large
E (19344) I2S: i2s_calculate_clock(1264): Common clock calculate failed
E (19354) I2S: i2s_check_set_mclk(253): ESP32 only support to set GPIO0/GPIO1/GPIO3 as mclk signal, error GPIO number:1061222028
E (19364) I2S: i2s_set_pin(314): mclk config failed

I worked around this in my application, which does not use mclk by adding the following in machine_i2s.c

i2s_config.mclk_multiple = I2S_MCLK_MULTIPLE_256;

to i2s_config

and

pin_config.mck_io_num = I2S_PIN_NO_CHANGE;

to pin_config

Obviously this needs a more thorough fix, especially if mclk is to be suppported...

[miketeachman]

Thanks for reporting this. I'll install esp-idf v4.4-beta1 and check this out.

[miketeachman]

I did some investigation into the ESP-IDF v4.4 Beta release. Adding the code you suggested fixes the breaking changes made in the ESP-IDF. Thank you!

I also found two more problems:

1. uasyncio support in I2S is broken from a bug introduced into the I2S implementation. I submitted an Issue:
   espressif/esp-idf#8121

2. Correct I2S operation has a dependency on the code that runs before the I2S constructor. For example, the tone example does not run, but adding a SPI constructor ahead of the I2S constructor gets the example working -- very strange. More debugging needed to identify the root cause of this peculiar behaviour.

ESP-IDF v4.4 is still in Beta, so I hope that Espressif will fix at least problem 1 above before the final release.

[miketeachman]

I submitted a PR for this issue:
micropython/micropython#8110

[miketeachman]

The PR has been merged.