Debug and re-enable I2S tests on ESP32

[bugadani]

Disabled in #2439 due to noise.

[bugadani]

A (probably separate) ESP32 I2S issue can be reproduced by liebman/esp-hub75@e5f8392

Placing an info between the following lines makes the example work:

https://github.com/liebman/esp-hub75/blob/e5f83929d527bbc716b5dffe3c9b6c508ab84f64/examples/i2s_parallel.rs#L260-L264

This issue feels very similar to other i2s/spi issues (like #2909) where we have to wait for some amount of time, otherwise an async operation hangs.

[liebman]

A (probably separate) ESP32 I2S issue can be reproduced by liebman/esp-hub75@e5f8392

Placing an info between the following lines makes the example work:

https://github.com/liebman/esp-hub75/blob/e5f83929d527bbc716b5dffe3c9b6c508ab84f64/examples/i2s_parallel.rs#L260-L264

This issue feels very similar to other i2s/spi issues (like #2909) where we have to wait for some amount of time, otherwise an async operation hangs.

This also happens on the esp32s3 using lcd_cam - example here.

On the esp32c6, however it works as expected - example here.

All the above examples are very similar.

Seems like something with the DMA interrupts on xtensa not playing nice with embassy executor. A logic analyzer shows that the fist transfer is sent - it seems like the task is never awakened from the interrupt.

[Dominaezzz]

The I8080 driver doesn't use the DMA interrupts fwiw, it has it's own "I'm done" interrupt.

[liebman]

More on the i8080 - it never sends anything - not even one clock.

[bugadani]

Thank you both, a number of problems have been discovered around DMA and the LCD_CAM peripheral, but neither of those are actually related to this particular issue.