test: reference image

[CarlFK]

test: supply a known image on hdmi input0/1, compare that to a known image.
image would be one of:

1. image file supplied at build time
2. image uploaded over usb/eth
3. frame grabbed from input0/1
4. internally generated pattern.

Any one of these would be good for testing the critical bit: consuming hdmi.
having all of them would also test other bits of the system.

[mithro]

I would start with (4). If you turn off the uptime part of the current test pattern, that image should be a known starting point and testing that you get it from the USB device would be useful.

Once you know that you can reliably capture an image via USB, it should then be pretty trivial to use something like an RPi to act as a HDMI source and generate a bunch of test patterns in the "test-patterns" repo.

[CarlFK]

added issue - option: text on/off of pattern #434

Even with uptime changing and all the other text, I would think (hope) the output and input buffers could be compared before the new pattern image steps on the output. I wonder if testing a moving target makes a better test. like if it fails, it means some timing thing fell apart.

I can reliably capture an image via USB (smug face). I am slowly getting to a legitimate test suite.