CBcmFrameBuffer can't set 32 bit color on RPI5

[andreas-jonsson]

Hi!

Using the CBcmFrameBuffer object and setting the color depth (and resolution?) to 32 bit on RPI5. It seems to quietly, without errors fall back to 16 bit.

Initialize do not return any error and the object reports back that bit depth is 32 bit when it clearly still is 16 bit.

I'm a bit new to Circle so I don't know if I'm missing something here? But the code works just as expected on RPI3&4 so I assume something is different with the RPI5?

I found this post. Not sure if it is relevant to this: raspberrypi/firmware#1904

[rsta2]

The frame buffer support in the firmware of the RPi 5 is unfortunately not as comfortable as for earlier models. This is a problem, which is caused by a non-functioning property mailbox function in the RPi 5, which Circle uses to set the display depth. Unfortunately I don't know an other programmatic way to realize this. I guess it does not exist.

Could you please try the option framebuffer_width=32 in the file config.txt? I think I read somewhere, that it may work to set the depth. I have an other RPi model currently on my desk and cannot change this quickly, so I cannot test this by myself at the moment.

[andreas-jonsson]

Thanks for the quick response. I'll have a look as soon as I have some spare time.

[rsta2]

You are welcome. Sorry, I mean framebuffer_depth=32 of course, not the width. ;)

https://www.raspberrypi.com/documentation/computers/legacy_config_txt.html#framebuffer_depth

[rsta2]

Unfortunately framebuffer_depth=32 does not work on the RPi 5. The display image is not clean. So there is currently no way to use this depth with Circle on the RPi 5. :(

[andreas-jonsson]

I'm sorry. I have been pulled in to another project and I don't have a easy to test environment setup atm. But I will follow up on this.

May I suggest that setting bit depth to anything else then 16bit on RPI5 will result in an error while calling Initialize or at the very least the CBcmFrameBuffer object don't falsely report 32bit?

[rsta2]

No problem.

Yes, I will implement this.

[rsta2]

Added a test for depth 16 on the develop branch.