Re: Mesa 25.2 patch

On 10/20/25 14:08, Damien Stewart wrote:
> On 20/10/25 9:33 pm, Michel Dänzer wrote:
>> The patch is incorrect regardless of endianness, it could break things even on big-endian architectures. It can only help for anything by accident.
> 
> So the patch is simple and removes BGRA ordered pixels from a list? But keeps ABGR ordered pixels. Does the order of items in table list matter?

The order shouldn't matter[0], the problem is that the correct mappings for PIPE_FORMAT_X/A8R8G8B8_UNORM are replaced by incorrect ones.

[0] Not for correctness, it can matter for which formats end up used in a given scenario though.

> I never got that ABGR order, it's almost like a confused PDP 11 pixel order.

It's a matter of convention, it's not possible to determine what the order actually is without knowing the applicable convention(s).

The important point to keep in mind here is that both PIPE_FORMAT_?8?8?8?8_UNORM & DRM_FORMAT_????8888 are defined independently from endianness, so there's an unambiguously correct mapping between them regardless of endianness.