[Mesa-dev] [PATCH] r300g: only allow byteswapped formats on big endian

[Alex Deucher]

On Tue, Feb 14, 2017 at 10:27 PM, Michel Dänzer <michel at daenzer.net> wrote:
> On 14/02/17 08:25 PM, Marek Olšák wrote:
>> I've changed my mind. The patch can be merged if nobody disagrees.
>
> It would be nice to better understand what exactly the problem is. It
> seems unlikely that it's actually an endianness specific problem in the
> driver code, more likely an endianness specific problem in state tracker
> code, or maybe a non-endianness-specific problem in the driver code
> (which may or may not be possible to hit on big endian hosts as well).
>
> In other words, I suspect this isn't a fix for the actual problem, but
> just a workaround. That said, if you guys are happy with that, I'm not
> holding it up.

I tend to agree, but OTOH, r300 is pretty old and BE systems are
pretty few and far between these days.  Neither of which are conducive
to spending much time on them.  Someone with an interest in BE systems
really needs to step up and fix up endian handling in mesa for real.

Alex