On Sun, Apr 26, 2020 at 09:30:57AM +1000, Michael Ellerman wrote: > Mark Brown writes: > > I am doing that but that still doesn't mean that the architectures > > shouldn't be updated - to me this is like the architectures that don't > > implement standard APIs, we should fix the issues they bring up but it'd > > be a lot less noisy to sidestep the issue. > I don't think it's really like architectures that don't implement > standard APIs. > It's more like architectures not building code they don't need, AFAICS > none of the drivers under there can ever be used on powerpc. That's generally the reason people give for not doing the standard APIs - there were a few architectures that didn't have DMA hardware so didn't even have stubs for the DMA API for example, until those were added you used to end up with build testing tripping over them constantly. It's much rarer to have an incompatible implementation of the same thing. > Similarly we don't build drivers/acpi. ACPI has stubs so doesn't cause issues here. > But if there's a good reason that we should be building it then I'm > happy to take a patch adding it. The build coverage seems useful.