[PATCH] pci: Add identifier for Cirus Logic GD-7556

[Tormod Volden]

On Sun, Oct 30, 2011 at 12:12 AM, Jeremy Huddleston <jeremyhu at apple.com> wrote:
>
> On Oct 29, 2011, at 14:01, Tormod Volden wrote:
>
>> On Mon, Oct 17, 2011 at 10:53 AM, Jeremy Huddleston <jeremyhu at apple.com> wrote:
>>> So, what's the point of having all these in common?  Why don't they just live in the drivers?  It seems silly to have these defined in the server's SDK rather than the driver that will consume them, since it means the driver will need to do
>>>
>>> #ifndef PCI_CHIP_GD7556
>>> #define PCI_CHIP_GD7556 0x004C
>>> #endif
>>>
>>> if it wants to work on older servers without the change ...
>>
>> Yes, should we move this to the drivers and make them stop including
>> xf86PciInfo.h? Are the PCI_CHIP_ macros used by anything else than the
>> drivers, or can they be taken out as soon as drivers are updated? The
>> server itself is not using them, according to git grep. I will follow
>> up with a radeon patch, it currently has a mix of both, and see how
>> that is received.
>>
>> The #ifndef dance it not necessary on gcc at least, as long as a
>> redefine uses the same value.
>
> This is complicated by the fact that we probably want older drivers to continue to build against the newer server ... so maybe we can put a #warning in xf86PciInfo.h that it will be removed in xorg-server-1.13 or later and start migrating them to the drivers instead.  What do you think about that?
>

I see it as most important that newer drivers can be built against
older servers. Usually you will have server changes that force driver
updates. But sure, we can leave the macros in the server for a while
after a driver has copied them and the following driver release. You
mean a warning that appears while compiling drivers? Sounds good.
Regarding my comment in the radeon patch, what do you think about the
VENDOR macros?

Cheers,
Tormod