Can PYSMAP_COMPAT go?

[David Woodhouse]

On Tue, 2010-04-20 at 09:35 +0200, Lennert Buytenhek wrote:
> COMPAT needs to die.  There are at least two other ways of specifying
> the same info (device tree or manual instantiation of a platform device
> from your platform code), and we really shouldn't allow specifying this
> in the .config as that will just give people other crazy ideas. 

If we take it away, they'll get crazier ideas. I've seen a fully-fledged
driver for a new board, including a bizarrely hacked-up copy of all the
CFI code and a special char device interface of its own, which could
have been supported by a simple PHYSMAP setup.

And this was i386, which unfortunately doesn't use the device-tree yet,
although on these embedded SoC devices it really _should_ be doing so.
And it doesn't have 'platform code' either.

If you explicitly configure your kernel to tell it that there is a flash
device at a certain address, then it's reasonable for the kernel to look
there for it. Aunt Tillie can go screw herself.

-- 
dwmw2