Re: [PATCH 1/2] ARM: Differentiate SheevaPlugs and DockStars on the basis of the memory size.

[Russell King - ARM Linux <linux@arm.linux.org.uk>]

On Fri, Apr 08, 2011 at 01:04:01AM +0200, Alexander Holler wrote:
> Am 08.04.2011 00:08, schrieb Russell King - ARM Linux:
>
>>> Requiring a machine ID and the needed stuff to handle that for a board
>>> which just is using two GPIOs different than another board is why the
>>> ARM tree exploded.
>>
>> You can not be any more wrong than that.
>>
>> The reason the ARM tree exploded is because of the compartmentalized
>> sub-community structure, where the vast majority of (eg) OMAP development
>> is done independently of the (eg) Samsung development.
>>
>> Consequently, there's no attempt to consolidate code between the SoCs,
>> even for basic stuff like 32-bit up-counting timers.  We've ended up with
>> _nine_ implementations of clocksources all doing the same thing in that
>> respect.
>>
>> That's got precisely zilch to do with machine IDs, and your attempt to
>> blame the bloat on machine IDs just shows how misinformed you are.
>
> I had a look at what's going on in the OMAP linux world for more than a  
> year now and I think I've seen a lot of the stuff you are referring to.
>
> And I think one of the reasons that the mess happened is the same I've  
> got trapped in. Why should anyone try to submit patches if he must fear  
> to get caught in some senseless endless discussion about one line.

Because matching against the memory size is _technically_ the wrong thing
to do, as you've already been told.