Proposal to send GSM sysfs patches on stable

NeilBrown neilb at suse.de
Tue Mar 3 05:15:36 CET 2009


On Tue, March 3, 2009 4:58 am, Paul Fertser wrote:
> Andy Green <andy at openmoko.com> writes:
>> | Andy Green wrote (ao):
>> |> Otherwise, they will simply propose people keep using U-Boot and not
>> Qi
>> |> as their "fix". To the extent we pull some extra current until GSM is
>> |> turned on, Qi is then compatible with old and new kernels. So it's
>> the
>> |> best path right now AFAICT.
>> |
>> | Would it be too inconvenient to have a 'correct' Qi and a
>> | 'backwards compatible' Qi?
>>
>> It's not inconvenient if it can choose what to do at runtime, based on a
>> sign from the U-Boot header that the kernel it's going to run can cope
>> with the right thing.
>
> Not compatible with zImage btw.
>
> Andy, i'm sad to see how you want to spend time to introduce ugliness
> that no distro maintainer seem to want.

Distro maintainers aren't your only customers.
Some people really like to build their own setup from bits and pieces.
a kernel here, a boot loader there.  Take a root fs and mess with it for
a while until it looks like nothing you've seen before.
Different hobbyists want to work at different levels...

So I think that assuming no known distro maintainers want this doesn't mean
that no-one will benefit.
As you really don't know who all your customers are, having a strong
"no regressions" policy is really important.

I think Andy's current direction is exactly the right one.  He'll probably
never know how many people were saved hours of frustration by his efforts.
And maybe it won't be very many.  But the policy is still a good one,
because when you do cause a regression like this, you can be sure you'll
hear about it, and responding to all those emails is just dead time.

NeilBrown
Qi + Andy-tracking + Debian - FSO + assorted Python Hacks.




More information about the openmoko-kernel mailing list