Re: [PATCH][RESEND] ARM: mach-shmobile: sh7372 generic suspend/resume support

[Magnus Damm <magnus.damm@gmail.com>]

On Fri, Aug 26, 2011 at 6:23 PM, Russell King - ARM Linux
<linux@arm.linux.org.uk> wrote:
> On Fri, Aug 26, 2011 at 05:44:04PM +0900, Magnus Damm wrote:
>> Good question. This was posted before the 3.1-rc merge window, but
>> didn't make it into mainline for some unknown
>> reason. I'm all for merging things sooner than later if possible, but
>> I understand if 3.1-rc is out of the question at this point.
>>
>> If you are offering to queue this up for the 3.2 merge window then I
>> am instead suggesting that Rafael picks this up in his PM tree because
>> there are other PM related patches that build on top of this one.
>
> If Rafael wants to pick it up, that's fine.

Let's see what he says about that.

>> While at it, for the 3.1 release, would it be possible to update the
>> mach-types list in mainline? We have patches for the Kota2 board that
>> depend on the mach-type update.
>
> I dropped my mach-types update (which was required to fix a build error -
> because a platform was introduced into mainline without its entry in
> mach-types) as the update was causing a number of problems with other
> platforms and drivers in mainline.

Ouch.

> I'm not sure what the long term future of mach-types is, as it's become
> quite a burden to manually update it each time due to the amount of crap
> now contained in there, and it seems automated filtering results in
> breakage.

Honestly, I think your mach-types list and the patch tracker have
served their purpose for a long time. They make totally sense to
handle the kind of work load and of diverse hardware platform support
that you have been dealing with over the years. These days I suspect
there are other solutions available that may make it easier for you.

> I think the whole interface to it needs to be revised so that it is
> purely read-only, and editing is done via human interaction adding a
> level of manual approval to the system.  It's the only way I can think
> of stopping the 'ARM', 'IMX', etc like entries from being added, along
> with ensuring that the machine_is_xxx() and MACH_TYPE_xxx() names always
> match.

I guess you prefer to keep the information separate from the Linux kernel?

If you don't mind, then wouldn't it be possible to move over the
information to some more verbose form of the mach-types file and store
it in the kernel tree and use regular git to manage it? That's what I
would do.

Another option is to extend your web interface and separate addition
of new entries from modification of already existing ones. Perhaps new
entries can be merged at any time in -rc but modification needs to
happen early in the cycle.

Cheers,

/ magnus