[meego-packaging] Proposal for change in PulseAudio packaging of configuration

[Auke Kok]

On 12/02/10 08:45, Carsten Munk wrote:
> 2010/12/2 Arjan van de Ven<arjan at linux.intel.com>:
>>>
>>> Sounds like a good direction. Maybe we should provide a simple package
>>> wih '/usr/bin/meego-machinename' which outputs a proper machine name
>>> in an arch-indep manner? (I guess we shouldn't encourage people to
>>> parse /proc/cpuinfo or rely on /sys/devices/virtual/dmi/id/board_name
>>> existence.. #ifdef __arm__ or x86 is so ugly :)
>>>
>>> That, or a machine name being put in /etc/meego-machinename at startup
>>> or something to avoid launching a process each time.
>>
>> I like the /etc/meego-machinename thing  or even just /etc/machinename
>>
>> we can do the x86 side no big deal in a short time...
>> (heck it could just be a symlink to the sysfs file there)
>
> I guess a good place would be rc.sysinit for this then, CC'ing Auke..
>
> To be a bit future proof, I'd like to propose following semantics then:
>
> * if exists /etc/machinename-override then symlink to that
> * else run machinename-detect
>     * x86 initial impl: symlink to
> /sys/devices/virtual/dmi/id/board_name, if containing 'INVALID' (my
> laptop does that), restore to platform name
>     * arm initial impl: parse /proc/cpuinfo for Hardware: line and
> write machine name to /etc/machinename, if invalid, restore to
> platform name (which we get from where?)
>
> -override is to deal with things like a BeagleBoard which may have
> differing configurations depending on what hardware is attached or for
> general development purposes.
>
> On ARM side it's complex as a board name also can contain spaces, so
> we might need some more parsing -- hence machinename-detect, but we'll
> contribute that part after talking to some vendors.. I think there is
> a unique board ID for each ARM board, but how to get it from
> userspace..

We'd need something extremely early - so currently rc.sysinig would be 
the place - right after all the mount stuff.

Auke