Re: [Qemu-devel] [Qemu-ppc] [PATCH] PPC: Move ppc specific hw emulation to hw/ppc

[Alexander Graf]

On 25.01.2013, at 20:41, Scott Wood wrote:

> On 01/25/2013 01:26:48 PM, Alexander Graf wrote:
>> On 25.01.2013, at 20:06, Blue Swirl wrote:
>> > On Fri, Jan 25, 2013 at 1:12 AM, Alexander Graf <address@hidden> wrote:
>> >> This patch moves all PowerPC specific hardware emulation except for
>> >> PREP to hw/ppc.
>> >>
>> >> Signed-off-by: Alexander Graf <address@hidden>
>> >> ---
>> >> MAINTAINERS                  |   20 ++++++++++----------
>> >> hw/Makefile.objs             |   16 +---------------
>> >> hw/ppc/Makefile.objs         |   38 ++++++++++++++++++++++++++++++--------
>> >> hw/{ => ppc}/adb.c           |    2 +-
>> >> hw/{ => ppc}/adb.h           |    2 +-
>> >> hw/{ => ppc}/cuda.c          |    4 ++--
>> >> hw/{ => ppc}/dec_pci.c       |   10 +++++-----
>> >> hw/{ => ppc}/dec_pci.h       |    0
>> >
>> > Please don't, Sparc64 should also use this device.
>> I think it makes sense to move it back as soon as sparc64 actually starts 
>> using it.
> 
> ...and then you break the history chain again (not totally, but it requires 
> more effort to follow with things like git blame), and make it less likely 
> that someone looking to reuse this code for another architecture finds out 
> that it exists.
> 
> IMHO files that don't actually depend on arch details should not go in an 
> arch directory, just as drivers don't generally go into arch directories on 
> Linux even if they're only found on a single platform.  They can still be 
> hierarchalized based on function (which is also helpful for finding files 
> that use similar interfaces).  Even files that do depend on arch details may 
> make more sense grouped by something else, if the arch interaction is minor 
> compared to the other interfaces.

Fair enough. I've removed this patch from my queues again - both ppc-next as 
well as ppc-for-upstream. That way the pull request is still valid and we can 
discuss this patch further.


Alex