Re: [PATCH v5 08/15] acpi: move aml builder code for floppy device

[Thomas Huth <thuth@redhat.com>]

On 13/05/2020 22.43, John Snow wrote:
> 
> 
> On 5/7/20 10:05 AM, Philippe Mathieu-Daudé wrote:
>> +Hervé
>>
>> On 5/7/20 3:16 PM, Gerd Hoffmann wrote:
>>> Signed-off-by: Gerd Hoffmann <kraxel@redhat.com>
>>> Reviewed-by: Igor Mammedov <imammedo@redhat.com>
>>> ---
>>>   hw/block/fdc.c       | 83 ++++++++++++++++++++++++++++++++++++++++++++
> 
> oh no...
> 
>>>   hw/i386/acpi-build.c | 83 --------------------------------------------
>>>   stubs/cmos.c         |  7 ++++
>>>   stubs/Makefile.objs  |  1 +
>>>   4 files changed, 91 insertions(+), 83 deletions(-)
>>>   create mode 100644 stubs/cmos.c
>>>
>>> diff --git a/hw/block/fdc.c b/hw/block/fdc.c
>>> index 9628cc171ef8..40faa088b5f7 100644
>>> --- a/hw/block/fdc.c
>>> +++ b/hw/block/fdc.c
>>> @@ -32,6 +32,8 @@
>>>   #include "qapi/error.h"
>>>   #include "qemu/error-report.h"
>>>   #include "qemu/timer.h"
>>> +#include "hw/i386/pc.h"
>>
>> I'd rather not see this target-specific header used in the generic
>> device model... The culprit seems cmos_get_fd_drive_type(). Is the value
>> really PC specific?
>>
>> Hervé, do you know if such info & RTC nvram layout is used in other arch?
>>
> 
> do we need pc-only floppy code somewhere to keep the makefile pure?

Yes, please put this into a separate file that only gets build if the
corresponding CONFIG switches are set. That way you should also get
along without that new stubs/cmos.c file.

 Thomas