On 17/09/2015 18:16, Peter Maydell wrote:
> On 17 September 2015 at 17:00, Paolo Bonzini <pbonz...@redhat.com> wrote:
>>
>>
>> On 17/09/2015 16:24, Peter Maydell wrote:
>>> Can we revert this one, please? Checkpatch now warns about constructs
>>> like
>>> typedef struct MyDevice {
>>> DeviceState parent;
>>>
>>> int reg0, reg1, reg2;
>>> } MyDevice;
>>
>> It's interesting that qom/object.h documents this and start like:
>>
>> typedef struct ObjectClass ObjectClass;
>> typedef struct Object Object;
>>
>> typedef struct TypeInfo TypeInfo;
>>
>> typedef struct InterfaceClass InterfaceClass;
>> typedef struct InterfaceInfo InterfaceInfo;
>>
>> I have a patch to flag widely-disrespected rules that we still want to
>> encourage in patches. Would you agree with filing these typedefs under
>> this category?
>
> No, I think that having a separate typedef is worse. The
> only exceptions are (a) when you need it to be separate because
> you need to use the type within the struct itself (or some
> similar dependency loop) (b) when you want to put the typedef
> in include/qemu/typedefs.h.
>
> I really don't see any need to suddenly outlaw something
> that's been accepted as standard good QEMU style for a
> long time.
I think it varies depending on the maintainer. PPC, USB, SCSI, ACPI all
use a separate typedef. I'll prepare a revert.
Paolo
