On Wed, Mar 7, 2018 at 5:14 PM, Michael Clark <m...@sifive.com> wrote:
> > > On Wed, Mar 7, 2018 at 12:27 PM, Philippe Mathieu-Daudé <f4...@amsat.org> > wrote: > >> >> Ok until here. >> >> > diff --git a/include/hw/riscv/sifive_e.h b/include/hw/riscv/sifive_e.h >> > index 0aebc57..818fbdc 100644 >> > --- a/include/hw/riscv/sifive_e.h >> > +++ b/include/hw/riscv/sifive_e.h >> > @@ -19,16 +19,7 @@ >> > #ifndef HW_SIFIVE_E_H >> > #define HW_SIFIVE_E_H >> > >> > -#define TYPE_SIFIVE_E "riscv.sifive_e" >> > - >> > -#define SIFIVE_E(obj) \ >> > - OBJECT_CHECK(SiFiveEState, (obj), TYPE_SIFIVE_E) >> > - >> > typedef struct SiFiveEState { >> > - /*< private >*/ >> > - SysBusDevice parent_obj; >> >> I'd keep however a 'Object parent_obj' here, to stay QOM; but your patch >> is valid. >> > > Okay I'll keep parent_obj when I respin. > BTW is the string constant in DEFINE_MACHINE a QOM type? i.e. should I keep the type conversion and type name macros and use the type name macro in DEFINE_MACHINE?