On Thu, Aug 20, 2020 at 06:57:33PM +0100, Daniel P. Berrangé wrote:
> On Thu, Aug 20, 2020 at 01:45:06PM -0400, Eduardo Habkost wrote:
> > (CCing Eric Blake, who reviewed the original patch.  Sorry for
> > not CCing you previously, Eric)
> > 
> > On Wed, Aug 19, 2020 at 08:12:22PM -0400, Eduardo Habkost wrote:
> > [...]
> > > +/**
> > > + * OBJECT_DECLARE_TYPE:
> > > + * @ModuleObjName: the object name with initial capitalization
> > > + * @module_obj_name: the object name in lowercase with underscore 
> > > separators
> > > + * @MODULE_OBJ_NAME: the object name in uppercase with underscore 
> > > separators
> > 
> > We need to decide what to do with TYPE_SWIM:
> > 
> >   #define TYPE_SWIM "swim"
> >   typedef struct SWIM SWIM;
> >   #define SWIM(obj) OBJECT_CHECK(SWIM, (obj), TYPE_SWIM)
> > 
> > Both the typedef and type checking macros are called "SWIM".
> > This makes usage of OBJECT_DECLARE_TYPE impossible (because the
> > type checking function can't have the same as the typedef).  What
> > should be the recommended style here?  Rename the struct to
> > "Swim"?
> 
> This would be the gobject style approach to the problem - structs
> always have initial-caps-only even if they are abbreviations.
> 
> If that's unappealing we could add a suffix  "struct SWIMInst" 

I was considering "SWIMState", because "State" is a common
instance type suffix in QEMU device code.  But I will let the
maintainer (Laurent) decide.

> 
> > We have similar issues with RXCPU and ARMSSE, but these type
> > checking macros can be easily renamed to RX_CPU and ARM_SSE. so
> > they won't be an issue.
> 
> Or  RSCPUInst  and ARMSSEInst  ?

In these cases, renaming the type checking macro seems more
logical, because the type name constants are TYPE_RX_CPU and
TYPE_ARM_SSE.

-- 
Eduardo


Reply via email to