On Mon, 5 Jan 2026 15:54:32 +0800
Zhao Liu <[email protected]> wrote:

> Thanks for looking at this!
> 
> > > Therefore, rename ObjectPropertyFlags to ObjectPropertyAccessorFlags,
> > > and then the original name ObjectPropertyFlags can be used for other
> > > cases.
> > 
> > I wonder if we can  just make existing ObjectPropertyFlags per object as 
> > you are planing
> > and still continue using it for accessor flags.
> >
> > That basically will lets you use new flags everywhere without rewriting 
> > call sites
> > everywhere.
> 
> I'm not sure about this. Currently, these read/write flags are actually
> specific to pointer properties (as showed by the changes in this patch,
> which all involve object_property_add_*_ptr() / 
> object_class_property_add_*_ptr()).
> 
> Other property types doesn't yet support flag parameters, so additional
> interface modifications are still needed.
> 
> And for now other property types either need to explicitly specify get/set
> accessors (e.g., object_property_add_bool()) or directly use the default
> get/set methods (e.g., object_property_add_link()).
> 
> If we extend read/write flags to other property types, such as adding
> "flags" argument to object_property_add_bool(), we must ensure the
> OBJ_PROP_FLAG_READ flag align with "get" argument and OBJ_PROP_FLAG_WRITE
> flag align with "set" parameters.

Ain't thouse accessors callbacks?
/I mean to you still can check flags inside of generic object property code
without touching setter/getter./

>This would introduces additional complexity.

it still might be woth considering to compare this series with alternative 
approach.

> 
> Thanks,
> Zhao
> 


Reply via email to