Re: [PATCH] d3dx9_36: Beginning of constants parsing + tests (based on work from Luis Busquets) (try 3) (resend)

[Alexandre Julliard]

Christian Costa <titan.costa(a)wanadoo.fr> writes:
> @@ -650,6 +650,29 @@ HRESULT WINAPI D3DXCompileShader(LPCSTR pSrcData,
>      return D3DERR_INVALIDCALL;
>  }
>
> +typedef struct ctab_constant ctab_constant;
> +
> +struct ctab_constant {
> +    D3DXCONSTANT_DESC desc;
> +    ctab_constant* members;
> +};
> +
> +inline ctab_constant* get_constant(D3DXHANDLE ptr)
> +{
> +    if (!ptr) return NULL;
> +    return (ctab_constant*)~((UINT_PTR)ptr);
> +}
> +
> +inline D3DXHANDLE get_handle(ctab_constant* ptr)
> +{
> +    return (D3DXHANDLE)~((UINT_PTR)ptr);
> +}
> +
> +inline BOOL isstring(D3DXHANDLE ptr)
> +{
> +    return !(((UINT_PTR)ptr) & (1 << (sizeof(UINT_PTR)*8-1)));
> +}

This is broken, you can't make assumptions about the high bit of pointer
values. The handle management needs more thought.

-- 
Alexandre Julliard
julliard(a)winehq.org

[PATCH] d3dx9_36: Beginning of constants parsing + tests (based on work from Luis Busquets) (try 3) (resend)

[Christian Costa]

> Message du 09/04/10 11:33
> De : "Alexandre Julliard" 
> A : "Christian Costa" 
> Copie à : wine-devel(a)winehq.org
> Objet : Re: [PATCH] d3dx9_36: Beginning of constants parsing + tests (based on work from Luis Busquets) (try 3) (resend)
>
> Christian Costa writes:
> > @@ -650,6 +650,29 @@ HRESULT WINAPI D3DXCompileShader(LPCSTR pSrcData,
> > return D3DERR_INVALIDCALL;
> > }
> >
> > +typedef struct ctab_constant ctab_constant;
> > +
> > +struct ctab_constant {
> > + D3DXCONSTANT_DESC desc;
> > + ctab_constant* members;
> > +};
> > +
> > +inline ctab_constant* get_constant(D3DXHANDLE ptr)
> > +{
> > + if (!ptr) return NULL;
> > + return (ctab_constant*)~((UINT_PTR)ptr);
> > +}
> > +
> > +inline D3DXHANDLE get_handle(ctab_constant* ptr)
> > +{
> > + return (D3DXHANDLE)~((UINT_PTR)ptr);
> > +}
> > +
> > +inline BOOL isstring(D3DXHANDLE ptr)
> > +{
> > + return !(((UINT_PTR)ptr) & (1 << (sizeof(UINT_PTR)*8-1)));
> > +}
>
> This is broken, you can't make assumptions about the high bit of pointer
> values. The handle management needs more thought.
>
> -- 
> Alexandre Julliard
> julliard(a)winehq.org

Does a 16-bit index would be suitable ?