[PATCH 02/10] jscript: Store the necessary function and variable info in the TypeInfo.
Jacek Caban
jacek at codeweavers.com
Tue Dec 10 11:49:10 CST 2019
Hi Gabriel,
On 12/9/19 6:12 PM, Gabriel Ivăncescu wrote:
> static DWORD get_flags(jsdisp_t *This, dispex_prop_t *prop)
> {
> if(prop->type == PROP_PROTREF) {
> @@ -590,9 +604,21 @@ static HRESULT fill_protrefs(jsdisp_t *This)
> return S_OK;
> }
>
> +struct typeinfo_func {
> + dispex_prop_t *prop;
> + jsdisp_t *disp;
> +};
Storing the whole function object reference does not seem right here.
Maybe we should store function_code_t instead? We could have a single
getter that could replace both is_source_function() and
get_source_function_params() from your patches. Such getter should
probably use function_vtbl_t for that. Storing bytecode_t pointer inside
function_code_t could also be handy for reference tracking.
>
> + for (prop = This->props, end = prop + This->prop_cnt; prop != end; prop++)
> + {
> + if (!prop->name || prop->type != PROP_JSVAL || !(prop->flags & PROPF_ENUMERABLE))
> + continue;
> +
> + /* If two identifiers differ only by case, the TypeInfo fails */
> + pos = This->props[get_props_idx(This, prop->hash)].bucket_head;
> + while (pos)
> + {
> + cur = This->props + pos;
> +
> + if (prop->hash == cur->hash && prop != cur &&
> + cur->type == PROP_JSVAL && (cur->flags & PROPF_ENUMERABLE) &&
> + !wcsicmp(prop->name, cur->name))
> + {
> + return TYPE_E_AMBIGUOUSNAME;
> + }
> + pos = cur->bucket_next;
> + }
It would be good to limit dispex_prop_t usage inside ITypeInfo in
general. IDispatchEx support in jscript deserves pretty deep changes and
extending usage of its internal structure may not be helpful. I'm fine
with trying to use them here for now, but I'd rather avoid depending on
its internals and that's the kind of internal thing I'd rather not have
here. We could call GetDispID(fdexNameCaseInsensitive) and compare
returned id to current one, if we had support for that. However, is it
really important in practice? I wouldn't mind leaving it as FIXME for now.
Thanks,
Jacek
More information about the wine-devel
mailing list