[PATCH 1/7] widl: Generate helper macros for WinRT implementation.

[Rémi Bernon]

On 2/25/21 2:34 PM, Jacek Caban wrote:
> Hi Rémi,
> 
> On 22.02.2021 10:02, Rémi Bernon wrote:
>> +static void write_widl_using_method_macros(FILE *header, const type_t 
>> *iface, const type_t *top_iface)
>> +{
>> +    const statement_t *stmt;
>> +    const char *name = top_iface->short_name ? top_iface->short_name 
>> : top_iface->name;
>> +
>> +    if (type_iface_get_inherit(iface)) 
>> write_widl_using_method_macros(header, type_iface_get_inherit(iface), 
>> top_iface);
>> +
>> +    STATEMENTS_FOR_EACH_FUNC(stmt, type_iface_get_stmts(iface))
>> +    {
>> +        const var_t *func = stmt->u.var;
>> +        const char *func_name;
>> +
>> +        if (is_override_method(iface, top_iface, func)) continue;
>> +        if (is_callas(func->attrs)) continue;
>> +
>> +        func_name = get_name(func);
>> +        fprintf(header, "#define %s_%s %s_%s\n", name, func_name, 
>> top_iface->c_name, func_name);
>> +    }
>> +}
>> +
>> +static char *format_widl_using_macro(const type_t *type)
>> +{
>> +    char *name = format_namespace(type->namespace, "WIDL_USING_", 
>> "_", "", NULL);
>> +    int i = strlen(name);
>> +    if (i) name[i - 1] = 0; /* remove trailing underscore */
> 
> 
> Not a big deal, but maybe the helper should take care of that instead? 
> For example NULL suffix could case the separator not to be appended.
> 

Sure.

> 
>> diff --git a/tools/widl/widltypes.h b/tools/widl/widltypes.h
>> index 81646cae1d1..bedba40d903 100644
>> --- a/tools/widl/widltypes.h
>> +++ b/tools/widl/widltypes.h
>> @@ -504,6 +504,7 @@ struct _type_t {
>>     const char *c_name;
>>     const char *signature;
>>     const char *qualified_name;
>> +  const char *short_name;
>>     unsigned int typestring_offset;
>>     unsigned int ptrdesc;           /* used for complex structs */
>>     int typelib_idx;
> 
> 
> Do you expect it to use useful outside void write_widl_using_macros()? 
> It looks like something specific to that, so maybe we could just 
> generate it there, when needed.
> 

I'm not sure if we can do that easily there, as we don't have anymore 
the knowledge of which type was parameterized and which not. I'll 
probably keep it like this.

> 
> Thanks,
> 
> Jacek
> -- 
Rémi Bernon <rbernon at codeweavers.com>