[v10 1/4] d3dx9: Implement fxlc constants (expressions) in effect.

[Paul Gofman]

On 04/11/2016 08:05 PM, Paul Gofman wrote:
> On 04/11/2016 07:52 PM, Matteo Bruni wrote:
>> 2016-04-07 17:41 GMT+02:00 Paul Gofman <gofmanp at gmail.com>:
>>
>>
>>> +static HRESULT set_constants_param(struct d3dx_regstore *rs, struct d3dx_const_tab *const_tab,
>>> +        D3DXHANDLE hc, struct d3dx_parameter *param)
>>> +{
>>> +    ID3DXConstantTable *ctab = const_tab->ctab;
>>> +    D3DXCONSTANT_DESC desc;
>>> +    unsigned int const_count, param_count, i, j, n, table, start_offset;
>>> +    unsigned int minor, major, major_stride, param_offset;
>>> +    BOOL transpose, get_element;
>>> +
>>> +    ID3DXConstantTable_GetConstantDesc(ctab, hc, &desc, &const_count);
>> This isn't correct, although it currently works. const_count is
>> supposed to be an inout parameter for GetConstantDesc() (you could
>> potentially have multiple instances of the same constant in different
>> register sets). The current implementation of GetConstantDesc()
>> doesn't check the parameter and blindly returns only the first
>> descriptor, also setting const_count to 1 (actually the following
>> descriptors aren't even parsed currently by the constant table code
>> IIRC) so this ends up working, although that's a bug and at some point
>> it might get fixed which would in turn break this.
>> Also it's a bit of an "overload" to use the same variable here and as
>> the count of constant elements / members in the following lines, it's
>> probably nicer to use two separate variables.
>>
>> BTW you're using GetConstantDesc() correctly in get_constants_desc().
>>
> I was looking into GetConstantsDesc() implementation to get the idea how
> constants count can be > 1 there, and saw there that the count is
> ultimately 1, that's why left it like that. Do you think it is OK just
> to set count to 1 prior to the call? As I get MSDN text for this count >
> 1 is possible only for samplers and we should not get here for samplers
> anyway.
>
>
Oh, please ignore this, actually the value type constant may also appear
multiple types if it is used in different register index. So I suggest
to initialize count to 1 before the call and put a FIXME if count != 1
on return. Making the real implementation for > 1 will be just a dead
code for now until GetConstantsDesc actually support this.