[PATCH 3/3] d3dx9/tests: Add tests for array of samplers.

[Matteo Bruni]

2017-07-05 19:45 GMT+02:00 Paul Gofman <gofmanp at gmail.com>:
> On 07/05/2017 08:40 PM, Paul Gofman wrote:
>>
>> On 07/05/2017 08:29 PM, Matteo Bruni wrote:
>>>
>>> 2017-06-28 13:28 GMT+02:00 Paul Gofman <gofmanp at gmail.com>:
>>>
>>> Sorry for the delayed reviews. I did it at last...
>>>
>>>>   #define test_effect_preshader_compare_shader_bytecode(a, b, c, d) \
>>>> @@ -4505,12 +4551,38 @@ static void
>>>> test_effect_preshader(IDirect3DDevice9 *device)
>>>>       hr = IDirect3DDevice9_GetSamplerState(device, 0,
>>>> D3DSAMP_MAGFILTER, &value);
>>>>       ok(hr == D3D_OK, "Got result %#x.\n", hr);
>>>>       todo_wine ok(value == 3, "Unexpected sampler 0 magfilter %u.\n",
>>>> value);
>>>> +
>>>> +    hr = IDirect3DDevice9_GetSamplerState(device, 1, D3DSAMP_MINFILTER,
>>>> &value);
>>>> +    ok(hr == D3D_OK, "Got result %#x.\n", hr);
>>>> +    ok(value == 1, "Unexpected sampler 1 minfilter %u.\n", value);
>>>> +    hr = IDirect3DDevice9_GetSamplerState(device, 1, D3DSAMP_MAGFILTER,
>>>> &value);
>>>> +    ok(hr == D3D_OK, "Got result %#x.\n", hr);
>>>> +    todo_wine
>>>> +    ok(value == 1, "Unexpected sampler 1 magfilter %u.\n", value);
>>>
>>> Not new or really related to this patch: any idea about what's going
>>> on with the magfilter state computation? I didn't really look into it,
>>> does
>>> it have to do with default parameter values by any chance?
>>>
>>>
>> It needs type casting, the output table for FXLC is always float, but the
>> input for state is integer, thus an enormous value failing the test. I was
>> thinking of adding type specifier to state table in effect.c and doing the
>> cast, but did not come to that yet.
>>
> More exactly, the state parameter in sampler state is float for FXLC
> constants for some reason (which is not the case for "normal" states, which
> are assigned ok). Maybe it should be set to correct type during the parse
> actually.

Maybe FXLC preshaders are always supposed to output float, with the
optional conversion after the fact. Not sure it makes any difference
in practice.