[PATCH v3 4/5] d3dcompiler: Implement d3d12_shader_reflection_GetResourceBindingDesc[ByName]().

[Paul Gofman]

On 9/1/20 17:43, Matteo Bruni wrote:
> On Tue, Aug 11, 2020 at 2:31 PM Paul Gofman <pgofman at codeweavers.com> wrote:
>> Signed-off-by: Paul Gofman <pgofman at codeweavers.com>
>> ---
>> v2:
>>      - fix reflection test in d3d10.
>> v3:
>>      - test more precisely how interface aliasing works for GetResourceBindingDesc.
>>
>> +        if (ref12)
>> +        {
>> +            memset(&desc12, 0, sizeof(desc12));
>> +            desc12.Space = 0xdeadbeef;
>> +            desc12.uID = 0xdeadbeef;
>> +
>> +            hr = ref12->lpVtbl->GetResourceBindingDescByName(ref12, pdesc->Name, &desc12);
>> +            ok(hr == S_OK, "Got unexpected hr %x, i %u.\n", hr, i);
>> +
>> +            ok(!strcmp(desc12.Name, pdesc->Name), "Got unexpected Name \"%s\", i %u.\n", desc12.Name, i);
>> +            ok(!memcmp(&desc12.Type, &desc11.Type, sizeof(desc11) - offsetof(D3D11_SHADER_INPUT_BIND_DESC, Type)),
>> +                    "D3D11 and D3D12 descs do not match.\n");
>> +            ok(desc12.Space == pdesc->Space, "Got unexpected Space %u, i %u.\n", desc12.Space, i);
>> +            ok(desc12.uID == pdesc->uID, "Got unexpected uID %#x, i %u.\n", desc12.uID, i);
>> +
>> +            memset(&desc12, 0, sizeof(desc12));
>> +            desc12.Space = 0xdeadbeef;
>> +            desc12.uID = 0xdeadbeef;
>> +            hr = ref12_from_d3d11->lpVtbl->GetResourceBindingDescByName(ref12_from_d3d11, pdesc->Name, &desc12);
>> +            ok(hr == S_OK, "Got unexpected hr %x, i %u.\n", hr, i);
>> +
>> +            ok(!strcmp(desc12.Name, pdesc->Name), "Got unexpected Name \"%s\", i %u.\n", desc12.Name, i);
>> +            ok(!memcmp(&desc12.Type, &desc11.Type, sizeof(desc11) - offsetof(D3D11_SHADER_INPUT_BIND_DESC, Type)),
>> +                    "D3D11 and D3D12 descs do not match.\n");
>> +            /* Native d3dcompiler_47 returns the same interface pointer when queried for ID3D12ShaderReflection
>> +             * from ID3D11ShaderReflection. Space field still gets the correct value on x64 due to
>> +             * D3D11_SHADER_INPUT_BIND_DESC padding. */
>> +            expected = offsetof(D3D12_SHADER_INPUT_BIND_DESC, Space) < sizeof(D3D11_SHADER_INPUT_BIND_DESC)
>> +                    ? pdesc->Space : 0xdeadbeef;
>> +            todo_wine_if(expected == 0xdeadbeef) ok(desc12.Space == expected, "Got unexpected Space %u, i %u.\n",
>> +                    desc12.Space, i);
>> +            todo_wine ok(desc12.uID == 0xdeadbeef, "Got unexpected uID %#x, i %u.\n", desc12.uID, i);
> Interesting, and also worrisome... What happens in the opposite case
> i.e. ref11_from_d3d12? I assume it expects and overwrites the whole
> D3D12_ struct, given that's the same vtbl.
>
> Actually I hacked a quick test and that seems to be the case.

As far as I interpret the tests, the logic behind this behaviour is:

- the interface implementation is identical, querying d3d12 for d3d12 
and vice versa is equivalent a pointer cast;

- the size of the structure to be copied in _GetResourceBindingDesc() 
(as well as the values for d3d12 only desc structure member) is 
determined during the reflection structure creation in D3DReflect.