[PATCH v2 3/5] wined3d: Move the fill mode to wined3d_rasterizer_state.

[Zebediah Figura]

On 3/6/20 10:33 AM, Henri Verbeet wrote:
> On Fri, 6 Mar 2020 at 05:04, Zebediah Figura <z.figura12 at gmail.com> wrote:
>> +    if (set_rasterizer_state)
>> +    {
>> +        struct wined3d_rasterizer_state *rasterizer_state;
>> +        struct wined3d_rasterizer_state_desc desc;
>> +        struct wine_rb_entry *entry;
>> +
>> +        desc.fill_mode = state->rs[WINED3D_RS_FILLMODE];
>> +        desc.front_ccw = FALSE;
>> +        desc.depth_bias_clamp = 0.0f;
>> +        desc.depth_clip = TRUE;
>> +
>> +        if ((entry = wine_rb_get(&device->rasterizer_states, &desc)))
>> +        {
>> +            rasterizer_state = WINE_RB_ENTRY_VALUE(entry, struct wined3d_rasterizer_state, entry);
>> +            wined3d_device_set_rasterizer_state(device, rasterizer_state);
>> +        }
> ...
>> +static int wined3d_rasterizer_state_compare(const void *key, const struct wine_rb_entry *entry)
>> +{
>> +    const struct wined3d_rasterizer_state *state = WINE_RB_ENTRY_VALUE(entry, struct wined3d_rasterizer_state, entry);
>> +
>> +    return memcmp(&state->desc, key, sizeof(state->desc));
>> +}
> 
> This works, but note that it's a little fragile; you're depending on
> there not being any holes or padding in the
> wined3d_rasterizer_state_desc structure.
> 

Sure; I saw similar places using memcmp(), and figured since there
shouldn't be holes (at least on x86...) it'd be safe. Is it worth changing?

[And if so, is it worth changing e.g. glsl_blitter_args_compare()?]