Re: [PATCH 2/7] wined3d: Fix handling of SM4+ integer registers.

[Henri Verbeet]

On 23 February 2017 at 15:00, Józef Kucia  wrote:
> @@ -2972,40 +3016,27 @@ static void shader_glsl_add_src_param_ext(const struct wined3d_shader_instructio
>      shader_glsl_get_register_name(&wined3d_src->reg, glsl_src->reg_name, &is_color, ins);
>      shader_glsl_get_swizzle(wined3d_src, is_color, mask, swizzle_str);
> - if (wined3d_src->reg.type == WINED3DSPR_IMMCONST
> -
>        || wined3d_src->reg.type == WINED3DSPR_PRIMID
> -
>        || wined3d_src->reg.type == WINED3DSPR_LOCALTHREADINDEX
> -
>        || wined3d_src->reg.type == WINED3DSPR_THREADID
> -
>        || wined3d_src->reg.type == WINED3DSPR_THREADGROUPID
> -
>        || wined3d_src->reg.type == WINED3DSPR_LOCALTHREADID)
> - switch (wined3d_src->reg.type)
> {
> -
>    shader_glsl_gen_modifier(wined3d_src->modifiers, glsl_src->reg_name, swizzle_str, glsl_src->param_str);
> -
>    case WINED3DSPR_IMMCONST:
> -
>        param_data_type = data_type;
> -
>        break;

This matches the existing behaviour, and in that regard this is fine,
and it probably won't matter in practice, but note that
shader_glsl_get_register_name() uses "reg->data_type", which isn't
necessarily the same as "data_type" here.

I think the usage of sprintf() is more problematic. That too matches
the existing code, but we should be using string_buffer_sprintf() for
this kind of thing.