[PATCH vkd3d v2 5/5] vkd3d-shader: Make some FIXME messages into proper compiler errors.

[Zebediah Figura (she/her)]

On 8/12/21 10:53 AM, Henri Verbeet wrote:
> On Thu, 12 Aug 2021 at 06:47, Zebediah Figura <zfigura at codeweavers.com> wrote:
>> ---
>> Alternate approach I came up with, that I think I like better.
>>
>> At least I'd like to keep printing messages to the console via FIXME; they're
>> not going to show up otherwise.
>>
> I'm fine with this; I mostly just care that these messages end up in
> the compiler output. That said, is the above really true?
> vkd3d-compiler is certainly going to print error messages by default,
> and presumably e.g. d3dcompiler would print them using its own debug
> facilities, similar to what wined3d does with vkd3d_shader_compile()
> (and e.g. GLSL compiler) messages.

I was going to say that we probably don't want d3dcompiler to always 
print error messages, but on reflection it seems reasonable to print 
error messages iff vkd3d_shader_compile() returned an error. I'll get 
rid of the FIXME part.

> 
>> @@ -112,6 +112,7 @@ enum vkd3d_shader_error
>>       VKD3D_SHADER_ERROR_HLSL_INVALID_RETURN              = 5014,
>>       VKD3D_SHADER_ERROR_HLSL_OVERLAPPING_RESERVATIONS    = 5015,
>>       VKD3D_SHADER_ERROR_HLSL_INVALID_RESERVATION         = 5016,
>> +    VKD3D_SHADER_ERROR_HLSL_NOT_IMPLEMENTED             = 5299,
>>
> The above is probably fine, but note that it would make it harder to
> later split the 5000-5299 range if we wanted to for some reason. At
> the same time, the HLSL error codes are not in any released version of
> vkd3d-shader yet, so you could still grab e.g. 5000 if we want a
> "nice" number for this.
> 

I was just following Matteo's cue here; I don't really see any reason 
not to make it 5017. <shrug>