wined3d public api cleanup

[Riccardo Bortolato]

2015-09-11 16:48 GMT+02:00 Henri Verbeet <hverbeet at gmail.com>:
> On 11 September 2015 at 09:08, Riccardo Bortolato <rikyz619 at gmail.com> wrote:
>>
>> Here is a follow up on the small fixes I sent the other day.
>>
>> I'm storing the wined3d texture and sub level in the d3dX surfaces/volumes
>> and this allows to go further with the removal of surfaces/volumes from the public api.
>>
>> With these commits I also introduce a temporary function to handle mapping
>> for surfaces and volumes which I hope it will gradually handle more of that
>> eventually delegating only bound checks (and maybe other minor things) to helper functions.
>>
>> Finally the remaining volume-related exported functions are removed.
>>
> At first sight this looks mostly ok. You may want to merge patches 1/6
> and 2/6, since they mostly seem to be touching the same places and are
> closely related anyway. Once you pass the wined3d texture to
> device_parent_surface_created()/device_parent_volume_created() you no
> longer need to pass "container_parent", since you can just call
> wined3d_texture_get_parent() on the wined3d texture to get it. You'll
> probably want to rename "idx" to "sub_resource_idx" and just make it
> unsigned int.
>

all right

> In patch 3/6 the decision to use wined3d_volume_map() or
> wined3d_surface_map() should probably go through wined3d_texture_ops.

ok

> Just pass a wined3d_box structure insead of a void pointer. (And
> passing a wined3d_box structure to wined3d_surface_map() should
> probably be a preparation patch for that.)

I kept the RECT * because I get 24 hits in wined3d_private.h, it felt
like the change was too big and needed better thinking but I can
reverse the decision and pass around a wined3d_box * instead.

> Usage and introduction
> should go together in patches 3/6 and 4/6, but you can do map() and
> unmap() in separate patches. You could also potentially further split
> that per-dll/sub-resource type (E.g. "d3d9: Use wined3d_texture_map()
> in d3d9_surface_LockRect()."), although I don't think the patch is
> excessively large as it is.
> wined3d_texture_map()/wined3d_texture_unmap() is fine for the function
> names, no need for the extra "sub_resource" there.
>
>> -    hr = wined3d_surface_unmap(surface->wined3d_surface);
>> +    wined3d_texture_sub_resource_unmap(surface->wined3d_texture, surface->sub_resource_idx);
>>      if (SUCCEEDED(hr) && surface->surface_desc.ddsCaps.dwCaps & DDSCAPS_PRIMARYSURFACE)
> You're dropping the return code here.
>
> You can drop the __cdecl in patch 5/6. I think
> wined3d_volume_get_resource() should be unused after patch 6/6, and I
> think you should be able to just use volume->wined3d_texture instead
> of volume->texture->wined3d_texture.
>
>> What to do with buffers is still unclear to me.
>>
> I think it's mostly of later concern, after surfaces and volumes are
> unified, but the idea is to eventually have something like
> wined3d_resource_map() that just ignores the sub-resource index for
> buffers. (Or perhaps requires it to be 0.)

OK. I'll do the RECT to wined3d_box changes and then re-apply those
with your suggested changes.

Greets,
Riccardo