[PATCH vkd3d 1/7] build: Use linker version scripts to control visibility.

[Józef Kucia]

On Mon, Jan 15, 2018 at 5:59 PM, Henri Verbeet <hverbeet at gmail.com> wrote:
> On 15 January 2018 at 17:55, Józef Kucia <joseph.kucia at gmail.com> wrote:
>> On Mon, Jan 15, 2018 at 4:03 PM, Henri Verbeet <hverbeet at gmail.com> wrote:
>>> On 15 January 2018 at 13:49, Józef Kucia <joseph.kucia at gmail.com> wrote:
>>>> +++ b/libs/vkd3d/vkd3d.map
>>>> @@ -0,0 +1,18 @@
>>>> +{
>>>> +global:
>>>> +    vkd3d_create_device;
>>>> +    vkd3d_create_image_resource;
>>>> +    vkd3d_create_instance;
>>>> +    vkd3d_create_root_signature_deserializer;
>>>> +    vkd3d_get_vk_device;
>>>> +    vkd3d_get_vk_format;
>>>> +    vkd3d_get_vk_instance;
>>>> +    vkd3d_get_vk_physical_device;
>>>> +    vkd3d_get_vk_queue;
>>>> +    vkd3d_get_vk_queue_family_index;
>>>> +    vkd3d_instance_decref;
>>>> +    vkd3d_instance_incref;
>>>> +    vkd3d_serialize_root_signature;
>>>> +
>>>> +local: *;
>>>> +};
>>> Should these maps use a named version, e.g. VKD3D_1.0?
>>
>> They could, but I'm not sure if this is necessary. I think we do not
>> plan to use symbol versioning for incompatible ABI changes. The intent
>> is to not introduce any backward incompatible ABI changes after the
>> the first release. If we ever want to revisit API and/or introduce
>> backward incompatible ABI changes, we'll probably just bump the soname
>> version.
> Yeah, but I think a named version would make it easier to change our
> minds about that in the future.

Sure, I'll prepare a patch.