Re: [PATCH v2 0/5] MR5365: win32u: Introduce a new common vulkan source.

[Jacek Caban (＠jacek)]

Jacek Caban (@jacek) commented about dlls/winex11.drv/vulkan.c:
> - LOAD_FUNCPTR( vkCreateSwapchainKHR );
> - LOAD_FUNCPTR( vkCreateXlibSurfaceKHR );
> - LOAD_FUNCPTR( vkDestroyInstance );
> - LOAD_FUNCPTR( vkDestroySurfaceKHR );
> - LOAD_FUNCPTR( vkDestroySwapchainKHR );
> - LOAD_FUNCPTR( vkEnumerateInstanceExtensionProperties );
> - LOAD_FUNCPTR( vkGetDeviceProcAddr );
> - LOAD_FUNCPTR( vkGetInstanceProcAddr );
> - LOAD_FUNCPTR( vkGetPhysicalDeviceXlibPresentationSupportKHR );
> - LOAD_FUNCPTR( vkGetSwapchainImagesKHR );
> - LOAD_FUNCPTR( vkQueuePresentKHR );
>
> +#undef LOAD_FUNCPTR
> -
> - vulkan_hwnd_context = XUniqueContext();
> - *driver_funcs = vulkan_funcs;
> - return vulkan_handle;

IMO it's not great to have the library handle being part of the interface. Perhaps we could just load the vulkan loader in win32u and pass a struct containing common functions to drivers? That wouldn't work for platform-specific functions like `vkCreateXlibSurfaceKHR`. One way to solve that would be to query it using `vkGetInstanceProcAddr` inside `X11DRV_vkCreateWin32SurfaceKHR` where we have the instance handle.

(This would also bring us a step closer to supporting Vulkan in the null driver).