Andreas Süßenbach

@tomadamatkinson You're right, I somehow missed to get the check out the right version. Works now!

> I don't think we can just assert in all these cases can we? You're kind of right. As I stated in the Description above, that should be replaced by...

> I'm also not really sure what was wrong with just having a single call to request_extension_features? The problem, as described in #938, was, that currently you get the supported...

> Well throwing a `std::runtime_error` would be better, as it doesn't terminate batch mode, just the current sample I guess. Sounds good. I can change that accordingly.

> Will that actually work with e.g. code completion/intellisense? As it is with all macros, you probably don't have any support by code completion or intellisense when you use `[HPP_]REQUEST_[OPTIONAL|REQUIRED]_FEATURE`....

@SaschaWillems I assume, that `pNextDevice` is the `pNext` of some `VkPhysicalDeviceFeatures2`, that you hold somewhere. And you're using a pointer to that `VkPhysicalDeviceFeatures2` to create a `VkDevice` using `vkCreateDevice()`. That...

> If one requests a feature that isn't supported, instance or device creation will fail with an error code hinting at a feature not supported. Interesting approach! That is, you...

@SRSaunders I adjusted the two samples _async_compute_ and _multithreading_render_passes_. But the changes in _dynamic_line_rasterization_ would actually change that sample, which I'ld like to avoid with this already pretty big and...

I agree, the resulting types are somewhat longer than before. - I moved the class Subpass from namespace `vkb` to `vkb::rendering`, to make it clear where it resides. If it's...

Indeed, the `Swapchain` isn't used that often any more. The "unified" class `VulkanSample` holds a `vkb::core::HPPSwapchain` as a member, no matter whether it's providing the C- or the C++-bindings-based interface....