Encoding Video tests crashing on latest drivers in CI

[spencer-lunarg]

In the last few days, we have started to hit CI failures on the following tests

• NegativeVideo.EncodeRateControlH264ConstantQpNonZero
• NegativeVideo.EncodeRateControlH265ConstantQpNonZero
• NegativeVideo.EncodeRateControlH264InvalidCodecInfo
• NegativeVideo.EncodeRateControlH265InvalidCodecInfo

with stack dumps like

ERROR: return code 3 (0x3) - details on failing tests follow
ERROR: ----stderr---------------------------------------------------
abort() has been called
ERROR: -------------------------------------------------------------
ERROR: vvl: CRASH - test NegativeVideo.EncodeRateControlH264InvalidCodecInfo never completed
------------------------------------------------------------
[ RUN      ] NegativeVideo.EncodeRateControlH264InvalidCodecInfo
C:\j\vz1\Debug64\Vulkan-ValidationLayers\tests\framework\binding.cpp(1751): error: Expected equality of these values:
  VK_SUCCESS
  vk::EndCommandBuffer(handle())
    Which is: VK_ERROR_INITIALIZATION_FAILED

running on the NVIDIA GeForce RTX 2080 family HW with the latest NDA drivers

(https://github.com/LunarG/VulkanTests/pull/636)

[spencer-lunarg]

Found this happened when the NVIDIA driver updated to 560.94

cc @aqnuep - we have these to skip on our internal CI until we figure it out, not sure if this a driver regression or the driver is catching something we are doing bad in the layer or tests (if tests, then likely missing a VU to cover it)

[aqnuep]

This is a known driver bug. They are not allowed to return VK_ERROR_INITIALIZATION_FAILED there. If anything, VK_ERROR_INVALID_VIDEO_STD_PARAMETERS_KHR should be returned. I notified them about it. It should be fixed eventually.