Seeing this failure on elm, jaq, and samus in video_ChromeHWDecodeUsed:

https://wmatrix.googleplex.com/reasons?suites=bvt-cq&tests=video_ChromeHWDecodeUsed

Also seeing this error on those same machines:
"Media.GpuVideoDecoderInitializeStatus not loaded or histogram bucket not found or histogram bucket found at < 100%"

There tests seem to be running on some custom toolchain builds:

arm64-llvm-next-toolchain/R56-8935.0.0-rc1/bvt-cq/video_VideoSanity.h264.arc

You're right - all of these failures are on -rc builds.

GoldenEye claims the Android PFQ failed for at least the three most recent of these builds, though the PFQ looks quite green:
https://uberchromegw.corp.google.com/i/chromeos/builders/samus-android-pfq/builds/678

+llozano in case he is running these CQ tests on these bots with custom toolchains.

Rahul, 

can you please take a look? is this only failing when we use LLVM? or when use llvm-next?

thanks for pointing this out Rohit.

Checked toolchain buildbot logs from last couple of weeks. The video_VideoSanity test fails only on the llvm-next buildbots (all four: amd64-llvm-next-toolchain, arm-llvm-next-toolchain, arm64-llvm-next-toolchain, x86-llvm-next-toolchain). Passes on the other toolchain buildbots (both gcc/llvm for all four architectures).

reassigning to yunlian so that he can verify with new llvm-next compiler.

This does not happen anymore.