The GpuProcess.equal_bug_workarounds_in_browser_and_gpu_process test from src/content/test/gpu/gpu_tests/gpu_process.py and src/content/test/gpu/page_sets/gpu_process_tests.py is failing unexpectedly in https://codereview.chromium.org/2318313004/ .

During a first launch of Chrome, there is no cached GPU information available, so the browser's GPU blacklist is computed solely based on the GPU vendor and device IDs. Later, when the GPU process starts, the OpenGL vendor and version become available, and workarounds that are dependent on that can be applied.

In https://codereview.chromium.org/2318313004/ a new driver bug workaround is being added for OpenGL versions less than 4.4. The workaround is being enabled in the GPU process but not the browser process. This is somewhat expected -- though to be honest it's not 100% clear to me why it's being enabled in the GPU process on Windows, since it's supposed to only take effect for desktop GL, and not GLES, which is what ANGLE exposes.

On Linux and macOS it's easier to understand why this workaround is enabled in the GPU process. Upon first launch, I think it is allowed that the browser's and GPU process's notions of the driver bug workarounds are different.

Something needs to be done here. Perhaps a small exception list of workarounds allowed to be different needs to be added.

I'm blocking this on  Issue 634519  since a workaround will probably be added here to allow that CL to go through and that bug to be fixed. This can then be revisited later.