Line drawing / clipping in deqp/functional/gles3/clipping.html fails on all Intel ChromeOS devices

[kenrussell]

It was recently discovered that some of the line drawing tests in the WebGL 2.0 conformance test deqp/functional/gles3/clipping.html fail on Intel GPUs on ChromeOS. Per the Chromium CL which suppressed these failures, https://chromium-review.googlesource.com/c/chromium/src/+/5180916 , it's effectively an off-by-one difference on whether a line going to a fractional screen coordinate fills the fragment at its end, or not. It appears to be caused by a difference in Intel's graphics driver compared to other GPUs. (Apologies, the linked Chromium bug is restricted-view.)

This failure also seems fairly minor - I personally doubt many WebGL applications would break if the rules in this area were relaxed.

Let's compare the current version of this test in Khronos' VK-GL-CTS with the ported WebGL 2.0 conformance test and see if the test has been updated in the intervening years. If so, let's apply those edits - maybe the test failure will be resolved. If not, let's consider loosening the test's requirements since this is a platform with many users and it seems unreasonable to deny WebGL 2.0 conformance for this one reason.

[lexaknyazev]

The first issue is to figure out the origin of this test. These two similarly named source files have little in common at first glance:

• JS: https://github.com/KhronosGroup/WebGL/blob/main/sdk/tests/deqp/functional/gles3/es3fClippingTests.js
• C++: https://github.com/KhronosGroup/VK-GL-CTS/blob/main/modules/gles3/functional/es3fClippingTests.cpp