Once the test framework is capable of running WebXR tests, any existing WebVR tests that have equivalents in WebXR should be copied and made to work with WebXR. There are several WebVR tests that don't make sense for WebXR, but the vast majority do.
Blocking this on the final cutover from Win7 to Win10. At that point we can put vr_browser_tests on win_optional_gpu_tests_rel and add that optional tryserver to PRESUBMIT.py in all directories where it would be relevant.
Note: it's now possible to put vr_browser_tests on either win_optional_gpu_tests_rel or win7_chromium_rel_ng, depending on how long it takes to run and how frequently it might get broken. If the tests are reliable and don't take that long, they should just be put on win_chromium_rel_ng per comment on https://chromium-review.googlesource.com/906839 .
Thanks! They currently only take ~25 seconds to run, and while that will go up over time as more tests are added, they should always be pretty fast.
As for stability, I'm not seeing any failures in the last 200 builds other than the time we accidentally broke them, so I think they're stable enough to put on win7_chromium_rel_ng.
That depends on how much test coverage is required for an origin trial (and if we're still shooting for an M66 WebXR origin trial) - the last bit of this bug is to re-add the magic window tests which were catching a segfault. If we need end-to-end tests for that in order to launch the origin trial, then yes, we need to land these for M66.
The only remaining unported instrumentation tests are ones related to autopresent/NFC, which doesn't have an equivalent in the WebXR spec (at least not yet). So, I'd say all the work for this for M67 is done.
Comment 1 by bsheedy@chromium.org
, Jan 20 2018