WebVR: Write automated test to catch regression of 680209 |
||||
Issue descriptionIssue 680209 is currently missing a regression test. This layout test covers the same scenario, but since the issue was purely visual (WebVR thought it was no longer presenting, it's just that the overlay was stuck), the layout test didn't catch it https://cs.chromium.org/chromium/src/third_party/WebKit/LayoutTests/vr/requestPresent_resolve_then_reject.html. Worst case scenario, this can be tested with an image diff test, but we currently don't have a stable way of doing that (VR screenshots are prone to being slightly different from each other, and all the image diff tools that I'm aware are used for Chromium testing are 1:1 pixel comparisons)
,
Feb 23 2017
In theory, that should be pretty easy, although not currently supported by Render Tests (what I've been told to use for image diffing tests for Android instrumentation tests) https://cs.chromium.org/chromium/src/chrome/test/android/javatests/src/org/chromium/chrome/test/util/RenderUtils.java?q=RenderUtils+package:%5Echromium$&l=28. I can see about adding support for that in the near-ish future.
,
Feb 24 2017
While making the changes to RenderUtils, it occurred to me that the new functionality is actually unnecessary in this case since we are testing to make sure the overlay is gone - as long as the webpage isn't reactive to device movement, any screenshots of the page while not presenting should be identical. I'll still go ahead and add the functionality since it will be useful when we need to test the presence of the overlay, which we should be able to do by comparing things like the corners and center line.
,
Aug 9 2017
Brian, same question as another similar bug - should we bother with this, given the age of the bug?
,
Aug 9 2017
I think it's also safe to close this out.
,
Jul 4
|
||||
►
Sign in to add a comment |
||||
Comment 1 by ddorwin@chromium.org
, Feb 23 2017Labels: -OS-Android -Type-Feature Type-Task