Some jasmine tests fail depending on the screen resolution of the computer running the test

[monfera]

When running tests in the Default resolution of a 13" MacBook, all of click_test.js passed; however, four failures were reported after increasing the resolution to the maximum in System Preferences. After much fiddling on my part, @etpinard suggested it may be to do with the browser window size. Indeed, the browser window was of a different size, depending on the resolution:

• Default resolution: 1035 x 617
• Highest resolution: 792 x 867

Manually resizing the already existing Chrome window and rerunning the test made it pass, confirming the suspicion that failure is due to less than expected viewport width.

[mdtusz]

Interesting. Is this because of the viewport width (and plots are being rendered at a different pixel-width?) or because of how OSX/Chrome handles pixel scaling when using the increased resolution mode (as in, does it use real pixel values or visible pixel values?).

In any case, it would probably be a good idea if we make sure all our tests have explicitly set widths and heights.

[monfera]

@mdtusz For this specific issue, it's the former. The devicePixelRatio is the same in both resolutions, the width was different (and Chrome even gave a different aspect ratio for the higher resolution mode). But in general it's true that there are multiple variables to consider, e.g.

• The devicePixelRatio is important for Canvas and WebGL, because, without scaling up with it, the output will be pixelated (aliased). On a multimonitor setup, it can be different per monitor.
• The browser zooming level is another variable. Maybe it's possible for it to get stuck at a non-100% level, failing a bunch of test cases.
• Minimum font size might also impact pixel based comparisons, and potentially even jasmine tests.

I haven't tried to address these other variations in this little PR.

My $.02 is it's useful to set the screen size in test cases, e.g. to test window size responsiveness. It's also useful to have a basic window size for those type of tests that are not width/height sensitive or haven't (yet?) been sized explicitly. My immediate motivation was to be able to run the test suite without errors so it's but a small step.