Use a consistent size for testharness.js

[gsnedders]

See https://bugs.webkit.org/show_bug.cgi?id=178053

We guarantee a given size for reftests (well, modulo #7598), and not doing so appears to be causing problems for some testharness.js tests, so we should probably just guarantee a size for them too.

cc/ @youennf @cdumez [I can't find any GitHub account for Darin?] @jgraham @Ms2ger @foolip

[foolip]

Makes sense. What should the size be?

[gsnedders]

I suggest we go for the same as we decide in #7598 to be consistent everywhere.

[foolip]

https://chromium-review.googlesource.com/c/chromium/src/+/768430 is a review where this is a problem. What is the size we'd guarantee, is it 800x600?

[Ms2ger]

I think 600×600, but I thought that was a minimum, not a guaranteed size. I'm definitely not using that when running tests in the browser manually. Can that dialog test compute the expected result dynamically?

[foolip]

Yes, the dialog test could be written to work with any viewport size above some minimum size. But that would amount to saying that this issue is WontFix, and tests having to deal with it.

[gsnedders]

Practically this is nowadays RFC material, and see web-platform-tests/rfcs#188 for an issue on the RFCs repo.