After working around the test harness issue affecting viewport-resize-event-on-load-overflowing-page.html in bug 1550813, the test now performs the expected assertion, but it still fails. This bug tracks investigating and fixing the cause.

I did look at this failure a bit. We get two resize events on the test case, but the test assumes no resize event receives if the scrollbar thickness is zero (it's zero on our Android). The two resize events come from;

1. a PresShell::SetVisualViewportSize call triggered by adding a viewport meta tag
2. a PresShell::SetResolutionAndScaleTo call triggered by a before-first-paint event (the event calls MobileViewportManager::SetInitialViewport and the function calls RefreshViewportSize and it calls UpdateResolution)

As for 1), it's the first time we set the visual viewport size, so I think we don't need to fire a resize event for the case.
As for 2), it's the first paint for us, so I think we don't need to fire a resize event either.