In most recent failures, logs show that current_time is equal to expected_duration. I think what happens is that the video takes a long time to initialize in VideoPuppeteer, so that it's already done playing by the time _start_time is set to current_time [1]. The definition of playback_started should account for that. The test should also pause playback after the VideoPuppeteer init is done. [1] https://github.com/mjzffr/firefox-media-tests/blob/cdd155c1823a1a669246d3b093996fa3eb09dbc8/media_utils/video_puppeteer.py#L94

Moving discussion over from Bug 1211959: (In reply to Bryce Van Dyk (:SingingTree) from comment #69) > Sure thing, I'm fairly new to these tests still, so any guidance would be > appreciated. > > I'm seeing this fail regularly at the moment on our Windows 8 test machine, > but am uncertain as to what exactly is going on thus far. I haven't looked at debugging this issue in a long time. Does Comment 0 make sense with what you're seeing on Windows 8 lately?

Up-to-date reference for Comment 0 and playback_started: * https://dxr.mozilla.org/mozilla-central/source/dom/media/test/external/external_media_tests/media_utils/video_puppeteer.py#288

I'm seeing some conflicting results when I watch the tests run, but I do understand comment 0, and am going to investigate further. I also haven't looked at the linux servers in any detail, and it seems like we have fails happening on them currently too, so I'll do some digging there.

The media tests have been refactored such that this work is not needed any more.