search

soundcloud / areweplayingyet

html5 audio benchmarks
BSD 2-Clause "Simplified" License
311 stars 69 forks source link

timeupdate test fails in Opera #10

Closed foolip closed 13 years ago

foolip commented 13 years ago

http://areweplayingyet.org/support-consistent-timeupdate-interval fails for presumably bogus reasons, since we (Opera) do what the spec says. Note that several algorithms fire the timeupdate event regardless of when the last one was fired, which could cause it to fire with 0 ms delay.

tsenart commented 13 years ago

Given the context and code path it should be safe to assume that the timeupdate events that are being triggered are related with audio being played. Other browsers also assumed this bug. If you have any suggestion of how to test this better please contribute.

foolip commented 13 years ago

I jumped the gun a bit on this one, it's a bit tricky. It turns out that Opera fires a timeupdate event in the transition to HAVE_ENOUGH_DATA, right after one sometimes starts autoplaying. This is immediately followed by the timeupdate event triggered by calling play(). If I change this to not fire the timeupdate event a local copy of your test passes, but http://areweplayingyet.org/support-consistent-timeupdate-interval still eventually fails. I think that since you wait for 50 timeupdate events there's some timeout in the test framework that happens before 50 events have been fired.

I'll fix the Opera bug.

yvg commented 13 years ago

Changed labels to "Audio Browser Bug" & "Reported to vendor"

foolip commented 13 years ago

It's also a bug in the testsuite, since it times out before the pass condition.

tsenart commented 13 years ago

Oops. Thanks! Fixed.

yvg commented 13 years ago

Can we close this?