GitLab

This is a reland of the following with an updated TestExpectations:
https://chromium-review.googlesource.com/c/chromium/src/+/882283

For the same reasons as the have metadata test is marked as pass
or timeout, so to may the have future data test timeout if timing
aligns just right to elide the resume call (not an issue in
practice since the real suspend path elides the suspend call in
this case)..

-----------------------------

Testing is accomplished by a couple new window.internals.* methods
which allow modification and inspection of suspend state:
- forceStaleStateForMediaElement(element)
- isMediaElementSuspended(element)

Unfortunately due to the async nature of suspend the method for
testing suspend must be polled once a idle suspend is initiated via
the forceStaleStateForMediaElement(). Test helper methods have been
added to media/suspend-util.js to alleviate test cruft.

forceStaleStateForMediaElement() does not actually force a suspend,
it simply marks the element as stale. If an element is playing,
seeking, or otherwise in a non-suspendable state, no suspend will
occur. This is intentional to ensure tests cover our state machine
as close to reality as possible.

Unfortunately the only way to force given ready states for the
media element is through throttled loading; this CL makes extensive
use of throttled loading and adds a throttling limit to our layout
test loader to avoid timeouts with slow initial rates. These tests
were run with a --repeat-each=200 several times without flake, but
I'll also watch the flakiness dashboard for these afterward.

As a result of the above combined with the fact that we are not
yet propagating play state to WMPI, we can reach a timeout state
in the 'suspend after metadata' src= test. It's marked as pass
and timeout in TestExpectations until the following lands:

https://chromium-review.googlesource.com/c/chromium/src/+/876924

This is part 1 of 2, the 2nd part will add MSE based variants of
these tests. No new internals methods are expected though, just
a appendBuffer() based mechanism for triggering ready states.

BUG=694855, 756897
TEST=it's all tests!
TBR=mlamouri, peter, foolip

Change-Id: Ie8f45d2d9572fd09c1943dc0674b25835527a1ab
Reviewed-on: https://chromium-review.googlesource.com/895856Reviewed-by: Dale Curtis <dalecurtis@chromium.org>
Commit-Queue: Dale Curtis <dalecurtis@chromium.org>
Cr-Commit-Position: refs/heads/master@{#533534}