AudioContext, HTMLMediaElement, and MediaStream provide currentTime to content scripts

[arthuredelstein]

The worst offender is probably AudioContext, where you can get
the current time thus:

var audioContext = new AudioContext();
audioContext.currentTime;

Similarly, HTMLVideoElement.currentTime, HTMLAudioElement.currentTime, and CanvasCaptureMediaStream.currentTime can each be used as a time source with resolution better than 10 ms.

[arthuredelstein]

Here is a branch with two patches.

​https://github.com/arthuredelstein/tor-browser/commits/19193

b91f47cce42a2e048f970a18710f9c15672a6cfc is a fixup for our #1517 patch that rounds the values returned by the currentTime attributes mentioned above.

936cadcc8dd2b0e6784be69cabe44880e36839b0 provides regression tests for all time sources we are rounding in the old and new patches.

[gk]

Looks good to me.

[gk]

Moving reviews to June

[mcs]

For the regression test patch (936cadcc8dd2b0e6784be69cabe44880e36839b0), the test code references silence.ogg but the patch adds a file named audio.ogg. Shouldn't they match?

The currentTime rounding patch (b91f47cce42a2e048f970a18710f9c15672a6cfc) looks good.

I tried to run the regression tests myself, but mach mochitest is not cooperating at the moment (I think this is due to Mac OS 10.11 Python differences; it is definitely not related to the tests themselves).

[mcs]

Replying to mcs:

I tried to run the regression tests myself, but mach mochitest is not cooperating at the moment (I think this is due to Mac OS 10.11 Python differences; it is definitely not related to the tests themselves).

OK, I was able to run tbb-tests/test_tor_bug1517.html on a different system and it worked correctly. I guess the test does not care about the silence.ogg vs. audio.ogg difference (but it should probably be fixed).

[arthuredelstein]

Replying to mcs:

Replying to mcs:

I tried to run the regression tests myself, but mach mochitest is not cooperating at the moment (I think this is due to Mac OS 10.11 Python differences; it is definitely not related to the tests themselves).

OK, I was able to run tbb-tests/test_tor_bug1517.html on a different system and it worked correctly. I guess the test does not care about the silence.ogg vs. audio.ogg difference (but it should probably be fixed).

Thanks for catching that mistake. I've corrected it here:

​https://github.com/arthuredelstein/tor-browser/commits/19193+1

[gk]

Fixed on tor-browser-45.2.0esr-6.5-1 with commit 49d0c6389eea055e94a8fb492bd95de56b9111a4 and b60b8871fa08feaaca24bcf6dff43df0cd1c5f29, thanks.