Fix stalls on WebOS

[avelad]

Fix for #1704

[vaage]

From a technical stand-point, this looks good, but there are some structural changes I would like, but I want @joeyparrish to chime in:

• Could the check for WebOS be in the callback passed into |onStall|? That way the core code of StallDetector won't have platform-specific code.
• Could we have some more documentation around why we are not reacting to stalls on WebOs?
• Long term, @joeyparrish would it make sense to have StallDetector be an optional component that would only be initialized if we are on a platform that should react to stalls.

@avelad Could you offer me some feedback? The files you changed were created from some recent refactoring in which I move the stall detector and platform code into their own classes. Did that make it easier for you to find where to make your changes?

[joeyparrish]

• Could the check for WebOS be in the callback passed into |onStall|? That way the core code of StallDetector won't have platform-specific code.

Given that we have platform checks isolated in shaka.util.Platform now, I don't see a strong need to isolate StallDetector from that. If we want to audit platform-specific logic in the future, it's easy to find these checks by looking for call sites of Platform. If we had more complex logic for when to use StallDetector and when not to use it, I could see doing it differently.

• Long term, @joeyparrish would it make sense to have StallDetector be an optional component that would only be initialized if we are on a platform that should react to stalls?

Yes, long term, I think greater modularity could be nice. Stall detection could even be disabled via configuration at that point. For now, though, it's not needed. We can file that as an issue to come back to at a later time.

• Could we have some more documentation around why we are not reacting to stalls on WebOs?

Yes, please. @avelad, if you could add a brief comment to the code, that would help. I left a suggestion on the code itself.

Looks good to me otherwise! Thanks!

[avelad]

@joeyparrish I update the comment with your suggestion (that is the real problem)

[skipness]

This PR seems fixes the problem on WebOS.
For Tizen 4.0, I found that the stall issue seems related to DRM, by using the same code I can play live stream without any issue on Tizen 3.5 but not on 4.0, I will open an issue if needed.
Thank you for the PR.

[avelad]

thanks for your info @skipness

[shaka-bot]

All tests passed!

[jamescahall]

Hi Joey,

We are getting consistently good results on VOD content on Web OS but have frequent buffering issue when we pass over discontinuities for SSAI ad breaks where bitrates / resolutions change. They eventually recover, but sometimes the buffering can take up to 5-10 seconds which leads to a really bad user experience. We don't see this with other players but need to use ShakaPlayer to support DRM for VOD content (no DRM on live linear) on WebOS. We also experience quite slow load times (5 seconds+).

        var configuration = { 'abr': { 'defaultBandwidthEstimate' : 500000 }, 'streaming': { 'bufferingGoal' : 15, 'rebufferingGoal': 4 } };

I can send a test stream via email but not here.

Thanks
James