Cap LCP load duration to LCP time

[tunetheweb]

Fixes #509

Videos can have LCP time before finished download so the resourceLoadDuration should be capped to LCP time.

I can't see an easy way to add a test case for this, but think the change is fairly innocuous.

[brendankenny]

I can't see an easy way to add a test case for this, but think the change is fairly innocuous.

How hard would it be to add a video LCP test?

[tunetheweb]

I can't see an easy way to add a test case for this, but think the change is fairly innocuous.

How hard would it be to add a video LCP test?

Well it's not just that. We'd need to load a video, and draw the LCP from the first frame, but ensure the video is still downloading after that LCP. That sounds incredibly racy. Plus the implementation in Chrome seems to be broken at the moment which Ian is looking into.

Plus need to consider other browsers (e.g. Firefox), which don't support video yet (I don't think) so might be OK, but another complication...

[philipwalton]

Well it's not just that. We'd need to load a video, and draw the LCP from the first frame, but ensure the video is still downloading after that LCP. That sounds incredibly racy.

I'm not sure if it's possible in Express, but it seems like we should be able to create a stream object that reads the video file on disk and adds an arbitrary setTimeout half way through to simulate this case without having to rely on the racy-ness of the network.

[tunetheweb]

I'm not sure if it's possible in Express, but it seems like we should be able to create a stream object that reads the video file on disk and adds an arbitrary setTimeout half way through to simulate this case without having to rely on the racy-ness of the network.

Still depends on Chrome working as it's supposed to though. So point is kinda moot for now.

[tunetheweb]

Raised this bug for the Chrome issue: https://issues.chromium.org/issues/364860066

Once that's resolved (and that version of Chrome feeds through to our test suite) we should see how to add a test, but until then I still think we should merge this change in preparatation.

[brendankenny]

Should this wait until we have a clear picture of what exactly ships (and when) in Chrome?

[tunetheweb]

Should this wait until we have a clear picture of what exactly ships (and when) in Chrome?

I don't see any reason NOT to cap it at the metric value anyway. Especially as I know of at least one reason (that's supposed to have shipped already).

[philipwalton]

Agreed. This seems low risk and may help avoid future bugs.