UseCase gets stuck in LOADING after navigating within Youtube |
|||||||
Issue descriptionIf you go to youtube.com and navigate to any video, the RendererScheduler will stay in UseCase::LOADING indefinitely. This is because clicking a video on youtube does not actually navigate to a new page but uses window.history.pushState and then modifies the current page. This doesn't reset the FirstMeaningfulPaintDetector which will never fire again because it thinks the network connection for the page is quiescent. While in theory we could reset the FMPD, this may not give the results we want since the page is now partially painted anyway, so there may not be a more significant paint. The solution is almost certainly to ignore such navigations and not enter LOADING at all in those cases.
,
Jun 23 2017
According to reports this may have regressed?
,
Jul 5 2017
I can confirm that this bug is back.
,
Jul 5 2017
,
Jul 5 2017
Looks like Alex added the bug back in https://chromium-review.googlesource.com/c/532902/ as part of fixing the perf regressions of his original patch (which only fixed this bug as a side-effect).
,
Jul 5 2017
I found this using the attached test page, rather than youtube which doesn't seem to be fixed at all by any of the previous changes. It's possible, the original change never fixed it or youtube itself has changed.
,
Jul 5 2017
Attaching correct version of test page.
,
Jan 24 2018
,
Aug 15
,
Aug 17
,
Oct 2
Can't repro this anymore on Linux (and I assume all platforms are the same), whether using the test from #7 or on youtube.com. In the test from #7, the use case never switches to loading again when clicking the buttons. |
|||||||
►
Sign in to add a comment |
|||||||
Comment 1 by delph...@chromium.org
, Jun 8 2017Status: Fixed (was: Started)