Collect + alert data on how frequently we hit the 5s limit for queued postMessages / how often we can't find idle time for prolonged periods
Categories
(Core :: DOM: postMessage, task)
Tracking
()
People
(Reporter: Gijs, Unassigned)
References
Details
In bug 1840762 we hit some PDF.js issues because before the load event fires we queue postMessage for either load or idle tasks, rather than firing them immediately. If load takes more than 5 seconds we start dispatching anyway - but this can still seem very long in some cases (like the PDF.js one in the aforementioned bug). We've had other issues where on machines with higher display refresh rates we've struggled to find idle time because of the increased frequency of refresh driver / vsync. It's also possible that the 5s cutoff is too long or too short, but right now we're flying a bit blind in this department.
It would be nice if we had data (and ideally alerting) that we could use to (a) calibrate the 5s timeout as needed and (b) alert/notice if we regress/improve how frequently people hit this case.
Description
•