1644284 - Investigate postponing setTimeout/Interval runs when there is important user input pending

Status: RESOLVED WONTFIX

(Core :: DOM: Core & HTML, enhancement, P3)

Description

[Olli Pettay [:smaug][[e-mail]]]

In some cases timeout callbacks get run while there might be more important user input
(non-move events basically) pending.
Same issue applies to xhr.onload and fetch() and such. But perhaps worth to investigate if setTimeout/Interval case could be improved, and whether that could help with responsiveness.

Comment 1

[Olli Pettay [:smaug][[e-mail]]]

Note, input events do already have their own queue. So it is possible that it should be tweaked instead. Or that setTimeout/Interval should get their own priority in the main thread (even though they already use a ThrottledEventQueue).

Comment 2

[Sean Feng [:sefeng]]

We started this task by implementing a telemetry to report the number of pending input events in IPC channel when the timeout handler is about to run.

Looking at the first 2 days data, 99% were 0, so perhaps we don't need to do any tweaks since most of the time there are just 0 pending inputs. But yeah, we should check the graph again once we have more data.

Comment 3

[Sean Feng [:sefeng]]

latest beta data https://telemetry.mozilla.org/new-pipeline/dist.html#!cumulative=0&end_date=2020-07-10&include_spill=0&keys=__none__!__none__!__none__&max_channel_version=beta%252F79&measure=PENDING_CRITICAL_INPUT_WHEN_TIMEOUT&min_channel_version=nightly%252F70&processType=*&product=Firefox&sanitize=1&sort_by_value=0&sort_keys=submissions&start_date=2020-06-29&table=0&trim=1&use_submission_date=0

It just looks like we don't need to worry about this, as the majority is just having 0 pending input event.

What do you think Olli?

Comment 4

[Olli Pettay [:smaug][[e-mail]]]

I think the current input event prioritization is good enough.