1393912 - Register the pending input event annotator on the main thread

Status: RESOLVED FIXED

(Core :: XPCOM, enhancement)

Description

[Nika Layzell [:nika] (ni? for response)]

In bug 1384238, the pending input event annotator was added, but unfortunately it was registered on the wrong thread.

This patch makes us register it on the main thread, so the annotation actually starts showing up.

Comment 1

[Nika Layzell [:nika] (ni? for response)]

Attachment: Register the pending input event annotator on the main thread

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/38498a94b5bd
Register the pending input event annotator on the main thread, r=smaug

Comment 3

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/38498a94b5bd

Comment 4

[Nika Layzell [:nika] (ni? for response)]

Any chance you could take another shot at integrating this into hangs.html :)?

Comment 5

[Alex Thayer [:alexical] (she/her)]

Should be up now. Just for my understanding - is this not feasible for the parent process? Should I keep using the old UserInteracting annotation for the parent, or do we have plans to get something similar for it at some point?

Comment 6

[Nika Layzell [:nika] (ni? for response)]

I don't know how I would go about doing this in the parent process. Do you have some idea of what it would look like olli?

Comment 7

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

If UserInteraction annotation uses user-interaction-active/inactive notifications, it is rather useless.
But perhaps better than nothing. However, it is very very different to pending input event check, so the UI should be different there. If it is difficult to have different kinds of UIs, perhaps parent process shouldn't have any indicator about input event handling.

I wonder if OSes have something useful here, like do they expose useful timestamps in their low level events?

What kinds of hangs do we get in parent process?

Comment 8

[Nika Layzell [:nika] (ni? for response)]

(In reply to Olli Pettay [:smaug] from comment #7)
> If UserInteraction annotation uses user-interaction-active/inactive
> notifications, it is rather useless.
> But perhaps better than nothing. However, it is very very different to
> pending input event check, so the UI should be different there. If it is
> difficult to have different kinds of UIs, perhaps parent process shouldn't
> have any indicator about input event handling.

That's what I'm inclined to do right now. The UserInteracting annotation does in fact use user-interaction-active/inactive and is active most of the time :-/.

> I wonder if OSes have something useful here, like do they expose useful
> timestamps in their low level events?

Yes, we do have timestamps for events at least on windows, but we don't currently have any way in BHR to go back in time and tag previously detected hangs after they have been detected. I could work on implementing that if we decide we want it though.

> What kinds of hangs do we get in parent process?

In the parent process, the majority of hangs right now are due to unnamed runnables (which are currently bugged in the UI - I filed https://github.com/squarewave/bhr.html/issues/15). 

After that, the next big lump of runnables which fail are nsPipeInputStream::Wait runnables, which seem to generally hang in nsInputStreamPump::OnInputStreamReady.