Wiggling the mouse can make the whole display freeze

[Daniel van Vugt]

Having spent a lot of time observing input latency I've also noticed a bug:

1. Run an input-driven client like mir_demo_client_eglsquare or mir_demo_client_target.
2. Wiggle the mouse around constantly for a minute or so.

Expected: Mouse always moves and client responds.
Observed: Roughly a couple of times per minute the cursor and the client freezes for about one second.

[Daniel van Vugt (vanvugt)]

Actually, as the cursor is freezing too it would be the Mir server that's introducing the pause.

[Daniel van Vugt (vanvugt)]

The problem appears to get a bit worse with a higher input event rate (kernel parameter usbhid.mousepoll=1).

[Daniel van Vugt (vanvugt)]

This is happening a lot with mir_demo_client_eglsquare now. Strangely no longer happening in mir_demo_client_target.

[Andreas Pokorny (andreas-pokorny)]

We recently changed the way mir does pointer confinement .. this could have a positive effect on this issue.

[Andreas Pokorny (andreas-pokorny)]

.. but the hickups still exist and they are sometimes longer than a second.

[Daniel van Vugt (vanvugt)]

Found a reliable way to make this bug really obvious, bordering on unusable:

Start mir_demo_client_animated_cursor and Alt+drag it around (from the middle of the window), such that the cursor is animating while you're dragging. Mir will spend more time freezing than animating.

Maybe it's a different issue, but seems like something that should be investigated first.

[Daniel van Vugt (vanvugt)]

Also just noticed this happens very easily during mir_glmark2_performance_test. If you wiggle the mouse, the display freezes and even the test fails as a result.

[Daniel van Vugt (vanvugt)]

All of the test cases mentioned now work, and never freeze.

I'm bisecting to see if we fixed this in Mir, by accident (!).

[Daniel van Vugt (vanvugt)]

Reopened due to a duplicate bug report.

[Daniel van Vugt (vanvugt)]

It appears kernel bug 1625912 might explain why I can't reproduce this any more.

[Alan Griffiths (alan-griffiths)]

I can't reproduce this. Did we no notice it getting fixed?

[Daniel van Vugt (vanvugt)]

Not sure, but kernel bug 1625912 certainly made it more difficult to reproduce. The problem used to be easy to reproduce with a high rate of input but now that option is broken, so it's hard to tell.

[Launchpad Janitor (janitor)]

[Expired for Mir because there has been no activity for 60 days.]