1186770 - dom/events/test/test_bug607464.html fails on M-e10s with APZ enabled

Status: RESOLVED FIXED

(Core :: Panning and Zooming, defect)

Description

[Kartikaya Gupta (email:[e-mail])]

A recent try run with APZ enabled on Linux showed that this test was failing. I was able to reproduce locally by running the test in chaos mode, and grabbed an rr recording. The intention of the test is to synthesize a bunch of wheel events that should trigger "instant" scrolling despite smooth-scrolling being enabled. The test checks for this by using requestAnimationFrame callbacks to ensure the scrolling has taken effect by the desired deadline.

With APZ enabled, however, the wheel events get sent to the parent process and handled by APZ. The APZ does handle these wheel events "instantly" (i.e. going through the instant scroll codepath in AsyncPanZoomController) but the scroll offset in the child process doesn't get updated until the repaint requests make it back and get processed. This process is asynchronous relative to the requestAnimationFrame calls in the child process, and so this test is racy and fails intermittently.

The only thing I can think of to salvage this test is to change the way the check is done. Instead of using rAF, perhaps the test should listen for scroll events and make sure that each scroll event happens at a multiple of 3 pixels (i.e. none of the wheel events trigger a scroll of less than 3 pixels, which would happen if there were a smooth scroll animation). I think that should still exercise the thing the test is trying to exercise, but I'm not 100% sure.

Comment 1

[Markus Stange [:mstange]]

Your suggestion sounds fine to me. Let's not make this too complicated, it's not a very valuable test.

Comment 2

[Kartikaya Gupta (email:[e-mail])]

Attachment: Patch

Thanks for the quick response! I updated the test to implement what I said and it seems to work as expected. I also increased the scroll amount from 3px to 30px just to make sure any partial scrolls would be more likely to get caught.

Comment 3

[Legacy TBPL/Treeherder Robot]

log: https://treeherder.mozilla.org/logviewer.html#?repo=try&job_id=9686139
repository: try
start_time: 2015-07-22T20:52:03
who: kgupta[at]mozilla[dot]com
machine: tst-linux32-spot-034
buildname: Ubuntu VM 12.04 try opt test mochitest-e10s-1
revision: c74a24581234

1564 INFO TEST-UNEXPECTED-FAIL | dom/events/test/test_bug607464.html | Pixel scrolling should have finished after one refresh driver iteration. We shouldn't be scrolling smoothly, even though the pref is set. - got 12, expected 15

Comment 4

[Legacy TBPL/Treeherder Robot]

log: https://treeherder.mozilla.org/logviewer.html#?repo=try&job_id=9683881
repository: try
start_time: 2015-07-22T20:31:00
who: kgupta[at]mozilla[dot]com
machine: tst-linux32-spot-316
buildname: Ubuntu VM 12.04 try opt test mochitest-e10s-1
revision: c74a24581234

1564 INFO TEST-UNEXPECTED-FAIL | dom/events/test/test_bug607464.html | Pixel scrolling should have finished after one refresh driver iteration. We shouldn't be scrolling smoothly, even though the pref is set. - got +0, expected 15

Comment 5

[Legacy TBPL/Treeherder Robot]

log: https://treeherder.mozilla.org/logviewer.html#?repo=try&job_id=9683986
repository: try
start_time: 2015-07-22T20:37:41
who: kgupta[at]mozilla[dot]com
machine: tst-linux32-spot-623
buildname: Ubuntu VM 12.04 try debug test mochitest-e10s-1
revision: c74a24581234

1580 INFO TEST-UNEXPECTED-FAIL | dom/events/test/test_bug607464.html | Pixel scrolling should have finished after one refresh driver iteration. We shouldn't be scrolling smoothly, even though the pref is set. - got 12, expected 15

Comment 6

[Legacy TBPL/Treeherder Robot]

log: https://treeherder.mozilla.org/logviewer.html#?repo=try&job_id=9686921
repository: try
start_time: 2015-07-22T21:25:43
who: kgupta[at]mozilla[dot]com
machine: tst-linux32-spot-353
buildname: Ubuntu VM 12.04 try debug test mochitest-e10s-1
revision: c74a24581234

1580 INFO TEST-UNEXPECTED-FAIL | dom/events/test/test_bug607464.html | Pixel scrolling should have finished after one refresh driver iteration. We shouldn't be scrolling smoothly, even though the pref is set. - got 12, expected 15

Comment 7

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/cac48b8498ac

Comment 8

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/mozilla-central/rev/cac48b8498ac