AshTestBase non-determinism when initializing display |
||
Issue descriptionAshTestBase has some strange behavior. It sets up a display for tests to use[1], but doesn't actually initialize it. Adding Shell::GetInstance()->display_manager()->UpdateDisplays(); after the ash_test_helper_->SetUp(start_session_, material_mode_); call will initialize the display. Making this change breaks 10 or so tests, but I think it exposes some real issues, such as in the autoclick tests. The autoclick tests break because initializing the display generates some mouse events. I suspect that this is the reason behind the autoclick flakes. I'm planning on uploading a CL to test this fix for autoclick tests only. Other tests also fail for various reasons (I haven't fully investigated), ie, because the primary monitor is being set by AshTestBase instead of by the test itself. I don't think fixing this will take all that long, but I don't have the time right now, so I'm filing a bug. 1: https://cs.chromium.org/chromium/src/ash/test/ash_test_base.cc?q=ash_test_base&dr=C&l=130
,
Aug 18 2016
Display is initialized in Shell::Init. There may be a case that real events affects unit tests, because we don't distinguish them. can that be an issue?
,
Aug 29 2016
|
||
►
Sign in to add a comment |
||
Comment 1 by bugdroid1@chromium.org
, Aug 18 2016