Unable to repro any flakiness on Mac Pro desktop non-retina OSX 10.11.6 simulating retina config via --enable-prefer-compositing-to-lcd-text as add'l driver flag to layout test runner, --repeat-each=100.
See also my comment on another bug rebased: http://crbug.com/637251#c3
Am wondering whether my change has tickled some other issue re: not-consistently-reset-state that leads to flakiness when other tests are executed ahead of these. For example pdr@ recently landed http://crrev.com/2208603004 though I don't think this issue is related to that.
Also +ccameron in case it turns out to be his CL rather than mine. Still investigating.
A bit of a tangent but I really wish we had a way to track the issues of leaving state around between test runs. Every one of these sits around in the tree for a few weeks before anyone gets worked up enough to fix it, then they take at least a full day to really track down because there's no way to blame the original patch. We have the infrastructure that notices when a test passes on retry, but no way to track which preceding test could be causing it.
The auto-rebaseline http://crrev.com/2239283002 has bad rebaseline image for mac-retina details-add-summary-1-and-click-expected.png.
On non-Retina but with --enable-prefer-compositing-to-lcd-text and --reset-results we generate a correct expectation image, and --repeat-each=100 still shows no flakiness. Possibly a bad Retina rebaseline?
I'm preparing a build on my Retina laptop currently.
In case related, trchen@ notes he is reverting a change http://crrev.com/2229303002/ that landed around midnight last night, so roughly similar timeframe to my change, and could have introduced flakiness into LayoutTests.
Comment 1 by kolos@chromium.org
, Aug 12 2016