I think it would be good for the test suite to pass when layout.css.devPixelsPerPx > 1, or at least for 2. There are currently many failures with the pref set to 2: https://tbpl.mozilla.org/?tree=Try&rev=f349f614c174 which has 212 reftest, ~80 mochitest failures on OS X 10.8.

We have a similar problem with metro tests. EventUtils coordinates are not sent to the right place causing various failures. We've disabled hidpi for test runs in case people run tests locally on hidpi hardware. (bug 886624)

I think we should consider annotating all existing failures with fails-if() and standing up machines on tbpl running tests in hidpi mode.

Agreed, we would love to see hidpi hardware in our slave pool running windows 7/8. Not sure how to kick off a discussion on the subject though, maybe a dev.platform post?

This is relevant to my interests. There are some bugs that are affecting hidpi devices on Fennec and B2G and are probably related to this.

The results are no longer available for comment 0's Try run, so I triggered another Try run with the same patch (linux64-only, reftests/mochitests-only, to save resources): https://tbpl.mozilla.org/?tree=Try&rev=f349f614c174

Sorry, copypaste fail -- I linked to comment 0's Try run instead of my new one. :) The new Try run is: https://tbpl.mozilla.org/?tree=Try&rev=1d42ca49841d

Thanks Daniel. Interesting that it causes a continuous invalidation loop resulting in exceeding log file size on one of my SVG text tests. I'll see if I can get some time next week to look at bug 888906, which is what is causing some of the failures there I looked at (reftest snapshot not being taken at native resolution).