These reftests are annotated as skip-if(!asyncPanZoom) so currently they are not running on desktop. Enabling APZ makes them run. However, the tests are actually assuming that the meta-viewport tags are being interpreted and the page zoom is being adjusted (see bug 1139306 comment 55 - 57). On desktop, enabling APZ doesn't automatically mean that the meta-viewport tag is applied, so I think the skip-if condition is wrong. We should just run these tests on B2G (i.e. on platforms where APZ and meta-viewport are both enabled).
Created attachment 8626440 [details] [diff] [review] Patch
Comment on attachment 8626440 [details] [diff] [review] Patch If in the future we're going to have more cases of reftests needing metaviewport on/off we should maybe think about having a reftest environment variable for metaviewport support.
(In reply to Kartikaya Gupta (email:firstname.lastname@example.org) from comment #0) > We should just [only] run these > tests on B2G (i.e. on platforms where APZ and meta-viewport are both > enabled). In bug 1307332, :Aryx is performing the monumental task of removing all B2G/Mulet reftest.list annotations (basically making our harness assume B2G/Mulet are never defined). In part, that means all of the lines affected in this bug are removed -- which leaves all of the tests unused & worthy-of-being-deleted. Can/should we delete them (and all the reftest.list lines), or should we enable them under some alternate constellation of annotations?
(hoping kats [patch author] and/or tnikkel [patch-reviewer & test-author] will have some insight into comment 5.)
We should probably switch them from skip-if(!B2G) to skip-if(!Android). In theory the tests should run and pass fine on Fennec, although what happens in practice may be different.
What kats said.