These tests are failing with WebRender with what seems to be an off-by-one/rounding problem. ./layout/reftests/writing-mode/reftest.list: == 1193519-sideways-lr-3.html 1193519-sideways-lr-3-ref.html == 1193519-sideways-lr-4.html 1193519-sideways-lr-4-ref.html ./layout/reftests/svg/smil/style/reftest.list: == anim-css-strokewidth-1-from-by-pct-pct.svg anim-css-strokewidth-1-ref.svg == anim-css-strokewidth-1-from-to-pct-pct.svg anim-css-strokewidth-1-ref.svg == anim-css-strokewidth-1-to-pct-pct.svg anim-css-strokewidth-1-ref.svg ./layout/reftests/bugs/reftest.list: == 1291528.html 1291528-ref.html ./layout/reftests/svg/svg-integration/clip-path/reftest.list: == clip-path-polygon-013.html clip-path-stripes-003-ref.html https://treeherder.mozilla.org/#/jobs?repo=try&bugfiler=&group_state=expanded&revision=4aee7c413aa978c5cd41f6065c654d55025b70b0

There are also a couple form-related reftests that are fuzzy in a similar way. I'll look into these as I recently looked at bug 1444185 and such. A new one from bug 1495153: ./layout/reftests/forms/input/file/reftest.list: == dynamic-max-width.html dynamic-max-width-ref.html Which actually used to pass before my work in bug 1444185, and fails across the board. So I still think that the way we do snapping of fallback is not quite right.

(In reply to Emilio Cobos Álvarez (:emilio) from comment #1) > A new one from bug 1495153: > > ./layout/reftests/forms/input/file/reftest.list: > == dynamic-max-width.html dynamic-max-width-ref.html FYI, in bug 1500530 I'm making the anonymous <label> have display:inline-block (instead of the default display:inline). This made it pass on WebRender(OSX) but not on other platforms. So I'm tagging it random-if(webrender) for now.

Could you tag it as fails-if(webrender&&!cocoaWidget) instead?

Sure, no problem. I repeated the test a few more times and seems to be a stable pass.