Side note that the logs there do show a bit of difference if you view them in reftest-analyzer, even though we're reporting `max difference: 0, number of differing pixels: 0`. This is because we apparently "round down" fuzz in some cases via this autofuzz clause: https://searchfox.org/mozilla-central/rev/35cc00a25c4471993fdaa5761952bd3afd4f1731/layout/tools/reftest/reftest.jsm#1128-1137 As shown in [one log](https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=269376112&repo=autoland&lineNumber=4352): > INFO - REFTEST TEST-LOAD | http://10.7.205.221:8854/tests/layout/reftests/border-image/svg-as-border-image-4b.html | 306 / 1928 (15%) > INFO - REFTEST INFO | REFTEST wr-on-android dropping fuzz of (1, 36) to (0, 0) > INFO - REFTEST INFO | REFTEST fuzzy test (16, 69) <= (0, 0) <= (16, 150) > INFO - REFTEST TEST-UNEXPECTED-PASS | http://10.7.205.221:8854/tests/layout/reftests/border-image/svg-as-border-image-4b.html == http://10.7.205.221:8854/tests/layout/reftests/border-image/svg-as-border-image-4-ref.html | image comparison, max difference: 0, number of differing pixels: 0 Anyway, even the "real" difference here `(1,36)` is below the fuzz threshold that we added in this bug. And since autofuzz bumps it down 0,0, we need to allow 0,0 in our ranges.
Bug 1584760 Comment 8 Edit History
Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.
Side note that the logs there do show a bit of difference if you view them in reftest-analyzer, even though we're reporting `max difference: 0, number of differing pixels: 0`. This is because we apparently "round down" fuzz in some cases via this autofuzz clause: https://searchfox.org/mozilla-central/rev/35cc00a25c4471993fdaa5761952bd3afd4f1731/layout/tools/reftest/reftest.jsm#1128-1137 As shown in [one log](https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=269376112&repo=autoland&lineNumber=4352): > INFO - REFTEST TEST-LOAD | http://10.7.205.221:8854/tests/layout/reftests/border-image/svg-as-border-image-4b.html | 306 / 1928 (15%) > INFO - REFTEST INFO | REFTEST wr-on-android dropping fuzz of (1, 36) to (0, 0) > INFO - REFTEST INFO | REFTEST fuzzy test (16, 69) <= (0, 0) <= (16, 150) > INFO - REFTEST TEST-UNEXPECTED-PASS | http://10.7.205.221:8854/tests/layout/reftests/border-image/svg-as-border-image-4b.html == http://10.7.205.221:8854/tests/layout/reftests/border-image/svg-as-border-image-4-ref.html | image comparison, max difference: 0, number of differing pixels: 0 Anyway, even the "real" difference here `(1,36)` is below the acceptable-fuzz-range that we added in this bug. And since autofuzz bumps it down 0,0, we need to allow 0,0 in our ranges.