text-decoration-skip-ink causes minor visual differences in decoration lines that cause some tests to fail
Categories
(Core :: Layout, defect, P3)
Tracking
()
People
(Reporter: cmarlow, Unassigned)
Details
Attachments
(1 file)
|
211.55 KB,
text/x-log
|
Details |
On some platforms, text-decoration-skip-ink causes some tests to fail because it can cause an extra half pixel of an underline to be drawn, or cause darker shades on one pixel of an underline due to compositing. This may be due to a rounding error ( potentially rounding done when intercepts are calculated). This affects the OSX and Windows test cases here: https://treeherder.mozilla.org/#/jobs?repo=try&resultStatus=testfailed%2Cbusted%2Cexception&classifiedState=unclassified&revision=fb0cfc873b9f6c0c766c2a1443b3f8231a6e252c&selectedJob=261258199
Updating to text-decoration-skip-ink: none for all underlined elements temporarily resolves the issue.
Comment 1•6 years ago
•
|
||
FWIW, this bug is the reason that reftest decoration-color-standards.html has ink-skipping disabled (set to none via CSS) right now. Here's the reftest log output from a try run where I attempted to reenable it -- it failed on Android with a fuzzy pixel at the far right edge of one of the underlines.
This was for:
Job: android 4.3 api16+ opt reftests without e10s test-android-em-4.3-arm7-api-16/opt-fennec-reftest-1proc-26 r-1proc(r26)
Job name: test-android-em-4.3-arm7-api-16/opt-fennec-reftest-1proc-26
https://treeherder.mozilla.org/logviewer.html#/jobs?job_id=263602399&repo=try&lineNumber=3313
(That ^ log will expire within a few days, as Try logs tend to do; hence, attaching the relevant reftest log output for posterity.)
Updated•3 years ago
|
Description
•