Remove w3c-css/received/css-writing-modes reftests
Categories
(Core :: Layout, task)
Tracking
()
Tracking | Status | |
---|---|---|
firefox75 | --- | fixed |
People
(Reporter: TYLin, Assigned: TYLin)
References
Details
Attachments
(4 files)
Before removing, we need to investigate the pass/fail state is the same as those in the web-platform-tests, or figuring out why they are different.
Bug 1594937 comment 2 is a sample analysis.
Assignee | ||
Comment 1•5 years ago
|
||
Assignee | ||
Comment 2•5 years ago
|
||
See bug 1598754 comment 1 for an analysis and observation after aligning
the pass/fail state with wpt.
This patch is generated by
- Migrate the bug numbers from failures.list to corresponding wpt int files.
- removing
os.path.join("css-writing-mode"),
in import-tests.py. - removing the relevent annotations of css-writing-mode in failures.list.
- running import-tests.py on a wpt repository at commit
15f199c91a72b0d51bf0a12b3b77827ecb5051ff.
Assignee | ||
Comment 3•5 years ago
|
||
The last css suite (css-writing-mode) is removed in Part 1. We can
remove the infrastructure.
Depends on D62984
Updated•5 years ago
|
- Non-webrender OSX (e.g. OS X 10.14 debug) has a lot of failures
because of removing fuzzy annotations. Those fuzzy annotations are
really necessary in our own reftest framework, but not on wpt. I think
they're just noise.
I at least noticed some when I was looking at the bug number transfers where the reftest harness version had a fuzzy annotation for non-webrender OSX specifically, and the WPT was annotated as fail for all mac. That seems less than optimal. Could the wpt annotation be reduced to being a fuzzy and/or could it be limited to webrender? For example, outline-inline-vrl-006.html.ini, or inline-block-alignment-006.xht.ini .
I didn't search for this exhaustively, though, so I'm curious how this category of changes was deemed ok by the analysis you did above?
Assignee | ||
Comment 5•5 years ago
|
||
Could the wpt annotation be reduced to being a fuzzy and/or could it be limited to webrender?
You're right. I overlook this. It is better to preserve the fuzzy annotation. Also, I just realize we didn't run wpt reftests on macOS with webrender enabled, so it is important to annotate the non-webrender macOS runs to pass with fuzzy.
Here is a try that changes those wpt tests that fail only on macOS to PASS. All of them are fuzzy. So I'll use this as a guidance to update our wpt ini.
https://treeherder.mozilla.org/#/jobs?repo=try&revision=05afc9c90112d5423328fe924443c91750054252
Since Part 1 is too large to display properly in phabricator UI, making the review and writing inline comments difficult. I'll consolidate all the wpt ini modification to Part 0.
Assignee | ||
Comment 6•5 years ago
|
||
In this patch, we migrate the fuzzy annotations and bug numbers from
layout/reftests/w3c-css/failures.list to corresponding wpt ini files.
That is, some of the failures on macOS in wpt ini files are actually
successes with some fuzzy.
Updated•5 years ago
|
Comment 8•5 years ago
|
||
bugherder |
https://hg.mozilla.org/mozilla-central/rev/33e4b5e3f51b
https://hg.mozilla.org/mozilla-central/rev/b066d8667720
https://hg.mozilla.org/mozilla-central/rev/d9a3e282d399
Description
•