What is causing the score reduction for position:sticky?

[Chris Harrelson]

https://wpt.fyi/compat2021?feature=position-sticky

Chrome seems to have bounced up and down, and all browsers went down at about the beginning of May.

+Stephen McGruer @Philip Jägenstedt how can we see which tests bounced their passing status?

We can see CLs that were in the latest dev release though, list here.

[Philip Jägenstedt]

I asked Stephen about this a while ago and his hypothesis was TablesNG getting enabled and disabled.

I don't see a way in the graphs to get a link to the wpt.fyi runs that the point was computed from, but clicking the first point after a change leads to precisely the link you shared. In there I see "[TableNG] Turn off for M91 branch point" which was hard to find because I was initially looking for "tables".

Using the same method on the previous increase in score leads to https://chromium.googlesource.com/chromium/src/+/ce54fa1dbfc1fe5256f431431c4a3bf712706625, so this was indeed the LayoutNGTable feature being turned on and off.

Best regards,

Philip

[Chris Harrelson]

Ok now i see the TableNG CLs. I was searching for "tables" and not "table".

Now I'd like to know about the change here (early May). That reduced the score in all browsers, did we add a test to the suite?

[Philip Jägenstedt]

Yeah, that must be a change in the test suite, because Chrome has been pinned to 91.0.4472.19 for almost 3 weeks. (This makes me realize getting it unpinned will be increasingly important...)

To figure this one out, I picked a run from May 1 and today in https://wpt.fyi/runs and compared them:

https://wpt.fyi/results/css/css-position/sticky?q=is%3Adifferent&run_id=5725812823162880&run_id=5645041693360128

It looks like one test has gone from passing to failing, and another failing test has been added. This would be enough to explain a visible difference with only 43 tests included, but neither test is actually in the list AFAICT... +Stephen McGruer can you tell what's going on here?

I've gone ahead and filed https://github.com/web-platform-tests/wpt.fyi/issues/2531 about making it a bit easier to figure stuff like this out.

[Chris Harrelson]

position-sticky-input-box-gets-focused-after-scroll.html: the test passes in the Chromium CQ (i.e., nothing marking it as failing in TestExpectations.

sticky-after-input.html: same.

+Robert Flack FYI

[Stephen McGruer]

> position-sticky-input-box-gets-focused-after-scroll.html: the test passes in the Chromium CQ (i.e., nothing marking it as failing in TestExpectations.

Looks like the test and code was changed in https://chromium-review.googlesource.com/c/chromium/src/+/2727895 . I assume that CL isn't in 91.0.4472.19 (which WPT is currently pinned to), and I assume the other two browsers fail the changed-to behavior as well?

> sticky-after-input.html: same.

Newly added in the same CL; I would assume the same root cause.

--

smcgruer • he / him