1840796 - WPT css/css-flexbox/alignment/flex-align-baseline-flex-003.html and flex-align-baseline-flex-004.html fail in Firefox

Status: RESOLVED FIXED

(Core :: Layout: Flexbox, defect)

Description

[Daniel Holbert [:dholbert]]

We fail various subtests in these two WPTs:
https://wpt.fyi/results/css/css-flexbox/alignment/flex-align-baseline-flex-003.html
https://wpt.fyi/results/css/css-flexbox/alignment/flex-align-baseline-flex-004.html

http://wpt.live/css/css-flexbox/alignment/flex-align-baseline-flex-003.html
http://wpt.live/css/css-flexbox/alignment/flex-align-baseline-flex-004.html

Looking at the first failure in -003 test: the test is doing e.g. first baseline alignment between a flex container and another element, and it's testing whether we export the correct baseline for that flex container.

In this case the test seems to be expecting that the flex container should derive its first baseline from an item in its first row, and that item happens to be last baseline aligned. (There are no first-baseline aligned items in that first row.) We instead fall back to just using the first flex item.

The relevant spec text is here:
https://drafts.csswg.org/css-flexbox-1/#flex-baselines

If any of the flex items on the flex container’s startmost/endmost flex line participate in baseline alignment, the flex container’s first/last main-axis baseline set is generated from the shared alignment baseline of those flex items.

This doesn't say anything about requiring the flex items to be participating in first vs. last baseline alignment; just whether they "participate in baseline alignment".

So we probably need to adjust the logic here:
https://searchfox.org/mozilla-central/rev/c0adc2160976e2c118e2e5709d08aac071fddce9/layout/generic/nsFlexContainerFrame.cpp#5257-5262
...to fall back from FirstBaselineOffset() to the line's LastBaselineOffset() before we finally give up and fall back to the sentinel value to trigger the first flex item.

Comment 1

[Ting-Yu Lin [:TYLin] (PST, UTC-8)]

Attachment: Bug 1840796 - Consider last baseline in the startmost line when computing flex container first baseline. r?dholbert

Likewise, we should also consider first baseline in the endmost line when
computing flex container last baseline.

Fixed the XXX comment by using logical coordinates. This is necessary to pass
flex-align-baseline-flex-004.html because it tests a vertical-rl flex container.

Comment 3

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/fb3b06cc9188
Consider last baseline in the startmost line when computing flex container first baseline. r=dholbert

Comment 4

[Cristian Tuns]

https://hg.mozilla.org/mozilla-central/rev/fb3b06cc9188