1591496 - word-break-break-all-inline-006 fails in Firefox 72

Status: NEW

(Core :: Layout: Text and Fonts, defect, P3)

Description

[Javier Fernandez]

Attachment: word-break-break-all-issue-2.html

User Agent: Mozilla/5.0 (X11; Ubuntu; Linux x86_64; rv:69.0) Gecko/20100101 Firefox/69.0

Steps to reproduce:

1- Load the attached test case

Actual results:

The first box breaks the second line after the last "b" character, moving the entire "ccccc" word to a new line.

The second box, instead, let the second line to overflow.

Expected results:

The first box must be rendered as the second box.

Comment 1

[Javier Fernandez]

The word-break-break-all-inline-006.html test was added by Mozilla developers in May this year, as it's shown [1] in the WPT repository, so I guess the tests passed at that time.

It seems that the test is failing now in Firefox 72, according to wpt.fyi [2]; I also verified locally using Firefox nightly.

[1] https://github.com/web-platform-tests/wpt/commit/6748dd2f8182ab3b0b199cb3c2c31744d6ee7d59
[2] https://wpt.fyi/results/css/css-text/word-break/word-break-break-all-inline-006.html?label=experimental&label=master&aligned

Comment 2

[Javier Fernandez]

BTW, it seems that the test is failing in Chrome as well; I'm working on it actually. So I'd like to verify whether this is a regression in Firefox or that the test is not correct now, according to the last version of the CSS Text 3 specification.

Comment 3

[Emilio Cobos Álvarez [:emilio]]

The test landed as expected-failure. My understanding is that safari was right here...

I think Jonathan may have done more spec digging. I'm not sure whether doing that at a boundary when break-all ends is correct.

Comment 4

[Javier Fernandez]

The test passes using Chrome with Legacy layout, so I guess the expected result is correct. I understand from your comments that it has been always failing for Firefox then.

As far as I understand, break-all applies only to the span element with the 'bbbbbbbbbbb' word, so that would explain the expected behavior. What's funny is that both Firefox and Chrome (LayoutNG) both renders the test in the same way, but wrongly as you are confirming now.

Comment 5

[Javier Fernandez]

In any case, if this is an expected fail, shouldn't be bug report in Firefox explaining that such behavior is not correct ?

Comment 6

[Javier Fernandez]

Umm, I guess the question here is whether the word-break property should be applied to the new inline child created when the 'bbbbbbbbbbb' word is broken.

Comment 7

[Javier Fernandez]

I filed an issue for the CSS WG to clarify the actual expected behavior of this test.

https://github.com/web-platform-tests/wpt/issues/19903

Comment 8

[Javier Fernandez]

JFTR, I filed a similar bug [1] for Chrome, which also fails in that tests but with a slightly different behavior.

[1] https://bugs.chromium.org/p/chromium/issues/detail?id=1018423