1280786 - word-break: break-all cuts the ligature in the middle

Status: NEW

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

Description

[Santhosh Thottingal]

Attachment: Screenshot showing the ligature cut. The text in red color has its first ligature cut.

User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:47.0) Gecko/20100101 Firefox/47.0
Build ID: 20160608010004

Steps to reproduce:

For a paragraph with Malayalam text, I gave the following css:

text-align:justify;
word-break: break-all;

Testcase: http://codepen.io/santhoshtr/pen/YWGVpW


Actual results:

The break happens in the middle of ligature. One half of ligature went to the next line.


Expected results:

At any case, a letter should not be cut into two parts. It is like splitting "w" to "v" and "v".

Comment 1

[Santhosh Thottingal]

Attachment: Same content rendered by Chrome. No ligature cut happened

Comment 2

[Jonathan Kew [:jfkthame]]

Yes, this is clearly bad.

Splitting a ligature is not quite the same thing as splitting a letter, though. It would obviously be wrong to split "w" into "v" and "v", but on the other hand if there were an "fi" ligature, it would be correct to split this into "f" and "i" when line-breaking.

One problem we have is that when a line-break occurs within a ligature, we don't re-shape the separate fragments of the text. This is bug 479829. If we fix that, we'd no longer get the ugly broken ligature here; instead, we'd presumably get a separate PA+VIRAMA (പ്) before the line-break, and the remainder of the word (from the RA onwards, രാദേശിക) on the next line.

It's not clear to me whether that would be the best result here, or whether a break within the ligated PA + VIRAMA + RA + AA-MATRA cluster should never be allowed (even with break-all) -- and if so, how to distinguish between that and cases like "ffi" where splitting the ligature (and re-shaping) is the desired result.