1945341 - Count characters more correctly when applying the CSS hyphenate-limit-chars length thresholds

Status: RESOLVED FIXED

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

Description

[Jonathan Kew [:jfkthame]]

Re-reading the code that implements hyphenate-limit-chars while fixing bug 1945227, I noticed an error in how "characters" are counted for the purpose of evaluating the word-length limits. The spec requires non-spacing marks and word-internal punctuation to be ignored, which we mostly do -- but at the end of the word, we unconditionally include the last character in the count without checking its category.

This leads to incorrect application of the word-length limit in cases like the Hindi word "मामले" in our auto-hyphenation-hi-1.html testcase, which should count as 4 characters long (म-ा-म-ले), and therefore with the default hyphenate-limit-chars setting that requires a word to be 5 characters long, no hyphenation should be applied. But we incorrectly count the last character, which is the vowel matra े (which combines with the consonant ल) as a separate unit, and so we improperly allow hyphenation.

A couple of the other Indic scripts have similar examples where we should be blocking hyphenation, but the code currently miscounts length.

Comment 1

[Jonathan Kew [:jfkthame]]

Attachment: Bug 1945341 - Count characters more correctly for hyphenate-limit-chars thresholds. r=#layout

In particular, this avoids counting a combining mark at the end of the word
as if it were a separate character.

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/ef9315641dfc
Count characters more correctly for hyphenate-limit-chars thresholds. r=layout-reviewers,emilio

Comment 3

[Cristian Tuns]

https://hg.mozilla.org/mozilla-central/rev/ef9315641dfc