[HarfBuzz] shaping of U+06C2 [uniscribe bug]

Andrew Glass (WINDOWS) Andrew.Glass at microsoft.com
Mon Aug 19 10:25:23 PDT 2013


Thanks for the report Jonathan, I've filed a bug in our system.

Andrew

-----Original Message-----
From: Jonathan Kew [mailto:jfkthame at googlemail.com] 
Sent: lundi 19 août 2013 03:00
To: Behdad Esfahbod; Harfbuzz; Andrew Glass (WINDOWS)
Subject: shaping of U+06C2 [uniscribe bug]

A further harfbuzz-vs-uniscribe discrepancy that I'm seeing is the shaping of the Arabic-script character U+06C2. Although (AFAIK) this character is normally used only in final position, it is classified by the Unicode standard as dual-joining (see ArabicShaping.txt), and therefore causes any following letter to take a right-linking (final or
medial) form.

In older versions of Unicode, U+06C2 was classified as right-joining, and so did not affect the form of a following letter. It looks like this change (for consistency with the behavior of its canonical decomposition
<06C1 0654>) was made in Unicode 4.1.

Unfortunately, uniscribe still (eight years later, even on Win8) seems to be treating U+06C2 as right-joining (and fonts shipped with Windows support it as such, and lack initial/medial forms).

This means that users are liable to omit the space or non-joiner that should be included in a phrase such as <U+062F,U+0631,U+062C,U+06C2,U+062D,U+0631,U+0627,U+0631,U+062A>, as uniscribe interrupts the joining after 06C2 even when a letter follows. 
When such text is rendered by a Unicode 4.1-or-later system, unwanted cursive-joining will occur.




More information about the HarfBuzz mailing list