Bug 70029 – Old Turkic characters behave as left-to-right instead of right-to-left, because they are encoded as surrogate pairs.

• Status: RESOLVED
• Resolution: FIXED
• Component: Text

Hello,

In Webkit browsers, including Chrome, Old Turkic text does not render correctly. It should be rendered from Right-to-Left, but we see it as LTR as usual. 

Old Turkic block was introduced with Unicode 5.2, and Gecko based browsers render it correctly. 

Sample texts:

http://fankibiber.wordpress.com

You can download the required fonts from here: http://bitigchi.tumblr.com/downloads

You can see the difference in rendering when tried with Firefox and Safari.

Thank you.

Created attachment 110854 [details]
Safari: Not rendered correctly

Created attachment 110856 [details]
Firefox: Rendered correctly

Although the font is different here, it is clear that the letters of the Safari screenshot here are right aligned on the first line.

[Radar WebKit Bug Importer]

<rdar://problem/10281541>

InlineIterator operates on UTF-16 code units rather than on characters. Since Old Turkic is encoded using surrogate pairs, InlineIterator::direction() incorrectly returns LeftToRight (for each of the two surrogates) instead of RightToLeft.

This is a very annoying bug. Is there any info on when this can be fixed?

!Ping!

[Darin Adler]

Created attachment 238440 [details]
Patch

[Darin Adler]

Mitz asked me to look at this bug, so I did. Patch isn’t ready to land because we need some regression tests; we need to write tests that don’t require installing a font.

[Darin Adler]

Also need to measure performance to see if I slowed things down.

[Darin Adler]

Comment on attachment 238440 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=238440&action=review

> Source/WebCore/rendering/InlineIterator.h:436
> +        if (LIKELY(U16_IS_SINGLE(codeUnit)))
> +            return u_charDirection(codeUnit);

If this code path is really hot, then we should have a faster patch so we don't have to call into u_charDirection for common characters.

(In reply to comment #8)
> Mitz asked me to look at this bug, so I did. Patch isn’t ready to land because we need some regression tests; we need to write tests that don’t require installing a font.

Great news anyway! Thank you very much!

[Darin Adler]

Created attachment 238599 [details]
Patch

[Darin Adler]

Committed r173943: <http://trac.webkit.org/changeset/173943>

Just for info, has this landed on a Safari release yet? I don't have an OS X machine at the moment, just wanted to check. 

Thanks in advance.