[HarfBuzz] Question regarding hb-view categorization of runs of text

[Khaled Hosny]

On Sat, Jan 18, 2014 at 04:51:51PM -0500, Ed Trager wrote:
> Hi, everyone,
> 
> If I run hb-view (v. 0.9.25) to compose a line of text that consists solely
> of Tai Tham, I get expected shaping results. However, if I add some Latin
> text at the beginning of a line of Tai Tham text, then I get incorrect
> shaping: specifically, Tai Tham prefixed vowels U+1A6E through U+1A72 and
> consonant sign U+1A55 MEDIAL RA are not re-ordered.

The script is taken from the first character with non-common
non-inherited script property, try setting it explicitly and see if it
fixes this.
 
> (sample image attached)
> 
> I'm wondering if this is just an artifact of hb-view?:  Does hb-view always
> take a single line of text as being a single run of text, based on the
> first non-neutral character encountered?
> 
> To me it looks like this is what happens.  That is to say, it looks like
> hb-view does not break single lines of text into separate runs of text even
> when that is required.

Yes, each line is shaped in one buffer with the same script, language
and direction.

> Since hb-view is a tool for testing, this may be acceptable behavior.
> I would just like to know if this is in fact the case, or whether I am
> seeing some sort of bug?

Any serious HarfBuzz client should be doing some sort of itemization, so
this is not an issue.

Regards,
Khaled