Zawgyi-One rendering broken

[behdad]

Eg. font from http://unobusa.org/2009/11/12/zawgyi-font-and-firefox-browser/
Text sequence: "\u1000\u103A\u1004\u1037\u1039\u1041" for zawgyi

The problem is, Zawgyi uses some gen_cat=Mn chars as non-spacing. The font does NOT have a GPOS table, so we invoke fallback mark positioning, which is not desired.

[behdad]

I wonder if we can detect this font (by size of it's GDEF etc) and disable fallback positioning...

[behdad]

@jfkthame any ideas?

[nona-google]

Just FYI, Android will use private script code for Zawgyi and we are passing language code to HarfBuzz.
You may be able to use that hint for detecting Zawgyi font.

[behdad]

Just FYI, Android will use private script code for Zawgyi and we are passing language code to HarfBuzz.
You may be able to use that hint for detecting Zawgyi font.

That actually helps. What script and language codes will you be passing? Script is enough for us to disable all kinds of processing.

[nona-google]

The new script code for Zawgyi will be "Qaag".
Currently this is in draft of UTS35.
http://unicode.org/repos/cldr/trunk/specs/ldml/tr35.html#unicode_script_subtag_validity

Once it is released, that should be written in here
http://www.unicode.org/reports/tr35/tr35.html#unicode_script_subtag_validity

[behdad]

Thanks Nona. I'm not sure whether to expose this hack script in the public API, or just silently wire it up internally. I'm leaning towards the latter. @jfkthame any preference?

[jfkthame]

Not sure what you mean by "wire it up internally". How will harfbuzz know when to apply the Zawgyi behavior in place of Myanmar behavior, unless the application passes in the new script code via the API?

[behdad]

I just meant whether to expose HB_SCRIPT_MYANMAR_ZAWGYI. Application can still create a hb_script_t tag without me exposing it. Indeed, for example, using hb-icu, or other bridges, it will go in automagically.

[behdad]

I'm worried that opening that can of worms will give legitimacy to encoding other pseudo- or non-Unicode- scripts.

[jfkthame]

Ah, OK. Yes, I don't think we need an HB_SCRIPT_ constant for this.

[behdad]

@nona-google Do you know what the language tag would be?

[nona-google]

Not yet decided in Android. Either en-Qaag-MM or my-Qaag-MM could be sent to HarfBuzz. The thing we can tell now is "Qaag" should be there if the text should be rendered in Zawgyi encoding.

[behdad]

Okay. As it happens, we don't care about the language tag. This should be fixed as far as I'm concerned.

[andjc]

@nona-google would use of Qaag for Zawgyi mean we have to add script tags for each of the major psuedo-Unicode myanmar script fonts out there? Or would my-Qaag specifically imply Zawgyi and other language tags would imply other pseudo Unicode fonts for Myanmar script?