827093 - [harfbuzz] bug rendering Cham text

Status: RESOLVED FIXED

(Core :: Graphics: Text, defect)

Description

[Gen Kanai [:gen]]

Danh Hong has identified a bug in harfbuzz that is creating problems in rendering of Cham language text.

http://www.khmertype.org/2013/01/cham-rendering-in-firefox.html

Is this something we can fix in Gecko or do we need to send this bug upstream to harfbuzz?

Comment 1

[Jonathan Kew [:jfkthame]]

It'll need to be addressed upstream. However, specific testcases (character sequences, and suggested OpenType fonts for testing) would be helpful; offhand, I'm not sure what font(s) we should be testing with. The khmertype.org page mentioned doesn't give any details...

Comment 2

[Behdad Esfahbod]

the sequence is aa40,aa34, but I can't find any font...

Comment 3

[Behdad Esfahbod]

Anyway, it's a pre-base reordering Ra.  It's a consonant medial, so there is no halant/coeng before it, so we don't apply the pre-base-reordering logic.  It's similar to the Myanmar cases we were checking in Vancouver.

Anyway, best way is to make the pref feature apply to the Ra alone, and I can change HarfBuzz to mark handle medial pre-base-reordering characters.

But first, need a font.

Comment 4

[Gen Kanai [:gen]]

I'll get a font for us to test with.

Comment 5

[Gen Kanai [:gen]]

Attachment: Cham truetype font for testing

Comment 6

[Gen Kanai [:gen]]

Danh has provided a .ttf font for testing. 

Also available here:

http://people.mozilla.org/~gen/files/testing/ChamPleiku.ttf

Comment 7

[Behdad Esfahbod]

I left a message on his blog a few days ago but it doesn't show up and I didn't hear from him.  Please tell him that the correct place to report bugs is the harfbuzz mailing list or bugzilla, not his blog...

Comment 8

[Gen Kanai [:gen]]

(In reply to Behdad Esfahbod from comment #7)
> Please tell him that the correct place to report bugs
> is the harfbuzz mailing list or bugzilla, not his blog...

Will do.

Comment 9

[Jonathan Kew [:jfkthame]]

I've filed bug 841313 about taking a harfbuzz update that should fix this issue.

Comment 10

[Jonathan Kew [:jfkthame]]

This should appear in Nightly builds tomorrow. If you can verify that the new behavior is correct for Cham once that build is available, that would be great.

Comment 11

[Mathieu Pellerin]

Jonathan, an additional issue was identify with Cham script and subsequently fixed by Behdad (see http://cgit.freedesktop.org/harfbuzz/commit/?id=b1f440759161cb5a54e1b1b910f3a8383b8e8472 for the fix).

It'd be nice for the above commit to be integrated into mozilla's harfbuzz prior to m-c moving to Aurora. That'd complete Cham support for Firefox v21.

Comment 12

[Jonathan Kew [:jfkthame]]

Attachment: cherry-pick Cham shaping fix from harfbuzz commit b1f4407..b8e8472.

To fix the remaining Cham shaping issue, we should cherry-pick this from upstream harfbuzz. I don't want to take a full HB update right now, as there has been extensive work on default-invisibles that potentially affects all scripts, so I'd like that to have some more testing/stabilization time, but this is a simple fix specific to the SEAsian shaper, used only for a few lesser-known scripts, and cannot affect anything else. As such, it would be good to include it with the SEAsian shaper already landed for FF21 in bug 841313.

Comment 13

[Jonathan Kew [:jfkthame]]

https://hg.mozilla.org/integration/mozilla-inbound/rev/323f90679193

Comment 14

[Gregory Szorc [:gps]]

https://hg.mozilla.org/mozilla-central/rev/323f90679193