79284 - Thai shaping not handled by gfx/src/nsRenderingContext

Status: RESOLVED FIXED

(Core :: Internationalization, defect)

Description

[Prabhat Hegde]

On Unix platforms, scripts which require generation of presentation
forms (context sensitive shaping) before being rendered are not
handled by mozilla, in this case for thai.

For example, rendering of URL http://www.thai.net using any of the
fonts specified for TIS620 or ISO8859-11 in nsFontMetricsGTK.cpp
is in-correct.

This needs to be corrected in order to present thai content on
*nix platforms. The output is correct on Windows platform since
Drawing and Measurement API in windows automatically handle shaping.

Comment 1

[Andreas Becker]

Changing QA contact to [e-mail] for now.

Comment 2

[nhottanscp]

Reassign to bstell.

Comment 3

[Prabhat Hegde]

Attachment: Patch for review with library and changed sources

Comment 4

[Prabhat Hegde]

hi brian,

I've attached a patch that fixes this problem. Patch is strictly for review
purposes (as some of the makefiles are for Solaris Make & not gnumake)

The approach is as below:

Create a library based on pango but eliminating the PangoFont and glib1.3
related dependencies:

A> Use API and implementation of shaper modules from pango.
B> Use API for layout from Pango.

Modifiy SingleByteConvert, currently used for thai FontCharsets to handle
thai shaping.

Note that this architecture allows us to transparently handle any complex
script that requires shaping with minor code changes in mozilla.

For ex, to add support for indic scripts:

A> Write a indic shaper.
B> Run tool to update pango.modules.
C> Add/Update SingleByteConvert, DoubleByteConvert implementation for the
   target script.

As you may expect, this does not solve the CTL problem. Code changes for Cursor
positioning and Selection are separate. 

I also have a build based on the above that works with all popular thai
presentation forms.

prabhat.

Comment 5

[Prabhat Hegde]

Attachment: Newer version of a pango based shaping/layout library

Comment 6

[Prabhat Hegde]

Attachment: Updated Object for CTL Interfaces to new pango library

Comment 7

[Prabhat Hegde]

Attachment: Diff w.r.t 05/11 nightly nsFontMetricsGTK.cpp

Comment 8

[Frank Tang]

Prabhat- can we wrap the majority of the TISConvert call into a 
nsIUnicodeEncoder interface so we will have similar call struct as the rest of 
the code in GTK GFX? I don't think we need a new XPCOM interface for the purpose 
of GTK GFX changes, right? We can add new interface later if we need to deal 
with layout issues. Using the existing interface will reduce a lot of issues. 

Comment 9

[Frank Tang]

please do not attach attachement in mime type application/octet-stream
What are they? zip ? tar? gz?

Comment 10

[Prabhat Hegde]

frank, Yes, TISConvert interface could be mapped onto nsIUnicodeEncoder;
As you say for GFX/GTK purposes only. However an XPCOM interface will be
needed in the future for layout additions.

Sorry for the mime attachment type (its tar.Z this time). 

prabhat

Comment 11

[Prabhat Hegde]

Added 84394 to track this bug. Brian, maybe you can choose which
bugs to close as duplicate and which ones to leave open - thanks.

QA team, if you would like to add any keywords to track thai ctl bugs,
please let me know so that i can add them as keywords when i update 
any thai related bugs - thanks.

Comment 12

[kill this account]

--> ftang

Comment 13

[Frank Tang]

bulk move NEW FUTURE bug to ASSIGN

Comment 14

[Roland Mainz]

prabhat:
Can we close this bug or mark this as duplicate ?

Comment 15

[Prabhat Hegde]

hi roland,

This bug can be marked closed.

Comment 16

[Roland Mainz]

Stealing...

Comment 17

[Roland Mainz]

Closing per comment 15, this bug has been fixed as part of bug 84380.
Marking bug as FIXED.

prabhat/katakai, can you verify, please ?