[Bug 24608] New: a regression in 1.3.4 about bidi support

http://bugs.winehq.org/show_bug.cgi?id=24608

Summary: a regression in 1.3.4 about bidi support
           Product: Wine
           Version: 1.3.4
          Platform: x86
        OS/Version: Linux
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: usp10
        AssignedTo: [e-mail]
        ReportedBy: [e-mail]

Created an attachment (id=31085)
 --> (http://bugs.winehq.org/attachment.cgi?id=31085)
wine 1.2 supports BiDi Partialy

There is a regression in 1.3.4 regards BiDi supporting. I can read Arabic texts
in wine 1.2 but not in 1.3.3.

After I run a regression test I found the commit number
7bf4db877ae402de72a4f1e7bf0f4f27b5b4fa67 causes  the regression.

the regression test result :

commit 7bf4db877ae402de72a4f1e7bf0f4f27b5b4fa67
Author: Aric Stewart 
Date:   Thu Sep 9 15:27:57 2010 -0500

usp10: Check font for required features and report error if missing.

:040000 040000 f13d84cd768f0757d04c9f750d4c97631aa26640
2c4447a6efdc88d9a5d43f83e6d4dce0fd4aaa08 M      dlls

[Bug 24608] a regression in 1.3.4 about bidi support

http://bugs.winehq.org/show_bug.cgi?id=24608

Fahad Al-Saidi  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

[Bug 24608] a regression in 1.3.4 about bidi support

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #1 from Fahad Al-Saidi  2010-10-02 23:37:44 CDT ---
Created an attachment (id=31086)
 --> (http://bugs.winehq.org/attachment.cgi?id=31086)
there is no BiDi supporting in wine 1.3.4

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

Austin English  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |download, regression
            Summary|a regression in 1.3.4 about |BiDi support broken for
                   |bidi support                |Arabic text

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

Paul Vriens  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #2 from Aric Stewart  2010-10-03 13:41:45 CDT ---
(In reply to comment #0)
> Created an attachment (id=31085)

--> (http://bugs.winehq.org/attachment.cgi?id=31085) [details]
> wine 1.2 supports BiDi Partialy
>
> There is a regression in 1.3.4 regards BiDi supporting. I can read Arabic texts
> in wine 1.2 but not in 1.3.3.
>
> After I run a regression test I found the commit number
> 7bf4db877ae402de72a4f1e7bf0f4f27b5b4fa67 causes  the regression.
>
> the regression test result :
>
> commit 7bf4db877ae402de72a4f1e7bf0f4f27b5b4fa67
> Author: Aric Stewart 
> Date:   Thu Sep 9 15:27:57 2010 -0500
> usp10: Check font for required features and report error if missing.
>
> :040000 040000 f13d84cd768f0757d04c9f750d4c97631aa26640
> 2c4447a6efdc88d9a5d43f83e6d4dce0fd4aaa08 M      dlls

Where are you seeing this issue? And what font is being used?  That commit
disable arabic rendering for fonts that do not support the minimum arabic GSUB
features (init, medi, fina etc)  As a result if you where using a font that had
arabic characters but no arabic features it would stop shaping.

Also you could post a +uniscribe log that would help confirm what feature is
missing.

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #3 from Fahad Al-Saidi  2010-10-03 23:50:54 CDT ---
I am using MS web fonts: Tahoma, Arial, and arabeyes fonts. They support Arabic
very well and All have the problem. I notice also that DejaVu Sans doesn't have
the problem.

I used to see the +uniscribe log but now disappears. Is there a way to show it?

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #4 from Andrew Nguyen  2010-10-04 00:08:58 CDT ---
(In reply to comment #3)
> I am using MS web fonts: Tahoma, Arial, and arabeyes fonts. They support Arabic
> very well and All have the problem. I notice also that DejaVu Sans doesn't have
> the problem.
>
> I used to see the +uniscribe log but now disappears. Is there a way to show it?

http://wiki.winehq.org/FAQ#head-16da35b6327024d6ea576e3678488b16862d0f5e with
WINEDEBUG=+uniscribe

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #5 from Fahad Al-Saidi  2010-10-04 00:28:08 CDT ---
Created an attachment (id=31103)
 --> (http://bugs.winehq.org/attachment.cgi?id=31103)
+uniscribe log

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #6 from Aric Stewart  2010-10-04 08:12:08 CDT ---
(In reply to comment #5)
> Created an attachment (id=31103)

--> (http://bugs.winehq.org/attachment.cgi?id=31103) [details]
> +uniscribe log

Interesting the font in question doe not appear to implement 'rlig' which
according to the documentation here
(http://www.microsoft.com/typography/otfntdev/arabicot/features.aspx) is
required for uniscribe to shape the font.

If you use a native usp10.dll from windows XP or later does it work?

I have been using the Arabic Typesetting font for testing myself.

It may be the case that we need to remove the 'rlig' requirement.

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #7 from Fahad Al-Saidi  2010-10-04 23:20:07 CDT ---
(In reply to comment #6)
> (In reply to comment #5)
> > Created an attachment (id=31103)

--> (http://bugs.winehq.org/attachment.cgi?id=31103) [details] [details]
> > +uniscribe log
>
> Interesting the font in question doe not appear to implement 'rlig' which
> according to the documentation here
> (http://www.microsoft.com/typography/otfntdev/arabicot/features.aspx) is
> required for uniscribe to shape the font.
>
> If you use a native usp10.dll from windows XP or later does it work?
>
> I have been using the Arabic Typesetting font for testing myself.
>
> It may be the case that we need to remove the 'rlig' requirement.

I copied usp10.dll and added as native from winecfg and it works fine with some
warnings like this:
fixme:font:WineEngRealizationInfo (0x167a58, 0x32eaec): stub!

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

Khaled Hosny  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[e-mail]

--- Comment #8 from Khaled Hosny  2010-10-05 03:02:56 CDT ---
I don't think hard coding a set of features is a wise idea, existing features
should be applied regardless of some others are missing or not. Most Arabic
fonts doing ligature I've seen use liga not rlig or may be both. I've been
building fonts that use neither and use calt instead. Even init, medi and fina,
one can have a perfectly usable Arabic font with only init and medi for example
(at least in theory). The point is that there are countless number of
possibilities, so there is no point to reject existing features in the font
because it is missing others, and if it was a broken font then the result is no
better than it was.

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #9 from Aric Stewart  2010-10-05 06:55:29 CDT ---
(In reply to comment #8)
> I don't think hard coding a set of features is a wise idea, existing features
> should be applied regardless of some others are missing or not. Most Arabic
> fonts doing ligature I've seen use liga not rlig or may be both. I've been
> building fonts that use neither and use calt instead. Even init, medi and fina,
> one can have a perfectly usable Arabic font with only init and medi for example
> (at least in theory). The point is that there are countless number of
> possibilities, so there is no point to reject existing features in the font
> because it is missing others, and if it was a broken font then the result is no
> better than it was.

That is very interesting. I was working off of the Microsoft typography website
where it states font requirements. I also know that native usp10.dll does in
fact reject some TTF fonts even if they do have the glyphs.

If you make fonts is there a chance you could create me a set of dummy fonts
that have some of the features but not all?  That way I can do some empirical
testing as to what uniscribe actually requires or not?

thanks!
-aric

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #10 from Khaled Hosny  2010-10-05 08:39:56 CDT ---
I don't know about Uniscribe behavior here, but FontConfig, for example,
detects language support based on certain character for each language (check
fc-lang/*.orth files in FC source tree), it is even smart enough to not count
empty glyphs for characters that should not be empty, so may be Uniscribe is
doing similar tricks.

It is easy to open a font in FontForge and remove some features, generate and
test. But if needed I can attach a set of fonts with various features
added/removed.

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #11 from Aric Stewart  2010-10-05 08:49:27 CDT ---
(In reply to comment #10)
> I don't know about Uniscribe behavior here, but FontConfig, for example,
> detects language support based on certain character for each language (check
> fc-lang/*.orth files in FC source tree), it is even smart enough to not count
> empty glyphs for characters that should not be empty, so may be Uniscribe is
> doing similar tricks.
>
> It is easy to open a font in FontForge and remove some features, generate and
> test. But if needed I can attach a set of fonts with various features
> added/removed.

Hi,

Yeah, I know uniscribe is doing more than just checking for characters because
I have some fonts with arabic characters but none of the features and windows
uniscribe rejects those fonts as not to shape.

I am working on a mac and fontforge here works very poorly when editing
features so if you can generate the set of fonts that would greatly speed up my
ability to test.

thanks!
-aric

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #12 from Khaled Hosny  2010-10-06 11:01:23 CDT ---
Created an attachment (id=31154)
 --> (http://bugs.winehq.org/attachment.cgi?id=31154)
Fonts for testing

Here is a set of fonts, each font name is suffixed by the set of features it
has; (I)nit, (M)edi, (F)ina, (R)lig and (N)one.

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #13 from Aric Stewart  2010-10-06 12:41:53 CDT ---
(In reply to comment #12)
> Created an attachment (id=31154)

--> (http://bugs.winehq.org/attachment.cgi?id=31154) [details]
> Fonts for testing
>
> Here is a set of fonts, each font name is suffixed by the set of features it
> has; (I)nit, (M)edi, (F)ina, (R)lig and (N)one.

Thanks,

I see that it will not shape with the "N" font.  But it shaped with all the
others.

Just out of curiosity can i get fonts with just M F and R?  Just so i can see
if it is the Init feature that is required or if it just wants one of the
features?

thanks, this is very helpful.
-aric

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #14 from Khaled Hosny  2010-10-06 13:17:10 CDT ---
Created an attachment (id=31156)
 --> (http://bugs.winehq.org/attachment.cgi?id=31156)
Font for testing

Here is a font with only MFR.

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #15 from Aric Stewart  2010-10-06 13:44:26 CDT ---
ok patch submitted to address this issue.

-aric

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #16 from Nikolay Sivov  2010-10-07 16:07:26 CDT ---
Fixed by 0c5d0f80bc739bfd14415f9d64570db06da0455b I suppose.

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

Nikolay Sivov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Blocks|                            |24655

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

--- Comment #17 from Fahad Al-Saidi  2010-10-10 00:10:59 CDT ---
(In reply to comment #16)
> Fixed by 0c5d0f80bc739bfd14415f9d64570db06da0455b I suppose.

I can confirm that is fixed now. Thanks

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

Dmitry Timoshkov  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |FIXED

--- Comment #18 from Dmitry Timoshkov  2010-10-10 00:15:14 CDT ---
Reported fixed.

[Bug 24608] BiDi support broken for Arabic text

http://bugs.winehq.org/show_bug.cgi?id=24608

Alexandre Julliard  changed:

What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|RESOLVED                    |CLOSED

--- Comment #19 from Alexandre Julliard  2010-10-15 12:50:35 CDT ---
Closing bugs fixed in 1.3.5.