438571 - Arabic digits are not enclosed in ARABIC NUMBER SIGN, ARABIC END OF AYAH, etc.

Status: RESOLVED FIXED

(Core :: Internationalization, defect)

Description

[arifkarim]

User-Agent:       Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9) Gecko/2008052906 Firefox/3.0
Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9) Gecko/2008052906 Firefox/3.0

I have worked on an indo-pak arabic quranic font, that works fine in internet explorer, but when i open the web page in Firefox, the numbers of quranic digits stay outside the circle: Here is the comparison between ie8 and firefox 3 rc2:
[img]http://i169.photobucket.com/albums/u218/arifkarim/a140.gif[/img]
[img]http://i169.photobucket.com/albums/u218/arifkarim/a141.gif[/img]

I am using the latest version of uniscribe on both browsers. All other complex features of this font are working better in firefox than any other competing web browsers :) This is the only buggy problem :( I hope u developers will fix this problem soon....

Note: these digits are defined as marks in opentype language, which firefox is unable to understand.... Plz solve this problem before release... thnx

Reproducible: Always

Steps to Reproduce:
1. U need to make sure Mozilla Firefox supports latest uniscribe, Thats All!
2.
3.

Comment 1

[Simon Montagu :smontagu]

Is there a version of the font available for download for testing purposes? This sounds similar to bug 290775.

Comment 2

[Simon Montagu :smontagu]

In fact bug 290775 no longer occurs, but the testcase there just has the End of Ayah characters without digits, so it doesn't help for testing this bug. Can you attach a testcase?

Comment 3

[Simon Montagu :smontagu]

Attachment: Testcase

I can confirm this using the SIL Scheherazade font from http://scripts.sil.org/cms/scripts/page.php?site_id=nrsi&item_id=ArabicFonts_Download. I added the hacks described at http://scripts.sil.org/cms/scripts/page.php?site_id=nrsi&item_id=arabicfonts#7f69be97 to the testcase, but they don't help.

Comment 4

[Simon Montagu :smontagu]

This happens because the End of Ayah character is defined in Unicode <=5.0 with a Bidi character type of "AL" (Arabic Letter) and so is right-to-left, while the digits have a character type of "AN" (Arabic Number) and so are left-to-right. This makes us render them in separate text runs.

The definitions are being changed in Unicode 5.1, so this should be fixed by bug 427350.

Comment 5

[arifkarim]

(In reply to comment #4)
> This happens because the End of Ayah character is defined in Unicode <=5.0 with
> a Bidi character type of "AL" (Arabic Letter) and so is right-to-left, while
> the digits have a character type of "AN" (Arabic Number) and so are
> left-to-right. This makes us render them in separate text runs.
> The definitions are being changed in Unicode 5.1, so this should be fixed by
> bug 427350.


thnx Simon
its good that u regenrated the problem using shehrazade font, actually i made my font on it......

but i need to inform u that the bug still exists even in the latest version of firefox: 3.1

Can u check if the problem is solved?

Comment 6

[Simon Montagu :smontagu]

It's not solved yet, since it depends on bug 427350, which is not yet checked in (there is a patch awaiting review).

Comment 7

[arifkarim]

(In reply to comment #6)
> It's not solved yet, since it depends on bug 427350, which is not yet checked
> in (there is a patch awaiting review).

so u think this patch will remove this problem? thnx again for ur kind help.........

Comment 8

[Simon Montagu :smontagu]

I'm not 100% sure that it will be enough to solve the problem. After it's checked in, we may need an additional patch.

Comment 9

[arifkarim]

(In reply to comment #8)
> I'm not 100% sure that it will be enough to solve the problem. After it's
> checked in, we may need an additional patch.

hmmm, a guy on this mozillazine forum thread says that this bug was reported in firefox 2, so it cannot be fixed in firefox 3 builds! is he right?
http://forums.mozillazine.org/viewtopic.php?f=23&t=664795&p=4003805

Comment 10

[Simon Montagu :smontagu]

I've reset the target milestone in bug 427350 to mozilla1.9.1.

Comment 11

[Simon Montagu :smontagu]

Attachment: Testcase (with typo fix)

Comment 12

[Simon Montagu :smontagu]

Attachment: Patch

This works in Windows, half works in OS X (only the third line of the testcase works, which is the same as in Safari), and doesn't work at all in Linux. Maybe a bug in Pango?

Ehsan, I'm asking you for review in case this conflicts in some way with your work on bug 441782

Comment 13

[(no longer active)]

Comment on attachment 351188 [details] [diff] [review]
Patch

No, I don't think this conflicts with my work on bug 441782.

Comment 14

[Simon Montagu :smontagu]

Jonathan, I just noticed that you are one of the developers of the Scheherazade font. Can you throw any light on why the testcase doesn't render correctly on Linux and OS X?

Comment 15

[Jonathan Kew [:jfkthame]]

In brief, because of limitations of the shaping engines we're using.

Note that although the 3rd line does the combination in Safari, the digits now read backwards (12 becomes 21) because of the override, so this isn't really a useful workaround. I'd guess the same is true with your patch (haven't tried it yet).

When I was working on Scheherazade, we never did come up with a satisfactory way to make this work through ATSUI, except by hacky use of overrides and messing with the encoding order, so I'm not surprised at the results you're seeing. I'll be interested to see what happens with CoreText on Leopard.... that's close to being ready to try.

Re Pango, it's simply not supported there, AFAIK.

Comment 16

[arifkarim]

Thnx to both Jonathan and Simon... Glad, finally a patch for Windows is working indeed. :). But I have a question: How do  I use this patch? Also I wonder if this patch fix would be included in the future releases of Firefox?

Comment 17

[Simon Montagu :smontagu]

Pushed as http://hg.mozilla.org/mozilla-central/rev/eff94ca7dcae.

Arif, this will be in forthcoming nightly builds of Firefox and certainly in 3.2. I will also be asking approval to check it in to 3.1.

Comment 18

[Simon Montagu :smontagu]

Comment on attachment 351188 [details] [diff] [review]
Patch

This is low-risk, and well-baked on trunk.

Comment 19

[Mike Beltzner [:beltzner, not reading bugmail]]

Comment on attachment 351188 [details] [diff] [review]
Patch

a191=beltzner

Comment 20

[Simon Montagu :smontagu]

http://hg.mozilla.org/releases/mozilla-1.9.1/rev/d139006e9a6d

Comment 21

[Khaled Hosny]

Not sure if it is a regression or a different bug, but using Firefox 12 on Linux, the digits are not enclosed by the end of ayah mark on the three lines, and for other signs there are enclosed on the last line only. If I switch the font to Amiri, end of ayah works fine with the 1st two lines, but not the third, all others are broken. Since my font is using the same OpenType lookup for all the 4 signs, I expect there is something wrong with the Unicode characters properties.

Comment 22

[Khaled Hosny]

Attachment: Screenshot with Scheherazade

Comment 23

[Khaled Hosny]

Attachment: Screenshot with Amiri

Comment 24

[Simon Montagu :smontagu]

As far as I remember, and judging by comment 12, comment 14 and comment 15, this never worked in Linux.

My results are even worse than yours: with Amiri (version 0.102) the digits are enclosed by the end of ayah mark on the first two lines, but are displayed overlapping.

Comment 25

[Khaled Hosny]

(In reply to Simon Montagu from comment #24)
> My results are even worse than yours: with Amiri (version 0.102) the digits
> are enclosed by the end of ayah mark on the first two lines, but are
> displayed overlapping.

That is a bug in the font, if you use v0.101 it should be fine.

Comment 26

[Behdad Esfahbod]

If this is going through HarfBuzz, I expect it to be broken.  Because HarfBuzz flips the Arabic run to RTL before processing OpenType features.  Which apparently is not what fonts expect.

Comment 27

[Khaled Hosny]

But, Behdad, since the digits and the mark are LTR there shouldn't be any flipping, no?

Comment 28

[Behdad Esfahbod]

HarfBuzz knows that Arabic OpenType is supposed to be run on logical RTL glyph order.  So when it sees LTR Arabic, it first flips it to RTL then applies OpenType.  Which wouldn't work in this case...

Comment 29

[Khaled Hosny]

OK, I understand that, what is puzzling me is the inconsistencies between end of ayah and other signs in Firefox, HarfBuzz’s command line tools give consistent results. Anyway, do you expect this to be changed soon? I now have a version of the font that works with HarfBuzz’s current behavior as well, if it will take time to change it.

Comment 30

[Behdad Esfahbod]

Well, I'll fix it when I know how to do so without breaking LTR Arabic text.  No idea right now.

Comment 31

[Lateef Shaikh]

Attachment: Tehreer.zip

ZIP file contains Tehreer font and a web page

Comment 32

[Lateef Shaikh]

Hi,
I am still facing this issue. Please unzip Tehreer.zip file and take a look at pqm_test.html file. It uses SIL Scheherzade and PakType Tehreer font. The End of Aaya mark encloses numbers properly in Internet Explorer but not in FireFox.

I have attached the font also so that it can help you in debugging the issue may be.

Thanks
Lateef

Comment 33

[Matthias Versen [:Matti]]

Please file a new bug. This report was closed 7 years ago.

Comment 34

[Lateef Shaikh]

(In reply to Matthias Versen [:Matti] from comment #33)
> Please file a new bug. This report was closed 7 years ago.

Thanks, filed https://bugzilla.mozilla.org/show_bug.cgi?id=1506415