Closed Bug 354585 Opened 18 years ago Closed 17 years ago

[Intel-only] Accented Greek letters rendered as questionmarks

Categories

(Core Graveyard :: GFX: Mac, defect)

PowerPC
macOS
defect
Not set
major

Tracking

(Not tracked)

VERIFIED WORKSFORME
mozilla1.9alpha6

People

(Reporter: thatha, Assigned: roc)

References

()

Details

(Keywords: intl)

Attachments

(6 files)

User-Agent:       Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.8.0.7) Gecko/20060909 Firefox/1.5.0.7
Build Identifier: Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.8.0.7) Gecko/20060909 Firefox/1.5.0.7

 firefox-3.0a1.en-US.mac.dmg (27-Sep-2006 06:00), seamonkey-1.5a.en-US.mac.dmg (27-Sep-2006 07:21), and Camino (2006091101 1.0.3) fail to render properly accented characters (they appear as question marks) This is apparent at http://www.in.gr/. The accented characters render nicely on Apple's Safari. A quick test is javascript:alert('ά<- is greek alpha with an accent ó <- omicron with an accent');. Firefox appears to be handling copying and pasting of such characters properly, thus the problem seems to be specific to rendering.

PowerPC based Macs do not suffer from this problem. Camino, Firefox, and Seamonkey over Rosetta on Intel-based Macs work fine. Thus, the problem seems to be specific to the Intel portion of the rendering module.

This problem also seems to appear after installing all the recent updates from Apple. I have been able to reproduce it on 10.4.7 (8J2135a), 10.4.4 (8G1172). It works fine (without Rosetta) on Mac OS X 10.4.7 (8J2135).


Reproducible: Always

Steps to Reproduce:
See above.



If you need specific parts of the following System Profiles, I'll be glad to provide them:
* Does not work on 10.4.4-8G1172.spx
* Does not work on 10.4.7-8J2135a.spx
* Works on 10.4.7-8J2135.spx
Confirming. I see this using Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.8.0.7) Gecko/20060909 Firefox/1.5.0.7 as well as the Firefox RC1 we just released.  As the reporter notes, this problem does not manifest itself on PPC or Safari.
Status: UNCONFIRMED → NEW
Ever confirmed: true
Summary: Gecko-based browsers running on Intel-based Macs render accented characters as questionmarks → [Intel-only] Accented characters rendered as questionmarks
I think jshin might know what's going on here (ref. bug 351941).
Flags: blocking1.9?
Is it possible that the set of fonts available to a PPC build of firefox run via Rosetta is different from that available to an intel x86 binary? 

What kind of Greek fonts do you have? Do those two updates include adding new fonts? 

Yamashita-san, do you have an Intel mac to test your patch there? Is your build mentioned in bug 351941 comment #11  a univeral binary? If not, can you build a universal binary with your patch for bug 121540)and put it up somewhere so that it can be tested by Intel mac owners? 
Keywords: intl
Other Universal applications don't seem to have any trouble in terms of fonts, thus, even though I am no expert I assume that both PPC and Intel applications access the same fonts.

I'm not using any particular Greek fonts, I'm just using the default Unicode fonts  that include Greek characters (such as Times and Helvetica).

http://www.ms.u-tokyo.ac.jp/~makotoy/ffmath/firefox-2.0b2.en-US.mac.dmg contains indeed a Universal binary that <b>works fine with accented characters</b>.

This issue to be critical (especially to non-English speakers) and I believe the patch should be incorporated in the core as soon as possible...
There is a strange aspect with this bug. At kybernetes.net, I noticed that in the body of the web page, the characters are rendered incorrectly. But in the title of the web page, the characters are rendered correctly.

The following web page would have the &sigmaf; of the title rendered correctly, and the &sigmaf; of the body rendered as a question mark:

<html>
<head><title>&sigmaf;</title></head>
<body>&sigmaf;</body>
</html>
(In reply to comment #5)
> There is a strange aspect with this bug. At kybernetes.net, I noticed that in
> the body of the web page, the characters are rendered incorrectly. But in the
> title of the web page, the characters are rendered correctly.

That's because the title is handled by the OS X itself while the body is rendered  by firefox. 

(In reply to comment #4)
> Other Universal applications don't seem to have any trouble in terms of fonts,
> thus, even though I am no expert I assume that both PPC and Intel applications
> access the same fonts.

That's not a clear-cut evidence. Even though a native binary and a PPC binary run via Rosetta see a different set of fonts, it can only result in a rendering problem for applications like firefox 1.x/2 that do not use modern rendering APIs of OS X
I bet 'other univeral applications' you tested use 'modern' APIs.

> http://www.ms.u-tokyo.ac.jp/~makotoy/ffmath/firefox-2.0b2.en-US.mac.dmg
> contains indeed a Universal binary that <b>works fine with accented
> characters</b>.
> 
> This issue to be critical (especially to non-English speakers) and I believe
> the patch should be incorporated in the core as soon as possible...

I wish we could do that for FF2. Unfortunately, it will never be incorporated because 1) it's too late for FF2 and 2) FF3 (due next year) will use an entirely different rendering mechanism which does NOT have the issue here. 

Nonetheless, we have to track down the cause of this bug for FF2 and make a fix if possible. 
Yamashita-san, what is the patch applied to the source of http://www.ms.u-tokyo.ac.jp/~makotoy/ffmath/firefox-2.0b2.en-US.mac.dmg?
The above build can show all persian characters but there are too much spaces between characters.
Any updates from the Mozilla-masters? This is rather important for non-English speakers running Mac OS X on Intel Macs.
My trunk build on Intel Mac (Mac Book Pro) with the source pulled out a few hours ago doesn't have this problem. It's not a universal binary, though. 

Anyway, this problem is likely to have been fixed on the trunk thanks to a recent switch to ATSUI (see bug 121540). 

As for FF 2.0, I really have no clue why this problem shows up only on Intel mac.
I cannot reproduce this problem with FF 1.5.0.7 (on Mac OS 10.4.8 / Intel Mac Book Pro). Accented characters at http://www.in.gr get rendered well. So do accented characters at http://www.lemonde.fr  

Fx 1.5.0.9 on Mac OS 10.4.8 (Build 8N1150) on an Intel MacBook does *NOT* render properly the accented characters at http://www.in.gr . The French accented characters at http://www.lemonde.fr never had a problem.

Jungshik-- would you be as kind as to provide a screenshot from http://www.in.gr/ and the build number of your Mac OS X?

I'm unable to verify or dispute if the Minefield trunk renders properly---there's something else going wrong with the general appearance of the window. I might investigate more!

This is an important enough issue that /something/ needs to be done prior to Minefield's release. On the other hand my Mozilla experience is minimal so I don't have any real suggestions.
I'm currently experiencing this problem with SeaMonkey 2006123008 on Windows, and it seems to be unrelated to font choice, as choosing different fonts in the preferences doesn't change the display of the characters. Furthermore, sites such as http://members.at.infoseek.co.jp/DrHell/ps1/index.html, which consist mainly of Japanese characters, mostly render with question marks, even though the site works fine in firefox 1.5.0.9, when the japanese encoding is selected. In SeaMonkey, changing the encoding in the encoding menu does NOT change the encoding, the encoding stays the same on the next visit to the encoding menu, and although the page is rerendered, its appearance does not change.
(In reply to comment #12)
> Fx 1.5.0.9 on Mac OS 10.4.8 (Build 8N1150) on an Intel MacBook does *NOT*
> render properly the accented characters at http://www.in.gr . The French
> accented characters at http://www.lemonde.fr never had a problem.

Then, why did you write that this is critical to 'non-English' speakers? Anyway, I'm changing the summary to reflect that this is an  issue only for Greek. Or, is it also an issue for Cyrillic? 
 
> Jungshik-- would you be as kind as to provide a screenshot from
> http://www.in.gr/ and the build number of your Mac OS X?

Sure, I'll. There's no question mark at all. All those accented Greek characters are rendered well. 


> This is an important enough issue that /something/ needs to be done prior to
> Minefield's release. 

I believe you do have a problem. However, I can't reproduce it. How about installing all the fonts on your Mac OS X CD/DVD? 



(In reply to comment #13)
> I'm currently experiencing this problem with SeaMonkey 2006123008 on Windows,

Please, file a *separate* bug for your issue. This bug is not about Windows problem. 
Summary: [Intel-only] Accented characters rendered as questionmarks → [Intel-only] Accented Greek letters rendered as questionmarks
here goes the pdf 'shot' of Firefox 2.0 rendering www.in.gr on Mac Book Pro (intel mac). 

What's your font setting for Greek,  Theocharis? Do changing it to use 'serif' and setting 'Serif' font for Greek to 'Lucida Grande' make any difference?
Another example showing question marks for accented Greek characters is http://en.wikipedia.org/wiki/Greece

In Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.8.1.1) Gecko/20061204 Firefox/2.0.0.1 on MacBook Pro, accented characters are shown as question marks. 

The same accented characters are displayed correctly in Safari Version 2.0.4 (419.3) on the same machine.
Here is a screenshot showing http://en.wikipedia.org/wiki/Greece rendered in Firefox and Safari on MacBook Pro, System Version:Mac OS X 10.4.8 (8L2127) Kernel Version:Darwin 8.8.1.

Shows greek accented characters rendered correctly in Safari, but not Firefox.
I do believe you have a problem. 

Please, tell me what's your font preference settings for Greek. Go to Preferences | Content. In Font & Color, press 'Advanced' button. In the dialog box popping up, select 'Greek' and set Proportional to 'Serif' and 'Serif' font to 'Lucida Grande'.  Moreover,  uncheck 'Allow pages to use their own fonts instead of my selections above'. And, then see you still have the problem.

If your problem disappears with the above setting, I think this is not Intel-mac specific. This issue is well-known and you just have to work around it with the above setting until FF 3.0 is released. 
Attached image settings for Greek
Thanks for your help. 

Switching to lucida grande does not seem to help. (My settings shown in the attachment). May just have to wait for FF 3.
(In reply to comment #14)
> Then, why did you write that this is critical to 'non-English' speakers?
> Anyway, I'm changing the summary to reflect that this is an  issue only for
> Greek. Or, is it also an issue for Cyrillic? 
Because as Ali Rastegar reported in bug ??? it is also an issue with Persian characters. IPA, some Cyrillic characters, and some Katakana extensions (U+31F0 ... U+31FF) are also messed up. It seems a problem spread over numerous different Unicode ranges.

> Sure, I'll. There's no question mark at all. All those accented Greek
> characters are rendered well. 
True indeed!

> I believe you do have a problem. However, I can't reproduce it. How about
> installing all the fonts on your Mac OS X CD/DVD? 
Doubtful! As comment #17 implies there are others having the exact same problem. I have personally reproduced the issue on the following versions of Mac OS: 10.4.4-8G1172 and10.4.7-8J2135a, while it worked fine on Mac OS 10.4.7-8J2135. All of these installations were by different people so it's nearly impossible that all of us did the same "mistake" while installing Mac OS. I believe I have all the available fonts installed.

(In reply to Comment #18)
The settings were exactly as you mentioned and all the above-mentioned scripts are still rendered improperly. The fact that there are more than one users experiencing problems in numerous scripts while Safari works fine implies that it's not a font-related installation-specific problem. The problem occurs in Camino too---too me, that means that's something in Gecko. Now, why this works fine on your Intel Mac beats me---I'm no Mac expert...

What's the difference between 10.4.7-8J2135a and 10.4.7-8J2135? It works fine on the latter, improperly on the former. Also, Yamashita's patched build works perfectly---he uses a different rendering mechanism: that also points us to the direction of the Fx2 rendering mechanism.
(In reply to comment #20)
> Because as Ali Rastegar reported in bug ??? it is also an issue with Persian
Ooops. bug ??? -> bug 357673
Flags: blocking1.9? → blocking1.9+
What is the status of this bug? Last comment was #20 on 1/2/07. Should we wait till FF3?
Erm, this sounds like a generic text rendering bug.. bouncing over to roc; do you have someone that can take a look and see if this is still a problem?  I think that this would probably be fixed by all the text rendering work that's been going on lately...
Assignee: joshmoz → roc
Target Milestone: --- → mozilla1.9alpha6
Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.8.1.3) Gecko/20070309 Firefox/2.0.0.3 is still rendering accented characters as question marks.

Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.8.1.4pre) Gecko/20070504 BonEcho/2.0.0.4pre is also rendering said characters improperly.

Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.8.1.1) Gecko/20061211 BonEcho/2.0.0.1, downloaded from http://www.ms.u-tokyo.ac.jp/~makotoy/ffmath/ff2mlte.html, is rendering said characters correctly.
works fine in my Intel Mac trunk build.
Robert, this bug seems to manifest only on some Intel Macs.
I'm still a PPC user, but I have received many reports from Intel Mac Users that this bug is fixed in Firefox 3.0 Alpha(Gran Paradiso).(at least for Persian/Farsi letters).
Greek accented characters appear to render correctly for me with the nightly build:

Mozilla/5.0 (Macintosh; U; Intel Mac OS X; en-US; rv:1.9a5pre) Gecko/20070505 Minefield/3.0a5pre
I am using MacBook Pro and have this bug. Comment#30 Joseph G. Briggs refers to a nightly build that corrects this bug would tell me how can I get it to try it on my system. Thanks

It seems as severe this bug is to most users other than Latin letters users it has been ignored for sometime. 
(In reply to comment #31)
> I am using MacBook Pro and have this bug. Comment#30 Joseph G. Briggs refers to
> a nightly build that corrects this bug would tell me how can I get it to try it
> on my system. Thanks

I downloaded it from here http://ftp.mozilla.org/pub/mozilla.org/firefox/nightly/latest-trunk/firefox-3.0a5pre.en-US.mac.dmg
Thanks. I used the above link and used this build. It has corrected the previous problem. However, Now it has now a new problems. it is corrected a font problem with a new fonts and spacing between the fonts. it is worse than before.
 
(In reply to comment #33)
> Thanks. I used the above link and used this build. It has corrected the
> previous problem. However, Now it has now a new problems. it is corrected a
> font problem with a new fonts and spacing between the fonts. it is worse than
> before.
> 
> 
Would you please also test this build?
http://download.mozilla.org/?product=granparadiso-alpha4&os=osx&lang=en-US
Does it have too many spaces between letters like the another build you have tested?
Thanks in advance
I have not tested any other build. But this one spacing is not very much but it is off horizontal and vertical. Also a square shows up in some. Again I did not try to document and look at all the fonts. I left that for experts. I continue to use the older version. since I kind of figured it out where and for what font "?" appears.
It sounds like this bug is fixed in trunk nightly builds.

If there's still a bug with http://www.in.gr, please open a new bug with a precise description of what the problem is.
Status: NEW → RESOLVED
Closed: 17 years ago
Resolution: --- → WORKSFORME
Where do I find the trunk nightly builds?
Thanks
(In reply to comment #37)
> Where do I find the trunk nightly builds?
> Thanks
> 

http://www.google.com/search?q=trunk%20nightly%20builds
1st attachment is from Safari. 2nd one is from firefox when bug was reported.
And finally 3rd attachment is screen shot of nightlyBuild partially corrected.
See next two comments for 2nd and 3rd attachments
Please file new bugs for your issues. They are quite likely to be different issues to the issues in this bug.
Why you think this is a new bug? same type problem with same letters except spacing and connecting letters.
I am getting a feeling you are given up on fixing it! 

First time I report it as a new bug. I was warned that this bug is already reported and should report to it. Now Being told to report it as new bug.
We cannot be sure this is the same bug. 
The first bug is: some characters in some languages display as question marks. The current fix for this bug resolves it completely for the most languages, but not for Farsi/Persian. 
For Persian/Farsi it can now show the characters but they are separated from other characters. It might be a new bug caused by the current fix.
Whatever it is, please file a new bug as suggested by developers.(Im a PowerPC user, If I were an Intel User I would file a new bug now.)
Is there a Nightly Build for thunderbird also. Same problem of ? exist in my thunderbird.
(In reply to comment #45)
> Is there a Nightly Build for thunderbird also. Same problem of ? exist in my
> thunderbird.
> 
I think this the link to latest nightly build of Thunderbird:
http://ftp.mozilla.org/pub/mozilla.org/thunderbird/nightly/latest-trunk/thunderbird-3.0a1pre.en-US.mac.dmg
Firefox 2.0.0.9 with Mac OS X 10.5 9A581 (Leopard Gold Master) renders Greek properly.
Firefox 3.0.a9pre nightly build (20071031) on Mac OS X 10.5 9A581 also renders Greek properly.
Status: RESOLVED → VERIFIED
Product: Core → Core Graveyard
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: