User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.6; rv:2.0) Gecko/20100101 Firefox/4.0 Build Identifier: On Mac OS X, Hebrew text falls back to the buggy font Arial Hebrew. Arial Hebrew will display the dagesh mark in the wrong place. While this is really not a mozilla bug, there is no knowing when Apple will fix this font (if ever). So for the time being, a better fallback font should be chosen. The other default Hebrew font, Raanana for serif text, suffers from the very same bug. Reproducible: Always Steps to Reproduce: View an unstyled sequence (e.g. in Firefox or Thunderbird) of: 1. any Hebrew letter, 2. any niqqud mark, 3. dagesh, 4. another Hebrew letter, for instance: בֵּית (bet-tsere-dagesh-yod-tav) 'house' Actual Results: The dagesh mark is displayed on the following letter (in the example בֵּית, on the letter י (yod). Expected Results: The dagesh mark should be displayed on the preceding letter (in the example בֵּית, within the letter ב (bet). For Mac OS X, I would suggest using the following fonts (preinstalled of course) instead of the buggy fonts Arial Hebrew and Raanana: default fallback/sans-serif: Lucida Grande or Arial serif: Times New Roman monospace: Courier New Safari and Chrome will default to Lucida Grande, so they have no problem. Incidently, the inverse order is displayed correctly (1. any Hebrew letter, 2. dagesh, 3. any niqqud vowel mark, 4. another Hebrew letter), for instance בֵּית (bet-dagesh-tsere-yod-tav). However, normalization operations will switch to the other order.
Yes, I can confirm the buggy rendering with Arial Hebrew. That looks like a font bug to me (though I haven't investigated deeply); it would make sense to specify something different as the default for Hebrew so as to avoid the issue. Simon, any thoughts on the most appropriate choices?
I have written a simple test page that illustrates the issue in two tables (with pictures): First the buggy current fallback fonts; second the alternative correct fallback fonts I would propose: http://unifraktur.sourceforge.net/testcases/bug648248illustration/Illustration%20of%20Bug%20648248.html
Created attachment 524635 [details] [diff] [review] Bug 648248 - Change fallback Hebrew fonts to non-buggy fonts I have tried to write my very first patch. It is supposed to change the fonts according to what I have proposed here (as illustrated on http://unifraktur.sourceforge.net/testcases/bug648248illustration/Illustration%20of%20Bug%20648248.html ). I do not know whom to put in as reviewer.
Comment on attachment 524635 [details] [diff] [review] Bug 648248 - Change fallback Hebrew fonts to non-buggy fonts Thanks for the patch! Asking smontagu for review, as he's knowledgeable about Hebrew.
Comment on attachment 524635 [details] [diff] [review] Bug 648248 - Change fallback Hebrew fonts to non-buggy fonts I would prefer changing Arial Hebrew to Arial, instead of Lucida Grande, which I find rather ugly in Hebrew.
(In reply to comment #5) > Comment on attachment 524635 [details] [diff] [review] > Bug 648248 - Change fallback Hebrew fonts to non-buggy fonts > > I would prefer changing Arial Hebrew to Arial, instead of Lucida Grande, which > I find rather ugly in Hebrew. One point to note is that older versions of Arial don't render Hebrew diacritics well under Core Text. In a brief test, they look OK with v5.01 (which I think is the version currently shipped with the OS), but fail with 3.05, which I had installed locally - it probably comes from MS Office or something like that. I wouldn't be surprised if it's fairly common for users to have local versions of Arial that "hide" the default system version, in which case they may see poorer results. Enabling harfbuzz for Hebrew fixes the rendering with Arial 3.05, so if we fix bug 623495 this will be less of a concern.
Created attachment 525063 [details] [diff] [review] patch, change Hebrew fallback fonts - updated Updated patch to use Arial for sans-serif. firstname.lastname@example.org, would you like to provide a full name to be included in the commit message when we make this change? Normally, we use "full name <email.address>" to record the patch author in the history.
Created attachment 525358 [details] [diff] [review] patch for checkin patch for checkin, carrying forward r=smontagu
Pushed to mozilla-central: http://hg.mozilla.org/mozilla-central/rev/b1365c6e3d08 This caused reftests bidi-004 and bidi-004-j to fail on OS X 10.5, because the version of Times New Roman there doesn't include the Hebrew presentation-form characters. Pushed a followup to fix this by specifying Raanana as the first-choice font for those tests: http://hg.mozilla.org/mozilla-central/rev/1156b1885571 http://hg.mozilla.org/mozilla-aurora/rev/e47cd3cac157
We are getting user complains on our community forum from osx users having fallback to wrong fonts and no text at all after disabling these fonts. Could it be caused by this issue? http://mozilla.org.il/board/viewtopic.php?f=9&t=10256&start=15
Sorry, but I'm not able to read discussions in Hebrew or identify specific problem reports there. If you can point me to a reduced test-case that exhibits these problems, I'll try to look into it. Probably best to file a new bug for any such issue, actually. Attaching a small test file that demonstrates the failure would be most helpful.