485179 - text in *-Latn does not follow sans-serif

Status: NEW

(Core :: Internationalization, defect)

Description

[Shreevatsa R]

User-Agent:       Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-GB; rv:1.9.0.7) Gecko/2009021906 Firefox/3.0.7
Build Identifier: Mozilla/5.0 (Macintosh; U; Intel Mac OS X 10.5; en-GB; rv:1.9.0.7) Gecko/2009021906 Firefox/3.0.7

For <span style="font-family: sans-serif" lang="yi-Latn">text</span> (for example), the font used is the default serif font even when it is supposed to be in sans-serif.

This is related to, but not a duplicate of, Bug 192636.

Reproducible: Always

Steps to Reproduce:
1. Have a block of sans-serif text.
2. Have some text enclosed in lang=xx-Latn where xx={ar,sa,yi,etc.}.
3. View text.
Actual Results:  
The text is displayed in the default serif font.

Expected Results:  
Text ought to be displayed in the default sans-serif font.

This sort of text is common on Wikipedia.

Comment 1

[Shreevatsa R]

Attachment: Test case with a few examples

Comment 2

[Shreevatsa R]

Attachment: Screenshot with default serif font set to Apple Chancery 30

Screenshot with default serif font set to Apple Chancery 30 for exaggerated effect, and sans-serif font set to Helvetica.

Note that this is *not* Bug 192636: that bug is about the last three (el-Latn ru-Latn ja-Latn) being displayed in a different font from the rest. This is about the first three (sa-Latn ar-Latn yi-Latn) being displayed in the *default* serif font when they should fall back to sans-serif instead. This bug can be fixed without fixing that bug.

Comment 3

[Shreevatsa R]

This bug might be Mac/Windows-specific: it seems to work fine on this Ubuntu system (in fact the last three also seem to look fine).

Comment 4

[Shreevatsa R]

I may have spoken too soon. What is actually seen here is that "normal" and "en-Latn" are in the default serif font or the default sans serif font on the respective lines, and the rest are all in either Bitstream Vera Serif or Bitstream Vera Sans  on the respective lines (whatever the default fonts are set to). Still, at least they follow the serif/sans-serif distinction, which they don't on Mac.

Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9.0.8) Gecko/2009032711 Ubuntu/8.10 (intrepid) Firefox/3.0.8

Comment 5

[Derek Harkness]

I have encountered this bug when using Pinyin romanizations of Chinese text and marking it up as zh-latn-pinyin. Firefox in Windows 7 displays the pinyin text in the same font as the Chinese characters (marked as zh-hans). It seems Firefox in Windows is following the language code zh and ignoring the script latn. Testing in Ubuntu with Firefox showed that this bug is Windows version specific.

Comment 6

[John Blackburne]

Attachment: Patch to add support for Latn subtag, as well as Bopomofo

It's not just a Windows problem. I was able to reproduce it on the Mac version with certain settings. I've had a look at it and created a patch for it. It adds support for all the scripts listed as well as a number of Chinese Romanisations/scripts that were my motivation for looking at this in the first place, after reports of problems on Wikipedia.

Comment 7

[Gordon P. Hemsley [:GPHemsley]]

(In reply to Shreevatsa R from comment #1)
> Created attachment 369271 [details]
> Test case with a few examples

To confirm, in this testcase (in Aurora 32 on Ubuntu Linux 12.04):

Always in serif:
* sa-Latn (Sanskrit)
* ar-Latn (Arabic)

Uses a different sans-serif from the rest:
* ja-Latn (Japanese)

Comment 8

[Gordon P. Hemsley [:GPHemsley]]

Attachment: Testcase: Comparison of font-face selection based on language and font-family

Here's a more detailed testcase that shows a number of different language and script combinations with the font-family keyword values of 'inherit', 'serif', and 'sans-serif'.

Comment 9

[Gordon P. Hemsley [:GPHemsley]]

Comment on attachment 8424410 [details] [diff] [review]
Patch to add support for Latn subtag, as well as Bopomofo

This surely is not the best approach, but it may be the only thing we can do until bug 251241 is fixed.

John, the review+ flag is reserved for when a patch is actually reviewed and approved by a relevant peer. To request review, use review?, as I have here.

Comment 10

[Gordon P. Hemsley [:GPHemsley]]

(In reply to Gordon P. Hemsley [:GPHemsley] from comment #9)
> This surely is not the best approach, but it may be the only thing we can do
> until bug 251241 is fixed.

Sorry, I meant bug 556237 and bug 756022.

Comment 11

[John Blackburne]

(In reply to Gordon P. Hemsley [:GPHemsley] from comment #9)

Sorry that was my first time here, so I had no idea how to ask for it to be looked at; having looked at it I just wanted some other eyes on it.

And yes, it's definitely not the best approach: The Latn subtag exists so you don't need to specify each and every *-Latn. But that's how it's working now.

Comment 12

[Jonathan Kew [:jfkthame]]

Comment on attachment 8424410 [details] [diff] [review]
Patch to add support for Latn subtag, as well as Bopomofo

I'm not sure offhand how many of these transliteration schemes would actually be covered by Latin-1, which is what x-western is presumably intended to mean (more or less). Certainly writing systems such as transliterated Arabic and Chinese Pinyin use characters that are not present in Latin-1. As such, I wonder whether it'd be more appropriate to map them to x-unicode?

Maybe it doesn't much matter, given that many "western" fonts now have extended character coverage; if the x-western font pref calls for a major system font like Arial, it'll probably be OK.

I'd like to see what Simon thinks about this; redirecting review there for now.

Comment 13

[Simon Montagu :smontagu]

Comment on attachment 8424410 [details] [diff] [review]
Patch to add support for Latn subtag, as well as Bopomofo

Review of attachment 8424410 [details] [diff] [review]:
-----------------------------------------------------------------

I'm happy with doing it this way for now, though unifying x-western, x-central-euro, x-baltic and I don't know what all else is long overdue! (bug 756022)

There may indeed be issues with a "ransom note effect" because of characters like e.g. ḥ or ḡ not being available in the default x-western fonts, but I don't expect that x-unicode will be any better in practise.

Comment 14

[BugBot [:suhaib / :marco]]

The bug assignee didn't login in Bugzilla in the last 7 months, so the assignee is being reset.