1164279 - Some characters of ruby text size is small in certain font

Status: RESOLVED FIXED

(Core :: Layout, defect)

Description

[Alice0775 White]

Attachment: index.html

Ruby text か,け,カ,ケ is small if font is Meiryo.

Steps To Reproduce:
1. Open attached
   or http://www3.nhk.or.jp/news/easy/k10010076251000/k10010076251000.html

Actual Results
Ruby text か,け,カ,ケ is small

Expected Results:
Ruby text か,け,カ,ケ should be same size as the other characters.

Comment 1

[Alice0775 White]

Attachment: screenshot of attached testcase

Comment 2

[Alice0775 White]

Attachment: screenshot of the web page

Comment 3

[Xidorn Quan [:xidorn] UTC+11]

Attachment: related glyphs in the font

This seems to be a bug of the font. We apply "font-variant-east-asian: ruby" property to ruby text conforming with the CSS Ruby spec. However, Meiryo has incorrect mapping for the characters you pointed out.

The attachment is the related glyphs I extracted from Meiryo. You can compare the glyph ids, and see how they map the glyphs reversely between normal kana and small kana.

I have no idea what should we do for this. The font authors ought to fix their font.

Comment 4

[Alice0775 White]

So, in this case, workaround is rt { font-variant-east-asian: normal; } .


This affect to Windows Vista, Windows7, Windows8.1, Windows10IP.
And I do not think that MS fixes the font....

Comment 5

[Xidorn Quan [:xidorn] UTC+11]

Yes, but that rule works fine for other fonts, and IIRC, Meiryo is not our default font. Hence we probably shouldn't remove that rule. Not sure, anyway.

jfkthame, do you think there is anything we should do here?

Comment 6

[Xidorn Quan [:xidorn] UTC+11]

We probably could have a quick hack on the specific fonts and version which swaps those substitution if we really care about that font. Not sure whether we should and how we could do that.

Comment 7

[Jonathan Kew [:jfkthame]]

This is clearly a font bug. I'd be pretty reluctant to add custom code to recognize the specific font/version and work around it -- there are probably multiple versions of Meiryo shipped with different Windows releases, so besides the ugliness and overhead of adding such a hack in the first place, it'd be a maintenance headache every time there's an update from MS.

I've reported the problem to some contacts at MS, but obviously we're not going to see an immediate fix; these things tend to grind pretty slowly.

For now, my inclination is to do nothing, and blame it on the font in case of complaints. Authors who are affected by it can explicitly set rt { font-variant-east-asian: normal; } on their content, and the result will be no worse than other browsers that don't apply the feature at all.

Comment 8

[Jonathan Kew [:jfkthame]]

Note that the same problem shows up in Chrome, for example, if you apply the ruby feature by adding

  rt { -webkit-font-feature-settings: "ruby"; }

to the testcase.

Comment 9

[Jonathan Kew [:jfkthame]]

Just FTR, the current "owner" of the Meiryo fonts at MS has responded saying "I’ll get it on the list of fixes", so I believe we can hope to see a corrected font shipping at some future date.

To avoid the bug in the meantime, authors should say

  rtc, rt { font-variant-east-asian: normal; }

if they're intending to use these characters in the Meiryo font.

Comment 10

[Xidorn Quan [:xidorn] UTC+11]

Attachment: patch to disable "ruby" feature on Windows

Since Meiryo is widely used on Japanese websites, this seems to affect a wide range of people, who think this is a bug of Firefox, because it doesn't exist on other browsers.

We should just disable that feature on Windows, and re-enable it when Microsoft fixes the font and most people get the update.

Comment 11

[Jonathan Kew [:jfkthame]]

I suppose this may be the best thing to do for now, but it makes me sad.... it'll mean that any fonts with a proper 'ruby' feature will behave differently by default in FF/Win than on other platforms.

Now that Win10 is released, is there any news here -- an updated version of Meiryo for Japanese Win10, possibly? (The font doesn't seem to be present at all on a default en-US install.)

:jdaggett, wdyt? Others in moz-japan?

Comment 12

[Alice0775 White]

Attachment: modified testcase (due to landing CSS Ruby autohiding Bug 1197765)

Comment 13

[Alice0775 White]

Attachment: modified testcase (due to landing CSS Ruby autohiding Bug 1197765)

Comment 14

[Xidorn Quan [:xidorn] UTC+11]

Actually, not only the four kanas mentioned in comment 0, some others are also weird with 'ruby' feature, noticeably チ, ナ, and ホ.

:jfkthame, could you also ask the owner of the font to fix them as well? Also could you ask them for a timetable when they are going to release the fixed version?

Comment 15

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

Would it make more sense to disable the feature (in C++ code) for fonts named Meiryo, rather than disabling it for all ruby text on Windows?

Comment 16

[Xidorn Quan [:xidorn] UTC+11]

(In reply to David Baron [:dbaron] ⌚UTC-7 from comment #15)
> Would it make more sense to disable the feature (in C++ code) for fonts
> named Meiryo, rather than disabling it for all ruby text on Windows?

I mentioned that in comment 6.

It adds cost in some common pathes. Also if we decide to workaround this issue, I hope to uplift the patch up to beta so that Firefox seems to "fix" this issue in the next release. But adding C++ code to disable it adds more risks which woule make me afraid to uplift it.

Comment 17

[John Daggett (:jtd)]

(In reply to Jonathan Kew (:jfkthame) from comment #11)
> I suppose this may be the best thing to do for now, but it makes me sad....
> it'll mean that any fonts with a proper 'ruby' feature will behave
> differently by default in FF/Win than on other platforms.
> 
> Now that Win10 is released, is there any news here -- an updated version of
> Meiryo for Japanese Win10, possibly? (The font doesn't seem to be present at
> all on a default en-US install.)
> 
> :jdaggett, wdyt? Others in moz-japan?

This is truly sad, indeed. The shipping version of Win 10 JA includes Meiryo with the problematic ruby variants. I think disabling this on Windows is the best choice. We should probably add an author suggestion on an MDN page somewhere to use font-variant: ruby when not using Meiryo.

Note that there is a difference between the normal hiragana/katakana か and け and the small variants, which appear only in certain situations:

  ２ヶ月
  ２ヵ月

Comment 18

[John Daggett (:jtd)]

Attachment: screenshot, buggy Meiryo variants vs. Yu Mincho / Yu Gothic under Win 10

Comment 19

[Jonathan Kew [:jfkthame]]

Comment on attachment 8659097 [details] [diff] [review]
patch to disable "ruby" feature on Windows

Review of attachment 8659097 [details] [diff] [review]:
-----------------------------------------------------------------

OK, r=me (sigh). I've also sent a brief nag-mail to the Meiryo owner at MS.

Comment 20

[Xidorn Quan [:xidorn] UTC+11]

https://hg.mozilla.org/integration/mozilla-inbound/rev/cd4f5bb41dd73ba0eebb6b597364b77358441d0f
Bug 1164279 - Disable font-variant-east-asian: ruby for <rt> on Windows. r=jfkthame

Comment 21

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/cd4f5bb41dd7

Comment 22

[Xidorn Quan [:xidorn] UTC+11]

Comment on attachment 8659097 [details] [diff] [review]
patch to disable "ruby" feature on Windows

Approval Request Comment
[Feature/regressing bug #]: bug 1039006
[User impact if declined]: see different font size on some specific characters on ruby when Meiryo is used. Meiryo is a widely-used Japanese font on Windows. Although it is not the default font on Firefox, many websites use this font on Windows.
[Describe test coverage new/current, TreeHerder]: n/a
[Risks and why]: very low risk, as it simply removes one cosmetic rule on Windows
[String/UUID change made/needed]: n/a

Comment 23

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

Comment on attachment 8659097 [details] [diff] [review]
patch to disable "ruby" feature on Windows

At this point in the 41 cycle (RC week), only bugs which are sec-crit + easy to exploit, severe crash/hangs and major regressions that make the release unusable will be considered. Given that, this is a wontfix for 41.

Comment 24

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

This is a wontfix for 41.

Comment 25

[Sylvestre Ledru [:Sylvestre]]

Comment on attachment 8659097 [details] [diff] [review]
patch to disable "ruby" feature on Windows

OK, still time with 42 to find potential regressions.

Comment 26

[Carsten Book [:Tomcat]]

(In reply to Sylvestre Ledru [:sylvestre] from comment #25)
> Comment on attachment 8659097 [details] [diff] [review]
> patch to disable "ruby" feature on Windows
> 
> OK, still time with 42 to find potential regressions.

does this need to beta now ?

Comment 27

[Sylvestre Ledru [:Sylvestre]]

Comment on attachment 8659097 [details] [diff] [review]
patch to disable "ruby" feature on Windows

Yes, let's take it.

Comment 28

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/releases/mozilla-beta/rev/7bea04fcb6d4

Comment 30

[Xidorn Quan [:xidorn] UTC+11]

It looks like in the latest version of Meiryo (version 6.26) shipped with Windows 10 Anniversary Update, this issue has been fixed.

The issue in comment 14 still presents, but given that the majority of furigana is written in hiragana, I guess it is fine to re-enable "ruby" font feature now.

Comment 31

[Xidorn Quan [:xidorn] UTC+11]

Attachment: Bug 1164279 - Re-enable ruby variant on Windows.

Review commit: https://reviewboard.mozilla.org/r/75838/diff/#index_header
See other reviews: https://reviewboard.mozilla.org/r/75838/

Comment 32

[Xidorn Quan [:xidorn] UTC+11]

jfkthame, could you inform me who is the owner of Meiryo in Microsoft? I'd like to send an email about the issues in ruby katakana mentioned in comment 14 as well. Although it is not as serious as the size issue, it would still make Firefox's rendering look worse than other browsers in some cases.

Comment 33

[Jonathan Kew [:jfkthame]]

(In reply to Xidorn Quan [:xidorn] (UTC+10) from comment #30)
> It looks like in the latest version of Meiryo (version 6.26) shipped with
> Windows 10 Anniversary Update, this issue has been fixed.
> 
> The issue in comment 14 still presents, but given that the majority of
> furigana is written in hiragana, I guess it is fine to re-enable "ruby" font
> feature now.

Do we have any stats for the number of Windows users who are on the Win10AU release at this point compared to those on older versions (where the bad font will still be present), or what the adoption curve is looking like for the new update?

Comment 34

[Xidorn Quan [:xidorn] UTC+11]

(In reply to Jonathan Kew (:jfkthame) from comment #33)
> Do we have any stats for the number of Windows users who are on the Win10AU
> release at this point compared to those on older versions (where the bad
> font will still be present), or what the adoption curve is looking like for
> the new update?

I don't know how to measure that, but I believe Telemetry has that data... Given that Microsoft's upgrading strategy is pretty progressive nowadays, I expect most of normal users should have upgraded.

I'll try to analyse that later.

Comment 35

[Jonathan Kew [:jfkthame]]

Note that according to the July 2016 chart currently at https://en.wikipedia.org/wiki/Usage_share_of_operating_systems#Desktop_and_laptop_computers, for example, Win10 is still substantially behind Win7. Even assuming all Win10 users get the update quickly, there are something like twice as many users on pre-Win10 versions (primarily 7, some 8.1) who will still be affected.

Comment 36

[Xidorn Quan [:xidorn] UTC+11]

Comment on attachment 8786997 [details]
Bug 1164279 - Re-enable ruby variant on Windows.

Hmmm... okay, so it looks like yeah, Windows 7 is still the majority [1], and the majority of Windows 10 users haven't upgraded to the Anniversary Update [2]...

I guess we don't want to land this patch at the moment, then.

[1] https://sql.telemetry.mozilla.org/queries/380#635
[2] https://sql.telemetry.mozilla.org/dashboard/windows-10-user-strata

Comment 37

[Alice0775 White]

(In reply to Xidorn Quan [:xidorn] (UTC+10) from comment #30)
> It looks like in the latest version of Meiryo (version 6.26) shipped with
> Windows 10 Anniversary Update, this issue has been fixed.
> 
> The issue in comment 14 still presents, but given that the majority of
> furigana is written in hiragana, I guess it is fine to re-enable "ruby" font
> feature now.

I can still reproduce the problem with the following CSS(stylish) on Windows10 Home Ver.1607 build 14393.105, Meiryo vaersion 6.26.

rtc, rt {
  font-variant-east-asian: ruby !important;
}

Comment 38

[Alice0775 White]

(In reply to Alice0775 White from comment #37)
> (In reply to Xidorn Quan [:xidorn] (UTC+10) from comment #30)
> > It looks like in the latest version of Meiryo (version 6.26) shipped with
> > Windows 10 Anniversary Update, this issue has been fixed.
> > 
> > The issue in comment 14 still presents, but given that the majority of
> > furigana is written in hiragana, I guess it is fine to re-enable "ruby" font
> > feature now.
> 
> I can still reproduce the problem with the following CSS(stylish) on
> Windows10 Home Ver.1607 build 14393.105, Meiryo vaersion 6.26.
> 
> rtc, rt {
>   font-variant-east-asian: ruby !important;
> }

Never mind.

Upgrading from Windows7 to Windows10 seem to be something corrupted.
I remove the following key and data from registry. It works as expected.

[HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\Fonts]
"メイリオ ボールド & メイリオ ボールド イタリック & Meiryo UI Bold & Meiryo UI Bold Italic (TrueType)"="meiryob_4.ttc"
"メイリオ & メイリオ イタリック & Meiryo UI & Meiryo UI Italic (TrueType)"="meiryo_4.ttc"

Comment 39

[Xidorn Quan [:xidorn] UTC+11]

I guess we should consider this fixed as of comment 21. We may still want to re-enable ruby feature on Windows at some point, but that can go to a different bug.

Comment 40

[Sarah Clements [:sclements]]

Bug 1758952 was filed to update the CSS condition for Windows11.