Open Bug 1941202 Opened 27 days ago Updated 16 days ago

[win] FPP font name changes

Categories

(Core :: Layout: Text and Fonts, enhancement)

enhancement

Tracking

()

People

(Reporter: thorin, Unassigned)

References

(Blocks 1 open bug)

Details

Attachments

(3 files)

In a recent windows 11 update (I update every month and just noticed this - super-likely the the 2024-12 cumulative update for 23HR as this was done 9th jan and the windows.old dir is dated then as well) .. the UD Digi Kyokasho range of fonts "changed" names (i think)

So previous TZP used the following (-R regular -B bold) in it's font enumeration tests

'UD Digi Kyokasho N-B','UD Digi Kyokasho N-R',
'UD Digi Kyokasho NK-B','UD Digi Kyokasho NK-R',
'UD Digi Kyokasho NP-B','UD Digi Kyokasho NP-R',

This worked (I would have tested, and I have all windows fonts including some downloadable ones such as Cascadia Code) so I could check names worked in font-families etc.

Now it fails. Looking in Windows.old/Windows/Fonts vs Windows/Fonts, the difference seems to be now those fonts have dropped the -R, -B parts (pics to follow), so I think what we want to do is add UD Digi Kyokasho N and UD Digi Kyokasho NK and UD Digi Kyokasho NP to kBaseFonts kLangPackFonts

https://searchfox.org/mozilla-central/source/gfx/thebes/StandardFonts-win10.inc#182-187

If you would like me to upload the fonts, let me know

Attached image pic.png

with font.vis/FPP disabled UD Digi Kyokasho N, UD Digi Kyokasho NK, and UD Digi Kyokasho NP are detected (if tested for in my local TZP), but not with FPP (e.g. ETp strict) enabled - and the -R and -B are not detected in any level (any more)

Attached image test.png

https://arkenfox.github.io/TZP/tests/fontdebug.html

testing UD Digi Kyokasho N which is a kLangPackFonts font, so should show up with ETP Strict

  • does not show in strict
  • does show in standard (to show that the font name works)
Attached image test -r.png

https://arkenfox.github.io/TZP/tests/fontdebug.html

testing UD Digi Kyokasho N-R which is a kLangPackFonts font, so should show up with ETP Strict

  • does not show in strict
  • does not show in standard (to show that the font name no longer works - assuming I am 99% sure it used to work)
Blocks: 1836535

This seems like another instance of OSes renaming fonts and it segmenting users. We saw the same thing on Android quite significantly.

I super-hate the idea of a mapping layer where we map font names to each other, but it seems like this might be important. (Hopefully we'll have data soonish to know for sure...)

We could map them but are we certain that it is only a rename? What if they added support for some character(s)? Even if we mapped them, fingerprinters could still check with that newly added character (or removed) and still figure out whether a machine has this change or not

I was going to say this is a great candidate for bug 1939286, but microsoft doesn't allow embedding fonts to applications freely, strictly only to documents unfortunately.

FWIW it also removed HoloLens MDL2 Assets and there is no easy way to anymore to add it (sure, I could grab it from windows.old but people aren't techy). The UI interface for adding fonts from MS has changed. My laptop (win 11 pro same as my desktop) hasn't updated yet, so I will do a compare. Now it just gives me an option to download "all language fonts" (before it was selective IIRC) - otherwise I can add what looks like crap from the MS store

I also would like to know why AlternateGothic2 BT is there - never seen or heard of it

Why do we need AlternateGothic2 BT or HoloLens MDL2 Assets at all?

I super-hate the idea of a mapping layer where we map font names to each other

on windows we have FontSubtitutes, but of course that wouldn't hide the differences in FontFace (the font has to actually exist IIUIC)

What if they added support for some character(s)

Unless you intend to bundle almost all fonts, like Tor Browser, you will never have control over font versions and their changes - which is a fairly expensive FP probe, touching the dom and measuring a lot of glyphs - that said textmetrics is quite speedy these days, but scripts have to await font fallback for stable results. I feel this is out of scope for Firefox, and the first step is to limit font name enumeration

microsoft lists HoloLens font as available since 2015 on Windows 10 version 1511. I'm not sure (literally) if we consider that widely available or not. About 7% of volunteers don't have it installed.

Regarding AlternateGothic2, I have no idea where that came from too. microsoft doesn't even list it, but I wasn't around at the time of this font list creation. Based on our volunteer data, no one had this font installed.

About 7% of volunteers don't have it [HoloLens] installed

likely win11 users who have installed the recent update, eventually all win11 users will catch up (about 42% of Firefox win users currently are win11)

Should I open a new bug to remove HoloLens MDL2 Assets (entropy for sure, and it doesn't add anything I assume) and AlternateGothic2 BT (no entropy since no has it, just a cleanup) - I'd ping jfkthame but I know he's always super busy, so can someone in FPP division headquarters please make a decision :) These aren't related to this bug, but would be nice to deal to and it's a simple patch

edited: to show the win stats are firefox user stats

(In reply to Thorin [:thorin] from comment #8)

FWIW it also removed HoloLens MDL2 Assets and there is no easy way to anymore to add it

Perhaps it's going away as part of the HoloLens product being discontinued. It's a bit surprising they'd remove it from Windows, though, given that the devices are still supported.

I wonder if they've merged the icons from there into some other system font? I guess it would be interesting to check whether any of the character codes it supported are not supported by some other available font. If the answer is "none", removal should hopefully be pretty harmless.

FWIW it is listed as being shipped with both Windows 10 and Windows 11, though I don't think we can trust those lists to track all updates with any consistency.

I also would like to know why AlternateGothic2 BT is there - never seen or heard of it

That one puzzles me too. I can only assume it was listed in https://learn.microsoft.com/en-us/typography/fonts/windows_10_font_list back when bug 1619350 was implemented, as I don't recall any other source being used; but I can't find any sign of it now. Unfortunately the Wayback Machine doesn't seem to have a version of the page from back then. Anyhow, it seems fine to delete it.

(In reply to Jonathan Kew [:jfkthame] from comment #12)

I wonder if they've merged the icons from there into some other system font? I guess it would be interesting to check whether any of the character codes it supported are not supported by some other available font. If the answer is "none", removal should hopefully be pretty harmless.

I found this repo from 7 years ago https://github.com/the6th/HoloLens-MDL2_Icons/, in png512 directory, icons seem to match with https://learn.microsoft.com/en-us/windows/apps/design/style/segoe-ui-symbol-font. So mayyyyybe, they might have merged it into this font.

Blocks: 1942875
Depends on: 1942883
No longer blocks: 1942875
Depends on: 1942875

I guess now we have to figure out what to do with UD Digi Kyokasho NK/N/NP fonts right? I'll check the volunteer data we have but I'm not sure it will give anything useful given that you said the update is from 2024-12. I don't know if we any recent data. We'll see.

No one in our volunteer data had any of UD Digi Kyokasho N-B|UD Digi Kyokasho N-R|UD Digi Kyokasho NK-B|UD Digi Kyokasho NK-R|UD Digi Kyokasho NP-B|UD Digi Kyokasho NP-R which I think is all of the fonts. We list them as Japanese supplemental fonts, so I guess we didn't have anyone with Japanese fonts installed.

I don't know what is the process for enabling/disabling fonts with updates. So I'll leave the decision to you and Tom.

I wonder if we would also benefit from font mappings on windows (see Tor Browser's implementation. Our font substitutes don't include anything for fingerprinting protection)

(In reply to Fatih Kilic [:fkilic] from comment #15)

I wonder if we would also benefit from font mappings on windows (see Tor Browser's implementation. Our font substitutes don't include anything for fingerprinting protection)

I mentioned this in https://bugzilla.mozilla.org/show_bug.cgi?id=1941202#c9 - it wouldn't help due to FontFace IIUIC (been doing lots of fontface testing lately)

also see Bug 1845105

You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: