TBA13: system/widget font tests: font-family not returned in getComputedStyle
getComputedStyle(el).getPropertyValue("font-family") is (mostly) missing in Android TB13
system fonts test
for example, I get all 17 items tested as
- windows:
"12px normal 400 sans-serif" - linux:
"15px normal 400 sans-serif" - android:
"16px normal 400 "(note the trailing space, because I didn't expect to have to parse for empty strings :))
widget fonts test
for example
windows:
"monospace 13.3333px": ["datetime", "datetime-local", "time"],
"monospace 13px": ["textarea"],
"sans-serif 13.3333px": [19 items],
"sans-serif 13px": ["image"]linux:
"monospace 12px": ["textarea"],
"monospace 13.3333px": ["datetime", "datetime-local", "time"],
"sans-serif 13.3333px": [19 items],
"sans-serif 13px": ["image"]android should be same as linux, but the 19 items is missing the font-family property: note the leading space
" 13.3333px": [19 items],
"monospace 12px": ["textarea"],
"monospace 13.3333px": ["datetime", "datetime-local", "time"],
"sans-serif 13px": ["image"]
cc @pierov
Designs
- tor-browser-build #41113
Activity
added Next label
added Bug label
added 13.5 stable label
@thorin does this differ from what we have in TBA 12.5?
can we fix this please - see https://arkenfox.github.io/TZP/tests/fontsystem.html
namely serif differs in another (secret) test and gives me unexpected results - if you look at
SYSTEM- the caption, icon etc has no font-family, where we would normally (in FF) expect serif, and it is serif that differs in my secret testedit: and in the
WIDGETStest it is only the13.3333pxlot that is missing the font-family (in Firefox this is reported asRobotowhich is a monospace font, so not sure if related)note: my secret test has nothing to do with widgets and everything to do with measuring elements
ping! @pierov
Edited by Thorin-
it's a sans-serif.
my bad, I lack debugging on android so I just googled it - stupid google first result (probably AI canned BS)
I think the easiest course of action is to ignore my patch in Android
I don't think so. Only Android 9+ has Roboto so there would still be entropy - but IDK the answer to that, because widgets still have to use the right font.
I am kinda lost on all this now. We set a system-ui font, we then set font-families to serif, sans-serif or -apple-system to ensure elements returned the correct/same getComputedStyle and actually used it (so measurements would be uniform, all things being equal e.g. zoom, subpixels etc)
So to abandon the patch on android just leaves the problem unsolved
side note: ignore my secret test, I can actually coax differences in desktop as well, and it doesn't help that android is scaled (e.g. dpi 2.26987 or whatever). But it also doesn't help when we have this issue unsolved
Only Android 9+ has Roboto
Citation needed?
Wikipedia says Android 4.0+, and it was redesigned for Android 5.0 (the minimum version we support).
IIRC, Emilio wanted me to hardcode to Roboto for Android, instead of sans-serif.
FWIW, if upstream ends up refusing our patch definitely, we can even start hardcoding precise fonts for our platforms (Segoe UI for Windows, Arimo for Linux, Roboto for Android, -apple-system for macOS).
-
from Tom's telemetry
https://bugzilla.mozilla.org/show_bug.cgi?id=1826412
added in 5
Roboto Black Italic Roboto Black Roboto Bold Italic Roboto Bold Roboto Italic Roboto Light Italic Roboto Light Roboto Medium Italic Roboto Medium Roboto Thin Italic Roboto Thinadded in 9
Robotoedit: also see https://github.com/arkenfox/TZP/blob/12871c5c7a0aafcc7bbbcca7a687e3b27c8e6f5b/js/fonts.js#L450-L455
} else if ("android" == os) { // see android list notes: Roboto is not guaranteed unless Android 9+ // ToDo: we should test a mini set of android fonts and use a detected one if (!isFontSizesMore) {isPlatformFont = "Dancing Script"}Edited by Thorin
removed 13.5 stable label
added 13.0 stable label
removed 13.0 stable label
added 13.5 stable label
removed Backlog label
added Roadmap::Future label
removed FF115-esr label
added Fingerprinting label
-
so following on from #41646 (comment 3010551)
In TZP and in the PoC, I split fonts into
moz,systemandwidgetwindows TB pic so know what I'm talking about
It is the last two that require our patches because otherwise it reveals the font name, but worse, the computedStyle can lie and different fonts will/can render different sized elements - hence why we set it to a generic font-family which we control in font prefs
The good news is that in windows and linux, for both
systemandwidgetwork across all app languages flawlessly and we have a health checksystem
WINDOWS "12px normal 400 sans-serif": [the lot] MAC "11px normal 400 -apple-system": ["message-box", "status-bar"], "11px normal 700 -apple-system": ["small-caption"], "12px normal 400 -apple-system": ["icon"], "13px normal 400 -apple-system": ["caption", "menu"] LINUX "15px normal 400 sans-serif": [the lot]widgets
WINDOWS "monospace 13.3333px": ["date", "datetime-local", "time"], "monospace 13px": ["textarea"], "sans-serif 13.3333px": [19 items], "sans-serif 13px": ["image"] MAC "-apple-system 13.3333px": [19 items], "monospace 13.3333px": ["date", "datetime-local", "time"], "monospace 13px": ["textarea"], "sans-serif 13px": ["image"] LINUX "monospace 12px": ["textarea"], "monospace 13.3333px": ["date", "datetime-local", "time"], "sans-serif 13.3333px": [19 items], "sans-serif 13px": ["image"]more to follow, need a break
Edited by Thorin -
The problem on android is the generic font-family computedStyle property is missing/blank - so how do we know it's being applied by the patch (well we could guess it is due to the missing property) and we can still check the sizes and weights - but I do not like this!! If we patch something it must be reliably measurable and added as a health check and I currently cannot do that
SYSTEM
FF "12px normal 400 Roboto": [6 items] TBA "16px normal 400 ": [6 items]The difference is the size and we are missing the generic font-family. IDK if this breaks things.
WIDGETS
FF "Roboto 13.3333px": [19 items], "monospace 12px": ["textarea"], "monospace 13.3333px": ["date", "datetime-local", "time"], "sans-serif 13px": ["image"] TBA " 13.3333px": [19 items], "monospace 12px": ["textarea"], "monospace 13.3333px": ["date", "datetime-local", "time"], "sans-serif 13px": ["image"]The difference,
besides the order of keys, is the[19 items]is missing the generic font-family property
Questions
some history
- Roboto is not on android phones by default until version 9
- FYI: the following are in version 5+: Roboto Black Italic, Roboto Black. Roboto Bold Italic, Roboto Bold, Roboto Italic, Roboto Light Italic, Roboto Light, Roboto Medium Italic, Roboto Medium, Roboto Thin Italic, Roboto Thin
- IDK what font is used prior to v9 for a) these 19 items [input/widgets] and b) these six items ["caption","icon","menu","message-box","small-caption","status-bar"]
- I think we support android 5 + higher, correct me if I'm wrong
questions
- we change the size in system (6 items) but not in widgets (19 items). Are we setting a different generic font-family, if not then why the size diff.
what to do?
- fix it so we return the computed style so we can do the health check
- and something something about sizes and emilio and upstream
- or remove it and leave android unprotected, relying on the fact that likely, everyone in 9+ is roboto and everyone < 9 is the same
class, discuss
cc: @dan if you can help with any old androids
Edited by Thorin - Roboto is not on android phones by default until version 9
-
I think we support android 5 + higher, correct me if I'm wrong
Yes. I don't know if Mozilla changed that after 115.
What is sans-serif on Android? Is it Roboto for > 9 and maybe Droid Sans in < 9? In case maybe we could set sans-serif just like Linux and Windows.
we change the size in system (6 items) but not in widgets (19 items). Are we setting a different generic font-family, if not then why the size diff.
We don't change widgets directly. I don't know where widget code is and what it does.
-
https://support.mozilla.org/en-US/kb/will-firefox-work-my-mobile-device ... still says 5.0 .. it's take a long time until 9.0 is the minimum
-
FYI: https://bugzilla.mozilla.org/show_bug.cgi?id=1894429#c1
summary by me
FF android users = Android 5-9 = 46%
https://gs.statcounter.com/android-version-market-share/mobile/worldwide/
March 2024 14 16.5 13 26.5 12 17.1 11 16.1 10 9.0 9 5.4 ----- 81.6% = 90.56% of supported versions [1] 8.1 1.9 8 2.9 7 1.4 6 1.3 5.1 0.7 5 0.3 ----- 8.5% = 9.43% of supported versions [1] [1] only 90.1% of reported versions are gecko supportedso yikes .. somewhere between 9% and 46% of users do not have Roboto (by default), give or take. I know, it's not very helpful, but interesting (and I have no idea what the FF sample size is)
marked this issue as a duplicate of #42529 (closed)
marked this issue as related to #42529 (closed)
assigned to @pierov
marked #42529 (closed) as a duplicate of this issue
removed Roadmap::Future label
added Doing label
mentioned in merge request !993 (merged)
marked this issue as related to tor-browser-build#41113 (closed)
mentioned in issue #42529 (closed)
removed Doing label
