Fix numbers in Nautilus and GTK apps showing in a calligraphic Arabic font
Held the already-corrected text against 4.0.4-1. 50-omarchy.conf (omarchy-settings 4.0.4-1) still ends with the unconditional Naskh append. Ran the record's Pango check: sans-serif 11 -> Liberation Sans, Adwaita Sans 11 -> Adwaita Sans, so the bug does not reproduce on a stock workstation, consistent with the check-first fix. Tested the override in a scratch XDG_CONFIG_HOME: no warning, and fc-match ':charset=0628' moves from Noto Naskh Arabic to Noto Nastaliq Urdu, exactly the cost the fix and danger describe. One wording defect: the fix says to 'wait for the upstream fix in #8928', but #8928 is an open issue with no fix merged and no linked PR, only a cross-reference from #11002. Fix reissued with that sentence corrected. Status kept as corrected so the earlier audit note is not lost.
Removes Noto Naskh Arabic as Chromium's last-resort fallback in the patterns it matches. Arabic text in Chromium/Electron may fall back to Nastaliq again.
Symptom
In Nautilus and other GTK4/libadwaita apps, digits look wrong. File sizes, dates, counts and progress numbers render in a serif, calligraphic face, while the letters around them are normal sans-serif. GTK Inspector says the font is Adwaita Sans.
Cause
/etc/fonts/conf.d/50-omarchy.conf (owned by omarchy-settings) ends with an unconditional rule that appends Noto Naskh Arabic to every pattern, as a last resort for Chromium's per-character fallback. Its comment assumes charset matching keeps Naskh away from codepoints it does not own. The reporter, on 4.0.1-1 with Nautilus 50.2.2 and locale es_AR.UTF-8, found that Pango's fallback for digits did not behave that way: Noto Naskh Arabic contains Latin digits and won the digit run on sans patterns. On 4.0.4-1 the same Pango check on a stock workstation drew 0 in Liberation Sans, so the bug depends on something in the reporter's setup that has not been identified. Check whether you have it before changing anything.
Fix
First confirm you have the bug. This asks Pango which font draws a digit:
python3 - <<'EOF'
import gi
gi.require_version('Pango','1.0'); gi.require_version('PangoCairo','1.0')
from gi.repository import Pango, PangoCairo
import cairo
for font in ("sans-serif 11", "Adwaita Sans 11"):
s = cairo.ImageSurface(cairo.FORMAT_ARGB32, 200, 80); c = cairo.Context(s)
l = Pango.Layout(PangoCairo.create_context(c)); l.set_text("0", -1)
l.set_font_description(Pango.FontDescription.from_string(font))
PangoCairo.update_layout(c, l)
print(font, "->", l.get_iter().get_run().item.analysis.font.describe().get_family())
EOF
If it prints Liberation Sans or Adwaita Sans, you do not have this bug. Stop here.
If it prints Noto Naskh Arabic, add a user override. It goes in conf.d/, so omarchy font set will not wipe it:
mkdir -p ~/.config/fontconfig/conf.d
cat > ~/.config/fontconfig/conf.d/99-naskh-digits-sans.conf <<'EOF'
<?xml version="1.0"?>
<!DOCTYPE fontconfig SYSTEM "urn:fontconfig:fonts.dtd">
<fontconfig>
<match target="pattern">
<test name="family" compare="eq"><string>Noto Naskh Arabic</string></test>
<edit name="family" mode="prepend" binding="strong"><string>sans-serif</string></edit>
<edit name="family" mode="delete"/>
</match>
</fontconfig>
EOF
fc-cache -f
nautilus -q
The delete edit takes no value. Writing <string>Noto Naskh Arabic</string> inside it, as the issue does, makes fontconfig print Expression doesn't take any effects for delete and delete_all on every call.
Know what this costs. Omarchy appends Naskh to every pattern, so this rule matches every pattern and removes Omarchy's last-resort Arabic fallback everywhere. Afterwards, Arabic text in Chromium and Electron falls back to Noto Nastaliq Urdu (the Urdu style) instead of Naskh:
fc-match ':charset=0628' # Noto Naskh Arabic before, Noto Nastaliq Urdu after
If you read Arabic, do not apply this. Issue #8928 is still open and no upstream fix has been merged as of 4.0.4-1. To undo the override:
rm ~/.config/fontconfig/conf.d/99-naskh-digits-sans.conf && fc-cache -f
Plain Arch: the Naskh rule is Omarchy's, so this does not occur.
Verify
Re-run the Python check above. Both lines should name a sans face, not Noto Naskh Arabic. Then fc-match ':charset=0628' shows which Arabic face Chromium will fall back to.