Summary
On one user's Mac, the tab-strip chrome icons (the + new-tab, ? help, and ⚙ settings glyphs) render on the left side of a tab instead of at their normal right-aligned positions. The click targets are unaffected: clicking where the icons should be still works. So the quad/hit-box math is right and only the shaped glyph positions are off.
Second data point: another Mac (macOS 26.3) renders them correctly.
Suspected cause
Those chrome glyphs (+ U+FF0B, ?/?, ⚙ U+2699) are not covered by every monospace font, so they shape through cosmic-text's font fallback. On a different macOS release (or a Mac with different fonts installed), fallback can resolve to a different font with different advances, shifting where glyphs land in the shaped run, while our hit-boxes are computed independently from cell metrics. A glyph-cluster/advance mismatch in the tab-strip's shaped text would produce exactly "drawn in the wrong place, clickable in the right place".
Needed to reproduce
Where to look
- Tab-strip shaping and the icon hit-boxes are computed in separate places by design; the fix likely pins the chrome glyphs to explicit sprite/atlas rendering (the box-drawing path already does this for exactly this reason) or measures the shaped run instead of assuming cell-width multiples.
Summary
On one user's Mac, the tab-strip chrome icons (the + new-tab, ? help, and ⚙ settings glyphs) render on the left side of a tab instead of at their normal right-aligned positions. The click targets are unaffected: clicking where the icons should be still works. So the quad/hit-box math is right and only the shaped glyph positions are off.
Second data point: another Mac (macOS 26.3) renders them correctly.
Suspected cause
Those chrome glyphs (
+U+FF0B,?/?,⚙U+2699) are not covered by every monospace font, so they shape through cosmic-text's font fallback. On a different macOS release (or a Mac with different fonts installed), fallback can resolve to a different font with different advances, shifting where glyphs land in the shaped run, while our hit-boxes are computed independently from cell metrics. A glyph-cluster/advance mismatch in the tab-strip's shaped text would produce exactly "drawn in the wrong place, clickable in the right place".Needed to reproduce
EMBER_FONT_DEBUG=1 /Applications/Ember.app/Contents/MacOS/ember-term 2>font.logand attach the log (it records which fonts cosmic-text actually resolved)Where to look