Highlighters emit one span per byte of a non-ASCII character, and the
ImGui and Qt renderers decoded each span on its own, so every such
character was drawn as replacement glyphs (reproduced under Xvfb:
"héllo" showed as "h??llo"). kte::SanitizeSpans (Highlight.h) widens
spans to whole UTF-8 characters and removes overlaps (the earlier span
wins), and replaces the three renderers' copies of the clamp/sort code.
The terminal renderer's token lookup rescanned the spans from the first
for every byte drawn; it now walks them with a cursor.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015rcVQiiWyTzoNrsvLSaAfB
Claude
c5df5035fa