[ddehilster]

Summary

Fixes #488: deaccent() mangles UTF-8 accented text — e.g. deaccent("fiancé") returned fiancA + garbage instead of fiance.

Root cause

Xml::de_accent walked the string byte-by-byte and switched on single bytes \xC0..\xFF — i.e. it assumed Latin-1. On the engine's UTF-8 text, é is two bytes (0xC3 0xA9): 0xC3 matched "Atilde" → A, and 0xA9 fell through the switch untouched. Hence fiancA<0xA9>.

Fix

Rewrite de_accent to walk UTF-8 codepoints (U8_NEXT) and map each accented codepoint to its ASCII base:

• The old byte table's values (0xC0..0xFF) are exactly the Latin-1 Supplement codepoints U+00C0..U+00FF, so the mapping carries over unchanged.
• The 0x8A..0x9F Windows-1252 letters map to their real Unicode codepoints (Scaron U+0160, OElig U+0152, …).
• ASCII bytes are copied unchanged → de-accent matching on ASCII is byte-identical to before (no regression); unmapped multi-byte codepoints are preserved verbatim.
• Replaces the byte-based de_accent_ch (removed; it was used only by these loops).

Verified

Input	Output
fiancé	fiance ✅
café / naïve	cafe / naive ✅
señor / Zürich / garçon	senor / Zurich / garcon ✅
Œuvre	Euvre ✅
hello (ASCII)	hello (unchanged) ✅

emailaddress analyzer still loads/extracts — no regression. de_accent is also used by rule-matching (Arun.cpp); since ASCII output is unchanged, existing matching is unaffected and UTF-8 matching now de-accents correctly.

⚠️ Merge order: bumps → 3.7.5, assuming #688 (→3.7.4) merges first. If merged before #688, keep 3.7.5 when resolving the trivial main.cpp version-line conflict.

Fixes #488.

🤖 Generated with Claude Code