Panic: 'end byte index 100 is not a char boundary' during search result rendering (src/search/mod.rs:1227)

[tony-nexartis]

Summary

codesearch search panics while rendering results when a matched chunk's context/snippet is truncated at a fixed byte offset that lands inside a multi-byte UTF-8 character.

Environment

• codesearch v1.0.213+1 (macOS arm64 prebuilt binary)
• Repo indexed locally (default settings), TypeScript codebase

Repro

Index a repo containing comment blocks with box-drawing characters (e.g. ─, U+2500, 3 bytes in UTF-8) near byte offset 100 of a chunk, then run a query matching that chunk:

codesearch search "plan universe"

Output

📄 .../src/lib/db/repositories/plan-universe.ts
   Lines 1067-1072 • Comment
   comment block (5 lines)
   ...
thread 'main' (2223939) panicked at src/search/mod.rs:1227:38:
end byte index 100 is not a char boundary; it is inside '─' (bytes 99..102 of string)

The matched file contains ASCII-art comment separators built from U+2500 characters; the panic occurs when the 100-byte truncation boundary falls inside one.

Expected

Truncate on a char boundary (e.g. s.char_indices().take_while(|(i,_)| *i <= 100).last()" or floor_char_boundary`) or on grapheme boundaries.

Notes

• Process exits mid-output (partial results printed, then panic). Exit code observed as 0 in a pipeline, which can mask the failure for scripted consumers.
• Happy to provide more details.

[flupkede]

Fixed in #157 — thanks for the detailed repro, @tony-nexartis! Root cause was byte-slicing UTF-8 at a fixed offset instead of a char-boundary-safe truncation (snippet.floor_char_boundary(100)).