Line references are pinned to main at bac2f765.
Pre-existing on main. Not introduced by #367 — that PR split config.rs into submodules and carried both copies over unchanged, so do not bisect to it.
Where
Two implementations of the same legacy migration:
src-tauri/src/config/persistence.rsL44-55, inline inload_config.src-tauri/src/config/library_registry.rsL285-298,migrate_legacy_library_path.
What goes wrong
The helper has no callers. migrate_legacy_library_path is referenced exactly twice repo-wide: its definition (library_registry.rs L285) and the re-export in config/mod.rs L11. Nothing calls it. The migration that actually runs is the inline copy in load_config.
The two copies disagree on the display name. The inline copy falls back to a literal:
// persistence.rs L46-50
let display_name = Path::new(&library_path)
.file_name()
.and_then(|name| name.to_str())
.unwrap_or("OpenKara Library")
.to_owned();
The helper uses library_display_name (library_registry.rs L277-283), which falls back to the full path string instead. For any legacy library_path with no final component — a drive or filesystem root, a path ending in a separator — the two produce different library names.
They also differ on who sets the active id: the helper assigns active_library_id itself (L292-295), while the inline copy leaves that to normalize_for_save (L246-259).
Effect
Dead code that reads as the authoritative migration. Anyone who routes load_config through the helper — the obvious cleanup — silently changes the display name for edge-case paths of users migrating from a pre-registry config.
Suggested resolution (not applied)
Keep one implementation, delete the other, and make the surviving fallback an explicit decision rather than an accident of which copy ran.
Not fixing here
Filed for tracking only.