fix(scan): clear stale legacy file columns when a book's folder is gone

[kevinheneveld]

Problem

Audiobooks can show "Downloaded" status with zero files and silently stop being re-searched.

LibraryListService derives file presence from a backward-compat fallback:

var hasTrackedFiles      = files != null && files.Count > 0;          // real AudiobookFiles rows
var hasLegacyFileSummary = !string.IsNullOrWhiteSpace(a.FilePath);   // legacy single-file column
var hasAnyFile = hasTrackedFiles || hasLegacyFileSummary;

So a book with a stale legacy FilePath (and no tracked files) reads as hasAnyFile == true → "Downloaded", while the UI's FileCount (from the real AudiobookFiles rows) is 0. It's also treated as complete by AudiobookWantedEvaluator, so it never re-searches.

Root cause

ScanJobProcessor.ProcessJobAsync has a dedicated branch for when a book's BasePath directory has vanished: it deletes the tracked AudiobookFile rows and sets audiobook.BasePath = null, then returns — but it never clears the legacy FilePath/FileSize columns. The later "legacy filePath migration" block that would clear them is downstream of this early return, so it's skipped.

Result: a book whose whole folder disappeared keeps its legacy columns set → permanent phantom "Downloaded".

Fix

Clear FilePath/FileSize alongside BasePath in the missing-folder branch:

audiobook.BasePath = null;
audiobook.FilePath = null;
audiobook.FileSize = null;
await audiobookRepository.UpdateAsync(audiobook);

The synchronous/manual scan path (LibraryManualScanWorkflow.MigrateLegacyFilePathAsync) already clears these for a missing file; this brings the background worker in line.

Test

Added ProcessJobAsync_BasePathMissing_ClearsLegacyFileColumns: a book with a non-existent BasePath and populated legacy columns is scanned, and FilePath/FileSize are asserted null afterward (read through a fresh DI scope to bypass the test context's identity map). Full suite green (1016 tests).

🤖 Generated with Claude Code