Fix queue worker failure handling and thumbnail fallbacks

queue_jobs.py

@@ -5,6 +5,7 @@
 import concurrent.futures
 import json
 import os
+import shutil
 import signal
 import sqlite3
 import subprocess
@@ -740,6 +741,33 @@ class DownloadQueue:
             if should_flush:
                 self._save_job_locked(job)
 
+    def _cleanup_staging_dir(self, job):
+        staging_dir = job_staging_dir(job)
+        if not staging_dir.exists():
+            return False
+        try:
+            shutil.rmtree(staging_dir)
+            debug_log("queue.cleanup.staging_removed", job_id=job.get("id"), staging_dir=staging_dir)
+            return True
+        except OSError as exc:
+            debug_log("queue.cleanup.staging_failed", job_id=job.get("id"), staging_dir=staging_dir, error=exc)
+            return False
+
+    def _fail_job_startup(self, job, exc):
+        message = f"Could not start download: {exc}"
+        with self.lock:
+            job["status"] = "failed"
+            job["exit_code"] = 1
+            job["pid"] = None
+            job["finished_at"] = now_iso()
+            job["updated_at"] = job["finished_at"]
+            job.setdefault("log", []).append(message)
+            self.current_process = None
+            self.current_job_id = None
+            self.current_job = None
+            self._save_job_locked(job)
+        self._cleanup_staging_dir(job)
+
     def _run(self):
         while not self.stop_event.is_set():
             job = self._next_pending()
@@ -781,16 +809,20 @@ class DownloadQueue:
             job["_dirty_log_lines"] = 0
             self._save_job_locked(job)
 
-        process = subprocess.Popen(
-            command,
-            cwd=str(PROJECT_ROOT),
-            env=env,
-            stdout=subprocess.PIPE,
-            stderr=subprocess.STDOUT,
-            text=True,
-            bufsize=1,
-            start_new_session=True,
-        )
+        try:
+            process = subprocess.Popen(
+                command,
+                cwd=str(PROJECT_ROOT),
+                env=env,
+                stdout=subprocess.PIPE,
+                stderr=subprocess.STDOUT,
+                text=True,
+                bufsize=1,
+                start_new_session=True,
+            )
+        except Exception as exc:
+            self._fail_job_startup(job, exc)
+            return
         with self.lock:
             self.current_process = process
             self.current_job_id = job["id"]
@@ -805,6 +837,7 @@ class DownloadQueue:
         finalize_error = None
         moved_files = []
         watchlist_sync_error = None
+        cleanup_staging = False
         if exit_code == 0 and not job.get("cancel_requested"):
             try:
                 moved_files = finalize_library_files(job)
@@ -841,7 +874,12 @@ class DownloadQueue:
             else:
                 job["status"] = "failed"
                 job.setdefault("log", []).append(f"Download failed with exit code {exit_code}.")
+                cleanup_staging = True
+            if canceled:
+                cleanup_staging = True
             self.current_process = None
             self.current_job_id = None
             self.current_job = None
             self._save_job_locked(job)
+        if cleanup_staging:
+            self._cleanup_staging_dir(job)