fix: limit atomic-write fallback to atomic-unsupported errnos

app/state_store.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import base64
 import collections.abc
+import errno
 import json
 import logging
 import os
@@ -17,6 +18,25 @@ STATE_SCHEMA_VERSION = 2
 _BYTES_MARKER = "__metube_bytes__"
 _DATETIME_MARKER = "__metube_datetime__"
 
+# Errnos that signal the filesystem cannot support the temp-file + rename
+# atomic-write strategy (for example an NFS-backed state dir returning EPERM on
+# mkstemp). These are safe to fall back on because they mean the atomic
+# mechanism is unavailable, not that the data write itself failed. Errors like
+# ENOSPC/EIO are deliberately excluded so a genuine storage failure surfaces
+# instead of silently truncating an existing good state file.
+_ATOMIC_UNSUPPORTED_ERRNOS = frozenset(
+    e
+    for e in (
+        errno.EPERM,
+        errno.EACCES,
+        errno.ENOSYS,
+        errno.EINVAL,
+        getattr(errno, "EOPNOTSUPP", None),
+        getattr(errno, "ENOTSUP", None),
+    )
+    if e is not None
+)
+
 
 def to_json_compatible(value: Any) -> Any:
     if value is None or isinstance(value, (bool, int, float, str)):
@@ -100,6 +120,8 @@ class AtomicJsonStore:
         try:
             self._atomic_write(payload)
         except OSError as exc:
+            if exc.errno not in _ATOMIC_UNSUPPORTED_ERRNOS:
+                raise
             self._warn_direct_write_fallback(exc)
             self._direct_write(payload)
 

app/tests/test_state_store.py

@@ -71,6 +71,28 @@ class StateStoreTests(unittest.TestCase):
             self.assertEqual(ctx.exception.errno, 13)
             self.assertFalse(os.path.exists(path))
 
+    def test_save_reraises_and_preserves_state_on_non_atomic_errno(self):
+        # A storage failure such as ENOSPC is not an "atomic unavailable"
+        # signal, so it must surface instead of falling back to a direct write
+        # that would truncate the existing good state file.
+        import errno as _errno
+
+        with tempfile.TemporaryDirectory() as tmp:
+            path = os.path.join(tmp, "queue.json")
+            store = AtomicJsonStore(path, kind="persistent_queue:queue")
+            store.save({"items": [{"key": "good"}]})
+
+            with patch(
+                "state_store.tempfile.mkstemp",
+                side_effect=OSError(_errno.ENOSPC, "No space left on device"),
+            ):
+                with self.assertRaises(OSError) as ctx:
+                    store.save({"items": [{"key": "new"}]})
+
+            self.assertEqual(ctx.exception.errno, _errno.ENOSPC)
+            # Existing state is untouched.
+            self.assertEqual(store.load()["items"], [{"key": "good"}])
+
     def test_invalid_file_is_quarantined(self):
         with tempfile.TemporaryDirectory() as tmp:
             path = os.path.join(tmp, "queue.json")