fix: tell a local DNS outage from a gone host

scripts/upstream.py

@@ -208,12 +208,29 @@ def _host_is_unresolvable(exc: BaseException) -> bool:
         if seen is None:
             break
         if isinstance(seen, socket.gaierror):
-            return True
+            # A resolver that answers nothing at all is the local network:
+            # every profile would read as gone, and the sweep would report
+            # nothing to review while it reviewed nothing.
+            return seen.errno == socket.EAI_NONAME and _resolver_answers()
         nested = getattr(seen, "reason", None)
         seen = nested if isinstance(nested, BaseException) else seen.__cause__
     return False
 
 
+# A name every forge call depends on: if it does not resolve, the fault is
+# local and no host can be declared gone.
+RESOLVER_PROBE = "github.com"
+
+
+def _resolver_answers() -> bool:
+    """Whether name resolution works here at all. Replaced in tests."""
+    try:
+        socket.getaddrinfo(RESOLVER_PROBE, 443)
+    except OSError:
+        return False
+    return True
+
+
 def _fetch(url: str, accept_json: bool = False) -> bytes | None:
     """Body of a GET, or None on 404.
 

tests/test_upstream.py

@@ -9,6 +9,7 @@ import sys
 import tempfile
 import socket
 import unittest
+from unittest import mock
 import urllib.error
 import urllib.request
 from pathlib import Path
@@ -291,14 +292,26 @@ class TestGoneUpstream(unittest.TestCase):
         self.assertEqual(self.calls, 1)
 
     def test_unresolvable_host_is_gone_and_not_retried(self):
-        self._serve(
-            urllib.error.URLError(socket.gaierror(-2, "Name or service not known"))
-        )
-        with self.assertRaises(upstream.GoneError):
-            upstream._fetch("https://host/x")
+        self._serve(urllib.error.URLError(
+            socket.gaierror(socket.EAI_NONAME, "Name or service not known")
+        ))
+        with mock.patch.object(upstream, "_resolver_answers", return_value=True):
+            with self.assertRaises(upstream.GoneError):
+                upstream._fetch("https://host/x")
         self.assertEqual(self.calls, 1)
         self.assertEqual(self.slept, [])
 
+    def test_a_local_resolver_outage_is_not_gone(self):
+        """With DNS down here, every forge looked gone and profile_sync
+        listed nothing to review, exiting 0."""
+        for errno, resolver in ((socket.EAI_NONAME, False), (socket.EAI_AGAIN, True)):
+            with self.subTest(errno=errno, resolver=resolver):
+                self._serve(urllib.error.URLError(socket.gaierror(errno, "no")))
+                with mock.patch.object(upstream, "_resolver_answers", return_value=resolver):
+                    with self.assertRaises(upstream.UpstreamError) as caught:
+                        upstream._fetch("https://host/x")
+                self.assertNotIsInstance(caught.exception, upstream.GoneError)
+
     def test_a_gone_error_is_still_an_upstream_error(self):
         self._serve(_http_error(451))
         with self.assertRaises(upstream.UpstreamError):