Fix a bug in the /event_reports Admin API which meant that the total count could be larger than the number of results you can actually query for.

changelog.d/13525.bugfix

@@ -0,0 +1 @@
+Fix a bug in the `/event_reports` Admin API which meant that the total count could be larger than the number of results you can actually query for.

synapse/storage/databases/main/room.py

@@ -2001,9 +2001,15 @@ def _get_event_reports_paginate_txn(
 
             where_clause = "WHERE " + " AND ".join(filters) if len(filters) > 0 else ""
 
+            # We join on room_stats_state despite not using any columns from it
+            # because the join can influence the number of rows returned;
+            # e.g. a room that doesn't have state, maybe because it was deleted.
+            # The query returning the total count should be consistent with
+            # the query returning the results.
             sql = """
                 SELECT COUNT(*) as total_event_reports
                 FROM event_reports AS er
+                JOIN room_stats_state ON room_stats_state.room_id = er.room_id
                 {}
                 """.format(
                 where_clause

tests/rest/admin/test_event_reports.py

@@ -410,6 +410,35 @@ def _check_fields(self, content: List[JsonDict]) -> None:
             self.assertIn("score", c)
             self.assertIn("reason", c)
 
+    def test_count_correct_despite_table_deletions(self) -> None:
+        """
+        Tests that the issue fixed by https://github.com/matrix-org/synapse/pull/13525
+        does not recur:
+        the count should match the number of rows, even if rows in joined tables
+        are missing.
+        """
+
+        # Delete rows from room_stats_state for one of our rooms.
+        self.get_success(
+            self.hs.get_datastores().main.db_pool.simple_delete(
+                "room_stats_state", {"room_id": self.room_id1}, desc="_"
+            )
+        )
+
+        channel = self.make_request(
+            "GET",
+            self.url,
+            access_token=self.admin_user_tok,
+        )
+
+        self.assertEqual(200, channel.code, msg=channel.json_body)
+        # The 'total' field is 10 because only 10 reports will actually
+        # be retrievable since we deleted the rows in the room_stats_state
+        # table.
+        self.assertEqual(channel.json_body["total"], 10)
+        # This is consistent with the number of rows actually returned.
+        self.assertEqual(len(channel.json_body["event_reports"]), 10)
+
 
 class EventReportDetailTestCase(unittest.HomeserverTestCase):
     servlets = [