MSC4140: delayed event content as text, not bytes (#19360)

Store the JSON content of scheduled delayed events as text instead of a
byte array. This brings it in line with the `event_json` table's `json`
column, and fixes the inability to schedule a delayed event with
non-ASCII characters in its content.

Fixes #19242

changelog.d/19360.bugfix

@@ -0,0 +1 @@
+MSC4140: Store the JSON content of scheduled delayed events as text instead of a byte array. This fixes the inability to schedule a delayed event with non-ASCII characters in its content.

scripts-dev/check_schema_delta.py

@@ -187,6 +187,14 @@ def check_schema_delta(delta_files: list[str], force_colors: bool) -> bool:
         sql_lang = "postgres"
         if delta_file.endswith(".sqlite"):
             sql_lang = "sqlite"
+        elif delta_file.endswith(".py"):
+            click.secho(
+                f"Skipping Python delta file: '{delta_file}'",
+                fg="yellow",
+                bold=True,
+                color=force_colors,
+            )
+            return True
 
         statements = sqlglot.parse(delta_contents, read=sql_lang)
 

synapse/storage/schema/main/delta/93/04_make_delayed_event_content_text.py

@@ -0,0 +1,62 @@
+#
+# This file is licensed under the Affero General Public License (AGPL) version 3.
+#
+# Copyright (C) 2026 Element Creations Ltd
+#
+# This program is free software: you can redistribute it and/or modify
+# it under the terms of the GNU Affero General Public License as
+# published by the Free Software Foundation, either version 3 of the
+# License, or (at your option) any later version.
+#
+# See the GNU Affero General Public License for more details:
+# <https://www.gnu.org/licenses/agpl-3.0.html>.
+#
+
+import logging
+
+from synapse.storage.database import LoggingTransaction
+from synapse.storage.engines import BaseDatabaseEngine, PostgresEngine
+
+logger = logging.getLogger(__name__)
+
+
+def run_create(cur: LoggingTransaction, database_engine: BaseDatabaseEngine) -> None:
+    """
+    Change the type / affinity of the `delayed_events` table's `content` column from bytes to text.
+    This brings it in line with the `event_json` table's `json` column, and fixes the inability to
+    schedule a delayed event with non-ASCII characters in its content.
+    """
+
+    if isinstance(database_engine, PostgresEngine):
+        cur.execute(
+            "ALTER TABLE delayed_events "
+            "ALTER COLUMN content SET DATA TYPE TEXT "
+            "USING convert_from(content, 'utf8')"
+        )
+        return
+
+    # For sqlite3, change the type affinity by fiddling with the table schema directly.
+    # This strategy is also used by ../50/make_event_content_nullable.py.
+
+    cur.execute(
+        "SELECT sql FROM sqlite_master WHERE tbl_name='delayed_events' AND type='table'"
+    )
+    row = cur.fetchone()
+    assert row is not None
+    (oldsql,) = row
+
+    sql = oldsql.replace("content bytea", "content TEXT")
+    if sql == oldsql:
+        raise Exception("Couldn't find content bytes column in %s" % oldsql)
+
+    cur.execute("PRAGMA schema_version")
+    row = cur.fetchone()
+    assert row is not None
+    (oldver,) = row
+    cur.execute("PRAGMA writable_schema=ON")
+    cur.execute(
+        "UPDATE sqlite_master SET sql=? WHERE tbl_name='delayed_events' AND type='table'",
+        (sql,),
+    )
+    cur.execute("PRAGMA schema_version=%i" % (oldver + 1,))
+    cur.execute("PRAGMA writable_schema=OFF")

tests/rest/client/test_delayed_events.py

@@ -278,17 +278,24 @@ class DelayedEventsTestCase(HomeserverTestCase):
         channel = self._update_delayed_event(delay_ids.pop(0), "cancel", action_in_path)
         self.assertEqual(HTTPStatus.TOO_MANY_REQUESTS, channel.code, channel.result)
 
-    @parameterized.expand((True, False))
-    def test_send_delayed_state_event(self, action_in_path: bool) -> None:
+    @parameterized.expand(
+        (
+            (content_property_value, action_in_path)
+            for content_property_value in ("test", "tест")
+            for action_in_path in (True, False)
+        )
+    )
+    def test_send_delayed_state_event(
+        self, content_value: str, action_in_path: bool
+    ) -> None:
         state_key = "to_send_on_request"
 
-        setter_key = "setter"
-        setter_expected = "on_send"
+        content_property_name = "key"
         channel = self.make_request(
             "PUT",
             _get_path_for_delayed_state(self.room_id, _EVENT_TYPE, state_key, 100000),
             {
-                setter_key: setter_expected,
+                content_property_name: content_value,
             },
             self.user1_access_token,
         )
@@ -300,7 +307,7 @@ class DelayedEventsTestCase(HomeserverTestCase):
         events = self._get_delayed_events()
         self.assertEqual(1, len(events), events)
         content = self._get_delayed_event_content(events[0])
-        self.assertEqual(setter_expected, content.get(setter_key), content)
+        self.assertEqual(content_value, content.get(content_property_name), content)
         self.helper.get_state(
             self.room_id,
             _EVENT_TYPE,
@@ -318,7 +325,7 @@ class DelayedEventsTestCase(HomeserverTestCase):
             self.user1_access_token,
             state_key=state_key,
         )
-        self.assertEqual(setter_expected, content.get(setter_key), content)
+        self.assertEqual(content_value, content.get(content_property_name), content)
 
     @parameterized.expand((True, False))
     @unittest.override_config({"rc_message": {"per_second": 2.5, "burst_count": 3}})