diff options
| author | Erik Johnston <erik@matrix.org> | 2019-06-10 16:21:42 +0100 |
|---|---|---|
| committer | Erik Johnston <erik@matrix.org> | 2019-06-10 16:22:45 +0100 |
| commit | c413540fb9e4b6ee2ec975a98676ea56d12249c8 (patch) | |
| tree | 7853d4b947451c9708cba12ed1ff722ea31cdbea | |
| parent | Merge branch 'erikj/fix_null_valid_until_ms' of github.com:matrix-org/synapse... (diff) | |
| download | synapse-c413540fb9e4b6ee2ec975a98676ea56d12249c8.tar.xz | |
Fix bug sending federation transactions with lots of EDUs
If we try and send a transaction with lots of EDUs and we run out of space, we call get_new_device_msgs_for_remote with a limit of 0, which then failed.
| -rw-r--r-- | synapse/storage/deviceinbox.py | 4 |
1 files changed, 4 insertions, 0 deletions
diff --git a/synapse/storage/deviceinbox.py b/synapse/storage/deviceinbox.py index 9b0a99cb49..4ea0deea4f 100644 --- a/synapse/storage/deviceinbox.py +++ b/synapse/storage/deviceinbox.py @@ -138,6 +138,10 @@ class DeviceInboxWorkerStore(SQLBaseStore): if not has_changed or last_stream_id == current_stream_id: return defer.succeed(([], current_stream_id)) + if limit <= 0: + # This can happen if we run out of room for EDUs in the transaction. + return defer.succeed(([], last_stream_id)) + def get_new_messages_for_remote_destination_txn(txn): sql = ( "SELECT stream_id, messages_json FROM device_federation_outbox" |