mirror of
https://github.com/element-hq/synapse.git
synced 2026-09-18 16:34:44 +00:00
83193b378d
If we failed to persist an event and then retried, this would fail due to a conflict error trying to reinsert the state groups that we are persisting. This blocks all new events being persisted in the room until the instance is restarted (which clears entries from the table matching the instance name). This is because we always mark the [latest persisted state group in the room as being persisted](https://github.com/element-hq/synapse/blob/781d08df263e640d85ca5c8b5ff514433e019c6f/synapse/storage/databases/state/deletion.py#L242-L243). Also add an `inserted_ts` field which we use to detect stale entries, e.g. from instances that have been taken offline permanently. These aren't a problem beyond blocking deletion of the associated state groups Introduced in https://github.com/element-hq/synapse/pull/18107 --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com> Co-authored-by: Andrew Morgan <1342360+anoadragon453@users.noreply.github.com>