While working on implementing the command monitoring spec tests in C++, spencer.mckenney found one test that was failing: unacknowledgedBulkWrite.json It fails because the reply in the command succeeded event is an empty document, but the spec tests expect a reply of ok:1. According to the spec:

For server versions that do not support write commands, the driver MUST treat an acknowledged write as a single command event, where the GLE command is ignored as a started event and the response to the GLE is treated as the reply in the CommandSucceededEvent. Unacknowledged writes must provide a CommandSucceededEvent with a ok:1 reply.

IIUC this is saying we should always generate an ok:1 reply. But the path for OP_MSG writes doesn't. Instead we fake the ok:1 in our tests.

I think we should actually generate the fake ok:1 for unacknowledged OP_MSG writes. From what I've looked at, it's a straightforward fix.