Segfault in NativeClient::rebalance_protocol

[bruceg]

We are seeing a sporadic segfault when handling rebalance events in our kafka client. I have traced this down to a null pointer being passed to CStr::from_ptr. It appears that rdsys::rd_kafka_rebalance_protocol can return a null pointer in some circumstances and the Rust library doesn't handle that. I have worked around it by returning RebalanceProtocol::None when this happens, but I don't know if there is a better solution or a deeper underlying problem.

[benesch]

Oy, yeah, looks like that's how it signals errors. None seems about right. The C example that I copied the incremental stuff from passes the return value to strcmp, which ignores any distinction between NULL and "NONE": https://github.com/edenhill/librdkafka/blob/f092c290995ca81b3afb4015fcc3350ba02caa96/examples/rdkafka_complex_consumer_example.c#L218

[benesch]

Would you be up for submitting a PR?

[bruceg]

Sure, coming right up.