Emanuele Sabellico

This seem to be caused by `rk->rk_rep` being already destroyed (rdkafka.c:954). Could you provide some example code for reproducing this?

Tests didn't run because the branch isn't starting with "dev_" or "feature/". While running locally it gave buffer underflow and disconnections from the broker because of parsing failures. ``` RDUT:...

I think this due to a bug discovered recently, affecting 2.1.0+ versions. When you subscribe a full metadata request is done. This request was emptying the metadata cache and causing...

Could you check latest pushed commit in same PR? I've reproduced and fixed your case, will add a new test too. What happens is that when you call subscribe, the...

> @emasab FYI: I tried librdkafka v2.0.2 & I dont see the bug occurring with that version. Yeah, there were changes with KIP 320 in 2.1.0

It seems like the offset that asked set wasn't available in that partition and it triggered an offset reset

Rebased, fix was already included in 848 so have left the CHANGELOG only and a change to TEST_CONSUMER_GROUP_PROTOCOL until we merge #4662

There seems to be a connectivity issue in your environment, the connection doesn't go beyond the SSL_HANDSHAKE state. Could you enable the "debug" property? Don't know how to pass it...

Hello, does it happen with versions >= 2.1.0 ? There's this fix in that version. > A reference count issue was blocking the consumer from closing. The problem would happen...

@lampt2010 you can probably use `rd_kafka_produceva` . Check a usage example in test [0074-producev](https://github.com/confluentinc/librdkafka/blob/be353be5f6ceaa11a95e21472e981d871ab1aaee/tests/0074-producev.c#L70).