Question of RdKafka: crash as I call rd_kafka_subscribe

[Benjaad]

Description

this stack like this:
#0 0x000000384d4093a0 in pthread_mutex_lock () from /lib64/libpthread.so.0
#1 0x00002abb6b775389 in mtx_lock (mtx=) at tinycthread.c:135
#2 0x00002abb6b76a11f in rd_kafka_cgrp_handle_SyncGroup (rkcg=0x19e0770, err=RD_KAFKA_RESP_ERR_NO_ERROR, member_state=) at rdkafka_cgrp.c:1984
#3 0x00002abb6b75a189 in rd_kafka_handle_SyncGroup (rk=, rkb=0x19e1580, err=, rkbuf=, request=0x2abbd0000c00, opaque=0x19e0770) at rdkafka_request.c:966
#4 0x00002abb6b754f7c in rd_kafka_buf_callback (rk=0x19e0290, rkb=0x19e1580, err=RD_KAFKA_RESP_ERR_NO_ERROR, response=0x2abc20000da0, request=0x2abbd0000c00) at rdkafka_buf.c:515
#5 0x00002abb6b76adcb in rd_kafka_cgrp_op_serve (rkcg=0x19e0770) at rdkafka_cgrp.c:1664
#6 rd_kafka_cgrp_serve (rkcg=0x19e0770) at rdkafka_cgrp.c:1746
#7 0x00002abb6b73d124 in rd_kafka_thread_main (arg=0x19e0290) at rdkafka.c:974
#8 0x00002abb6b77517f in _thrd_wrapper_function (aArg=) at tinycthread.c:611
#9 0x000000384d4079d1 in start_thread () from /lib64/libpthread.so.0
#10 0x000000384cce88fd in clone () from /lib64/libc.so.6

How to reproduce

Join a new kafka cluster, then subscribed the topic that not created before, finally it will be crashed

Temporary Solution;

I debug it and get the reason：broker return sysc group response without MemberState ,so when read MemberState ,it will call rd_kafka_buf_read_i16 and print the log. But rkbuf->rkbuf_rkb is NULL

code:

---

    rkbuf = rd_kafka_buf_new_shadow(member_state->data,
                                    RD_KAFKAP_BYTES_LEN(member_state));

    rd_kafka_buf_read_i16(rkbuf, &Version);

modify before

---

define rd_kafka_buf_parse_fail(rkbuf,...) do { \

            if (log_decode_errors) {                                \
                    rd_rkb_log(rkbuf->rkbuf_rkb, LOG_WARNING, "PROTOERR", \
                               "Protocol parse failure at %s:%i",   \
                               __FUNCTION__, __LINE__);             \
                    rd_rkb_log(rkbuf->rkbuf_rkb, LOG_WARNING,   \
               "PROTOERR", __VA_ARGS__);        \
            }                                                       \
            goto err;                                               \
} while (0)

---

modify after

---

define rd_kafka_buf_parse_fail(rkbuf,...) do { \

            if (log_decode_errors && rkbuf->rkbuf_rkb) {                                \
                    rd_rkb_log(rkbuf->rkbuf_rkb, LOG_WARNING, "PROTOERR", \
                               "Protocol parse failure at %s:%i",   \
                               __FUNCTION__, __LINE__);             \
                    rd_rkb_log(rkbuf->rkbuf_rkb, LOG_WARNING,   \
               "PROTOERR", __VA_ARGS__);        \
            }                                                       \
            goto err;                                               \
} while (0)

is this Solution suitable？？

[edenhill]

This is fixed on master: 9932f55