eclipse
mosquitto_loop_stop(force=0) hangs and the mosquitto thread blocks in connect() call forever
Mosquitto version: 1.6.15 Tested with latest mosquitto client library: 2.0.12 Platform: Linux
We are using the mosquitto client library with the mosquitto_loop_start/stop mechanism for our implementation. When mosquitto_connect_async is called with the "unreachable broker address" and keepalive=60 secs, The connect fails after 60 secs and disconnect callback is invoked with rc=MOSQ_ERR_KEEPALIVE (19) After some time, when mosquitto_disconnect is called, it fails with MOSQ_ERR_NO_CONN (in v1.6.15) and MOSQ_ERR_SUCCESS(in v2.0.12). The success error code in the latest version, I understand may be due to the bug fix (6ebbb4d)
After this, the mosquitto_loop_stop(force=0) call hangs forever. The backtrace of two threads (application thread which invoked the loop_stop and the mosquitto thread)
mosquitto thread #0 0x00007f109e27253b in __libc_connect (fd=13, addr=..., len=16) at ../sysdeps/unix/sysv/linux/connect.c:26 #1 0x00007f10a270ec60 in net.try_connect () from /home/gogopals/deps-edge-components/lib/libmosquitto.so.1 #2 0x00007f10a270ed97 in net.socket_connect () from /home/gogopals/deps-edge-components/lib/libmosquitto.so.1 #3 0x00007f10a27098bc in mosquitto_reconnect () from /home/gogopals/deps-edge-components/lib/libmosquitto.so.1 --Type for more, q to quit, c to continue without paging-- #4 0x00007f10a270baf8 in mosquitto_loop_forever () from /home/gogopals/deps-edge-components/lib/libmosquitto.so.1 #5 0x00007f10a2713af3 in mosquitto.thread_main () from /home/gogopals/deps-edge-components/lib/libmosquitto.so.1 #6 0x00007f109e267609 in start_thread (arg=) at pthread_create.c:477 #7 0x00007f109dc2c293 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95
Application thread #0 __pthread_clockjoin_ex (threadid=139708726155008, thread_return=0x0, clockid=, abstime=, block=) at pthread_join_common.c:145 #1 0x00007f10a2713be2 in mosquitto_loop_stop () from /home/gogopals/deps-edge-components/lib/libmosquitto.so.1 #2 0x0000559596064ad7 in MosquittoSDK::mosquitto_loop_stop(mosquitto*, bool) const ()
From the above backtrace, I understand the mosquitto thread hangs in socket connect() call forever. I added some logs before and after mosquitto_reconnect in loop.c:mosquitto_loop_forever log__printf(mosq, MOSQ_LOG_DEBUG, "Before reconnect. state=%d", state); rc = mosquitto_reconnect(mosq); log__printf(mosq, MOSQ_LOG_DEBUG, "After reconnect. state=%d", state);
The After reconnect log print is never seen. When the mosquitto_reconnect (blocking) is called the first time, then after that it never unblocks. Unfortunately, we are facing some issues with loop_end(force=1) which uses pthread_cancel. Appreciate your help on how to unblock the mosquitto thread from the connect() call gracefully using any timeout or disconnect option?
Sorry to hear you're having problems. Is there any chance you could provide me with a minimal compilable example that shows how you are using the mosquitto library? I could spend time trying to come up with an example based on your description, but there's a good chance I'd be doing something differently so it would be wasting time rather than trying to find the actual problem.
As an example, I was able to help out with https://github.com/eclipse/mosquitto/issues/2335 quite quickly because of the full code example.
@ralight hello，I encountered the same problem（mosquito lib version 2.0.22）, and I think the problem lies here
The current code returns 0
But 0 won't stop the loop, but mosquitto_loop_stop is waiting thread exits. Since this thread has never ended, the mosquitto_loop_stop interface is blocked
