The loss of IP address on the client side will cause the program to crash

Nobody has claimed this yet.

After establishing a connection between the MQTT client and the server, if the client's network becomes unstable and the IP is lost, it may cause the program to crash：
/**********************************************************************************/
gdb run sample/async_subscribe.cpp

Connection success

Subscribing to topic 'test/pqq'
for client paho_cpp_async_subscribe using QoS1

Press Q to quit

Subscription success for token: [1]
token topic: 'test/pqq', ...

Program received signal SIGPIPE, Broken pipe.
[Switching to LWP 2998]
0x0000007ff3e4f518 in writev () from /apex/com.android.runtime/lib64/bionic/libc.so
(gdb) bt
#0 0x0000007ff3e4f518 in writev () from /apex/com.android.runtime/lib64/bionic/libc.so
#1 0x000000555557c108 in ?? ()
#2 0x000000555557c604 in ?? ()
#3 0x000000555557f594 in ?? ()
#4 0x000000555557194c in ?? ()
#5 0x0000005555571e3c in ?? ()
#6 0x0000005555570388 in ?? ()
#7 0x00000055555702cc in ?? ()
#8 0x0000005555570154 in ?? ()
#9 0x000000555556e4cc in MQTTAsync_receiveThread ()
#10 0x0000007ff3e63814 in __pthread_start(void*) () from /apex/com.android.runtime/lib64/bionic/libc.so
#11 0x0000007ff3e032f4 in __start_thread () from /apex/com.android.runtime/lib64/bionic/libc.so
(gdb)

Start by reproducing the failure with gdb and sample/async_subscribe.cpp using the connection and subscription steps shown in the report. Inspect the SIGPIPE backtrace from MQTTAsync_receiveThread and determine the relevant client-side failure path. Done means the same network-loss scenario no longer crashes the program.

Written by the indexing model from the issue text.