[zeromq-dev] publish/subscribe hangs in libzmq

[Toralf Wittner]

Please consider the following program:

#include <zmq.hpp>
#include <iostream>

int main() {
    zmq::context_t ctx(0);

    zmq::socket_t pub(ctx, ZMQ_PUB);
    zmq::socket_t sub(ctx, ZMQ_SUB);

    sub.setsockopt(ZMQ_SUBSCRIBE, "", 0);
    sub.bind("inproc://endpoint");
    pub.connect("inproc://endpoint");

    char const* data = "hello";
    zmq::message_t hi(const_cast<char*>(data), 5, NULL);
    pub.send(hi);

    zmq::message_t msg;
    sub.recv(&msg);

    std::cout << static_cast<char*>(msg.data()) << std::endl;

    return 0;
}

Running against 0MQ 2.1.11 yields the expected result:

$ g++ -g -ggdb -O0 -Wall -o testcase-cpp -I/opt/zeromq2-1/include/
-L/opt/zeromq2-1/lib  testcase.cc -lzmq
$ ./testcase-cpp
hello

Compiling and running against libzmq (current HEAD
6f32361fea61619fec94348de693a9e3ff8981e0) will not terminate. Here the
stacktraces:

$ g++ -g -ggdb -O0 -Wall -o testcase-cpp -I/opt/zeromq3-1/include/
-L/opt/zeromq3-1/lib  testcase.cc -lzmq
$ gdb testcase-cpp
GNU gdb (GDB) 7.3.1
Copyright (C) 2011 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
and "show warranty" for details.
This GDB was configured as "x86_64-unknown-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
Reading symbols from /home/user/workspaces/zeromq/testcase-cpp...done.
(gdb) run
Starting program: /home/user/workspaces/zeromq/testcase-cpp
[Thread debugging using libthread_db enabled]
[New Thread 0x7ffff6c5f700 (LWP 1630)]
^C
Program received signal SIGINT, Interrupt.
0x00007ffff7151173 in poll () from /lib/libc.so.6
(gdb) thread apply all bt

Thread 2 (Thread 0x7ffff6c5f700 (LWP 2053)):
#0  0x00007ffff71599c3 in epoll_wait () from /lib/libc.so.6
#1  0x00007ffff7bba49f in zmq::epoll_t::loop (this=0x604610) at epoll.cpp:142
#2  0x00007ffff7bca1b6 in thread_routine (arg_=0x604680) at thread.cpp:74
#3  0x00007ffff6c67df0 in start_thread () from /lib/libpthread.so.0
#4  0x00007ffff71593ed in clone () from /lib/libc.so.6
#5  0x0000000000000000 in ?? ()

Thread 1 (Thread 0x7ffff7fca720 (LWP 2050)):
#0  0x00007ffff7151173 in poll () from /lib/libc.so.6
#1  0x00007ffff7bc5361 in zmq::signaler_t::wait (this=<optimized out>,
timeout_=<optimized out>) at signaler.cpp:145
#2  0x00007ffff7bbd33a in zmq::mailbox_t::recv (this=0x606c90,
cmd_=0x7fffffffe220, timeout_=<optimized out>) at mailbox.cpp:69
#3  0x00007ffff7bc64cd in zmq::socket_base_t::process_commands
(this=0x606aa0, timeout_=<optimized out>, throttle_=false) at
socket_base.cpp:663
#4  0x00007ffff7bc6c55 in zmq::socket_base_t::recv (this=0x606aa0,
msg_=0x7fffffffe350, flags_=0) at socket_base.cpp:600
#5  0x00007ffff7bcec58 in zmq_recvmsg (s_=0x606aa0,
msg_=0x7fffffffe350, flags_=0) at zmq.cpp:296
#6  0x0000000000401b0b in zmq::socket_t::recv (this=0x7fffffffe370,
msg_=0x7fffffffe350, flags_=0) at /opt/zeromq3-1/include/zmq.hpp:345
#7  0x000000000040134f in main () at testcase.cc:19

I did not look too closely but it seems that no subscriptions match in
zmq::xpub_t::xsend, i.e. no pipe is marked as matching and
dist.send_to_matching will not send out anything (with
dist.send_to_all this program works just like in 0MQ 2.1.11).

T.