[zeromq-dev] Problems with 2.0.6 poll using C++ api

[Martin Sustrik]

Matthew,

Yes, it looks like 0MQ believes that the two sockets belong to different 
  threads. Unfortunately, the problem doesn't reproduce on my box.

1. What OS you are running on?

2. Have a look at dispatcher.cpp, lines 146-154. The code there is 
called when socket is created. It checks whether there's another socket 
belonging to the same thread and if so, it chooses to share the same 
app_thread_t object. Something probably goes wrong there. Can you check 
it? Maybe current thread ID as returned by thread_t::id() is different 
even if called twice from the same thread? Maybe thread ID comparison 
function, thread_t::equal() is broken and returns false even if the two 
thread IDs are the same?

Martin

> I'm simply trying to get the simplest possible 'polling' example 
> working. I have two local PUB sockets sending a message every second, on 
> ports 5550, 5551. The recv printout just verifies that I'm able to 
> receive messages, prior to then attempting to poll.
> 
> I actually exported CXXFLAGS="-ggdb3" and re-configured, make, make 
> install. I still get the same error -- here's the latest gdb output:
> 
> 342        for (int i = 0; i != nitems_; i++) {
> (gdb)
> 345            if (items_ [i].socket) {
> (gdb)
> 349                zmq::socket_base_t *s = (zmq::socket_base_t*) items_ 
> [i].socket;
> (gdb)
> 350                if (app_thread) {
> (gdb)
> 351                    if (app_thread != s->get_thread ()) {
> (gdb)
> 352                        free (pollfds);
> (gdb)
> 353                        errno = EFAULT;
> (gdb)
> 354                        return -1;
> (gdb)
> 630    }
> (gdb)
> zmq::poll (items_=0x7fffffffe1e0, nitems_=2, timeout_=-1) at 
> /usr/local/include/zmq.hpp:54
> 54            if (rc < 0)
> (gdb)
> 55                throw error_t ();
> (gdb)
> terminate called after throwing an instance of 'zmq::error_t'
>   what():  Bad address
> 
> Program received signal SIGABRT, Aborted.
> 0x00007ffff5b0f4b5 in raise () from /lib/libc.so.6
> (gdb) bt
> #0  0x00007ffff5b0f4b5 in raise () from /lib/libc.so.6
> #1  0x00007ffff5b12f50 in abort () from /lib/libc.so.6
> #2  0x00007ffff65cacc5 in __gnu_cxx::__verbose_terminate_handler() () 
> from /usr/lib/libstdc++.so.6
> #3  0x00007ffff65c90f6 in ?? () from /usr/lib/libstdc++.so.6
> #4  0x00007ffff65c9123 in std::terminate() () from /usr/lib/libstdc++.so.6
> #5  0x00007ffff65c921e in __cxa_throw () from /usr/lib/libstdc++.so.6
> #6  0x0000000000404536 in zmq::poll (items_=0x7fffffffe1e0, nitems_=2, 
> timeout_=-1) at /usr/local/include/zmq.hpp:55
> #7  0x0000000000404274 in main (argc=1, argv=0x7fffffffe358) at 
> StrategyEngine.cpp:199
> 
> 
> Regards,
> -Matt
> 
> On Thu, Mar 18, 2010 at 4:13 PM, Martin Lucina <mato at kotelna.sk 
> <mailto:mato at kotelna.sk>> wrote:
> 
>     Hi Matthew,
> 
>     mgiedt at gmail.com <mailto:mgiedt at gmail.com> said:
>      > Here's the application that I'm using to test:
>      >
>      > [snip]
> 
>     Looks fine though I'm not sure what you're trying to achieve...
> 
>      > This produces:
>      >
>      >  sock1 recv: 0xfce9e8
>      >  sock2 recv: 0xfcece8
>      > poll...
>      > terminate called after throwing and instance of 'zmq::error_t'
>      >   what(): Bad address
>      > Aborted
>      >
>      > I've included the gdb output below -- it's thinking that the two
>     sockets were
>      > created from different threads?
> 
>     That's what it looks like, but the trace you posted looks like it's
>     bogus
>     due to optimization. Can you try doing
> 
>     $ export CXXFLAGS="-O0 -g"
> 
>     Then rebuild and reinstall, and see if you a) still get the problem
>     b) the
>     backtrace looks sane?
> 
>     Thanks
> 
>     -mato
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> zeromq-dev mailing list
> zeromq-dev at lists.zeromq.org
> http://lists.zeromq.org/mailman/listinfo/zeromq-dev