problem: pthread_cond_wait doesn't get signaled by ctrl+c

[somdoron]

Issue description

pthread_cond_wait is not being in interrupted by ctrl+c and I cannot stop an app that waits on a thread safe socket.

Environment

• libzmq version (commit hash if unreleased): master (320741f2c9740832929ec709f18b7ef6f60507c4)
• OS: fedora 29

Minimal test code / Steps to reproduce the issue

#include <zmq.h>
#include <stdio.h>

void my_handler(int s){
    printf("Caught signal %d\n",s);
}
int main() {
    struct sigaction sigIntHandler;

    sigIntHandler.sa_handler = my_handler;
    sigemptyset(&sigIntHandler.sa_mask);
    sigIntHandler.sa_flags = 0;

    sigaction(SIGINT, &sigIntHandler, NULL);

    void *ctx = zmq_ctx_new ();
    void *client = zmq_socket (ctx, ZMQ_CLIENT);
    zmq_msg_t msg;
    zmq_msg_init (&msg);
    zmq_msg_recv (&msg, client, 0);

    return 0;
}

# What's the actual result? (include assertion message & call stack if applicable)

# What's the expected result?
the application should stop after pressing ctrl+c

[sigiesec]

Can you provide an example using libzmq (rather than czmq)?

[somdoron]

@sigiesec changed.

If using dealer instead of client the app do exit.

[somdoron]

A solution might be, to provide a context function that signal all thread safe sockets. So a user can call it when interrupted. CZMQ will probably do it by default.

[bluca]

could the timed version be used instead? so that it can check for interruption

[somdoron]

How can you check for interruption?

[bluca]

can't remember off the top of my head, but we are somehow doing it for the non-safe sockets right?

[somdoron]

Also, it might make a performance problem, exiting and entering a mutex...

[somdoron]

@bluca, no, the behavior of poll/select and pthread_cond_wait is different. poll get interrupted and return EINTR, pthread_cond_wait is not.

[bluca]

ah true, forgot about that

[bluca]

I guess then an API as you suggested is the easiest thing to get it working