zmq_poll does not trigger ZMQ_POLLIN event in some successful connections #3161
Description
Issue description
I use ZMQ_DEALER connect to ZMQ_ROUTER, and listening to events with zmq_poll, but when a client reconnect, the zmq_poll may be not trigger ZMQ_POLLIN event.
Environment
- libzmq version (commit hash if unreleased):
4.2.3 and 4.2.5 - OS:
Ubuntu 16.04 and Windows 7
Minimal test code / Steps to reproduce the issue
Running these code, the issue will occur in less than 5 minitues.
server
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "zmq.h"
#ifdef WIN32
#include <winsock2.h>
#include <ws2tcpip.h>
#endif
int main()
{
int linger = 1000;
int ret = 0;
void *ctx = zmq_ctx_new();
void *conn = zmq_socket(ctx, ZMQ_ROUTER);
zmq_setsockopt(conn, ZMQ_LINGER, &linger, sizeof(linger));
ret = zmq_bind(conn, "tcp://*:1026");
if (-1 == ret) {
printf("bind failed, errno:%d\n", errno);
exit(-1);
}
zmq_pollitem_t items[1] = { 0 };
items[0].socket = conn;
items[0].events = ZMQ_POLLIN;
while (1) {
int rc = 0;
int more = 0;
size_t more_size = sizeof(more);
rc = zmq_poll(items, 1, 100);
if (rc <= 0) {
if (-1 == rc) {
printf("zmq_poll failed errno:%d error:%s\n", errno, zmq_strerror(errno));
}
continue;
}
if (items[0].revents & ZMQ_POLLIN) {
char ident[128] = { 0 };
char data[1024] = { 0 };
int rc = zmq_recv(conn, ident, sizeof(ident), ZMQ_DONTWAIT);
if (-1 == rc) {
printf("recv failed errno:%d, error:%s\n", errno, zmq_strerror(errno));
continue;
}
printf("recv ident:%s\n", ident);
zmq_getsockopt(conn, ZMQ_RCVMORE, &more, &more_size);
if (!more) {
printf("there is no more data\n");
continue;
}
zmq_recv(conn, data, sizeof(data), 0);
printf("recv data:%s\n", data);
zmq_send(conn, ident, strlen(ident), ZMQ_SNDMORE);
zmq_send(conn, data, strlen(data) + 1, 0);
}
}
zmq_close(conn);
zmq_ctx_term(ctx);
return 0;
}
client
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <time.h>
#include <unistd.h>
#include "zmq.h"
#ifdef WIN32
#include <winsock2.h>
#include <ws2tcpip.h>
#endif
int main()
{
char *ident = "123456-hello";
int linger = 10000;
int ret = 0;
char addr[] = "tcp://127.0.0.1:1026";
void *ctx = zmq_ctx_new();
void *conn = zmq_socket(ctx, ZMQ_DEALER);
zmq_setsockopt(conn, ZMQ_IDENTITY, ident, strlen(ident));
zmq_setsockopt(conn, ZMQ_LINGER, &linger, sizeof(linger));
ret = zmq_connect(conn, addr);
if (-1 == ret) {
printf("connnect failed errno:%d error:%s\n", errno, zmq_strerror(errno));
exit(-1);
}
int reconn = 1;
for (int i=0;;i++) {
int timeout = 2000;
char data[256] = { 0 };
char recv_data[1024] = { 0 };
if (reconn) {
zmq_disconnect(conn, addr);
zmq_connect(conn, addr);
}
sprintf(data, "hello world:%d", i);
ret = zmq_send(conn, data, strlen(data)+1, 0);
if (ret >= 0) {
printf("send %s\n", data);
}
zmq_setsockopt(conn, ZMQ_RCVTIMEO, &timeout, sizeof(timeout));
ret = zmq_recv(conn, recv_data, sizeof(recv_data), 0);
if (ret <= 0) {
printf("recv failed!!!\n");
reconn = 0;
}
else {
printf("recv %s\n", recv_data);
}
usleep(600*1000);
}
zmq_close(conn);
zmq_ctx_term(ctx);
}
What's the actual result? (include assertion message & call stack if applicable)
I can capture the send data using wireshark and the connection state is ESTABLISHIED, but the ZMQ_POLLIN not trigger, .
What's the expected result?
The ZMQ_POLLIN event will trigger normally.