Unverified Commit afd5d9f7 authored by Luca Boccassi's avatar Luca Boccassi Committed by GitHub

Merge pull request #2939 from sigiesec/fix-poll-timer-event-retiring

Fixed a deadlock and an assertion failure in poll_t
parents ac777bad 5873894c
......@@ -52,6 +52,7 @@ zmq::poll_t::poll_t (const zmq::ctx_t &ctx_) :
zmq::poll_t::~poll_t ()
{
stop ();
worker.stop ();
}
......@@ -141,12 +142,23 @@ void zmq::poll_t::loop ()
// Execute any due timers.
int timeout = (int) execute_timers ();
cleanup_retired ();
if (pollset.empty ()) {
// TODO yield? or sleep for timeout?
continue;
}
// Wait for events.
int rc = poll (&pollset[0], pollset.size (), timeout ? timeout : -1);
#ifdef ZMQ_HAVE_WINDOWS
wsa_assert (rc != SOCKET_ERROR);
#else
if (rc == -1) {
errno_assert (errno == EINTR);
continue;
}
#endif
// If there are no events (i.e. it's a timeout) there's no point
// in checking the pollset.
......@@ -168,20 +180,23 @@ void zmq::poll_t::loop ()
if (pollset[i].revents & POLLIN)
fd_table[pollset[i].fd].events->in_event ();
}
}
}
// Clean up the pollset and update the fd_table accordingly.
if (retired) {
pollset_t::size_type i = 0;
while (i < pollset.size ()) {
if (pollset[i].fd == retired_fd)
pollset.erase (pollset.begin () + i);
else {
fd_table[pollset[i].fd].index = i;
i++;
}
void zmq::poll_t::cleanup_retired ()
{
// Clean up the pollset and update the fd_table accordingly.
if (retired) {
pollset_t::size_type i = 0;
while (i < pollset.size ()) {
if (pollset[i].fd == retired_fd)
pollset.erase (pollset.begin () + i);
else {
fd_table[pollset[i].fd].index = i;
i++;
}
retired = false;
}
retired = false;
}
}
......
......@@ -79,6 +79,8 @@ class poll_t : public poller_base_t
// Main event loop.
void loop ();
void cleanup_retired();
// Reference to ZMQ context.
const ctx_t &ctx;
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment