[zeromq-dev] [PATCH] Fix pgm_receiver.cpp: zmq_assert (pending_bytes == 0) (LIBZMQ-205)

[Martin Lucina]

This patch fixes the problem described in LIBZMQ-205. The assertion itself
is probably caused by previously queued POLLIN events arriving after POLLIN
has been disabled on the socket.

The following additional bugs have been fixed as part of debugging this
problem:

- pgm_receiver_t does not flush messages written to the session in all
  cases which can lead to a stalled reader. Add calls to session->flush ()
  in the appropriate places.

- ensure to restart polling when a pending message is flushed in
  activate_in ().

Signed-off-by: Martin Lucina <martin at lucina.net>
---
 src/pgm_receiver.cpp |   13 +++++++++++--
 1 files changed, 11 insertions(+), 2 deletions(-)

diff --git a/src/pgm_receiver.cpp b/src/pgm_receiver.cpp
index 122d110..99e882b 100644
--- a/src/pgm_receiver.cpp
+++ b/src/pgm_receiver.cpp
@@ -117,8 +117,15 @@ void zmq::pgm_receiver_t::activate_in ()
     //  processed the whole buffer but failed to write
     //  the last message into the pipe.
     if (pending_bytes == 0) {
-        if (mru_decoder != NULL)
+        if (mru_decoder != NULL) {
             mru_decoder->process_buffer (NULL, 0);
+            session->flush ();
+        }
+
+        //  Resume polling.
+        set_pollin (pipe_handle);
+        set_pollin (socket_handle);
+
         return;
     }
 
@@ -128,6 +135,7 @@ void zmq::pgm_receiver_t::activate_in ()
     //  Ask the decoder to process remaining data.
     size_t n = mru_decoder->process_buffer (pending_ptr, pending_bytes);
     pending_bytes -= n;
+    session->flush ();
 
     if (pending_bytes > 0)
         return;
@@ -145,7 +153,8 @@ void zmq::pgm_receiver_t::in_event ()
     unsigned char *data = NULL;
     const pgm_tsi_t *tsi = NULL;
 
-    zmq_assert (pending_bytes == 0);
+    if (pending_bytes > 0)
+        return;
 
     if (has_rx_timer) {
         cancel_timer (rx_timer_id);
-- 
1.7.2.5