Reduces memory usage and CPU usage.

Thanks to Guido Hinderberger, Martin Rauscher and Michael Hekel.

Bug: https://bugs.eclipse.org/bugs/show_bug.cgi?id=468987

Change-Id: I3de322006623483cbf20218da071a9da5d7b2e2b
Signed-off-by: Anmol Sarma <me@anmolsarma.in>

Don't duplicate the increment when queueing.

Thanks to Joe McIlvain.

Bug: https://bugs.eclipse.org/bugs/show_bug.cgi?id=474935

Fix handling of outgoing QoS>0 messages for bridges that could not be
sent because the bridge connection was down.

Change-Id: I39db613c8e75c1437b909de263df4f96a7f76880
Signed-off-by: Anmol Sarma <me@anmolsarma.in>

Signed-off-by: Anmol Sarma <me@anmolsarma.in>

Thanks to Jaime Yu.

Bug: https://bugs.eclipse.org/bugs/show_bug.cgi?id=465438

Fix incorrect loop timeout being chosen when using the threaded
interface and keepalive = 0.

Thanks to Gianfranco Costamagna.

Bug: https://bugs.eclipse.org/bugs/show_bug.cgi?id=471334

Thanks to Martin Rauscher.

Bug: https://bugs.eclipse.org/bugs/show_bug.cgi?id=470660

Adding a new temporary variable for realloc memory in order to keep the reference to the allocated memory before the function returns.
In case that realloc fails (i.e. OOM), before exiting with "return(1)", the application is still able to free the corresponding allocated memory.

Signed-off-by: Eduardo Sanchez <ledusanchez@gmail.com>

Coverity CID 1292477.

Coverity CID 1292471.

If a custom notification topic was used, the bridge status would be set
to 1 before a connection was attempted, this has been fixed.

Initial updates to the status topic are now only set once, when the
broker starts, rather than each time the bridge attempts to reconnect.

Bug: https://bugs.eclipse.org/bugs/show_bug.cgi?id=467096

Thanks to Arjen van Elteren and Stephen Zeng.

Bug: https://bugs.eclipse.org/bugs/show_bug.cgi?id=464811

This was being incorrectly set as rc=4.

socket is an opaque object on Windows, so this check isn't valid.

Thanks to Stefan Rosskopf.

Bug: https://bugs.eclipse.org/bugs/show_bug.cgi?id=465384

Fix possibility of select() being called with a socket that is >FD_SETSIZE.

Bug: https://bugs.eclipse.org/bugs/show_bug.cgi?id=464632

Bug: https://bugs.eclipse.org/bugs/show_bug.cgi?id=464779