The functions implemented in Heap.c are not usable for Windows builds.
Therefore use only for none MSVC builds.
Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

* '373' of https://github.com/eclipse/paho.mqtt.c.git:
  Add latest restart test changes
  Add test changes

It turned out in https://github.com/eclipse/paho.mqtt.c/issues/385 that
Socket_getReadySocket() could remove a list element, which is currently
in use by a function called by Socket_noPendingWirte().
And this leads to memory corruption or segmentation fault.
To fix this, this call of Socket_PendingWrites() in MQTTAsync.c needs to be protected by mutex_socket.
Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Setting pointers to NULL after free never hurt, but may prevent
double free and other problems.
Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

The function myfree() is used as a replacement for free().
Therefore it must not cause any failure, if someone calls myfree(NULL).
Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

* test4issue373:
  Add log output for failed message queueing in test due to disconnect
  Add trace output to verify that reducing the tcp send buffer forces TCPSOCKET_INTERRUPTED
  Reduce the tcp snd buffer to provocate issue #385
  Move the evaluation of the not freed memory _after_ call of MQTTAsync_destroy()
  Fix Windows build of test_issue373 and increase number of iterations
  Consider any not freed memory as test failure
  Modified number of connect iterations and messages send per second
  Moved the decrement of pendingMessageCnt to another call back function
  Add test for issue #373 and #385
  Add skeleton of test_issue373

* '373' of https://github.com/eclipse/paho.mqtt.c.git:
  Forgot to remove test failure code!
  Make sure writeComplete deals with socket error #373
  Some corrections for failing socket writes #385
  Take account of not finding a response in writeComplete #385
  Add another continuous write test
  Add continuous send test
  More memory cleanup on network errors - #373

Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

The functions implemented in Heap.c are not usable for Windows builds.
Therefore use only for none MSVC builds.

Increase the workload and number of connect/disconnects.
Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

For some unknown reason, decrementing pendingMessageCnt in
test373DeliveryComplete() does not work. Therefore moved to other call
back functions.
Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

This test connects to a broker up to 10 times.
As long as the connection is established, 100 message should be published
each second.
At the end all memory should be freed and no segmentation occurred.
Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

The function MQTTPacket_free_packet() is not used at all.
Eliminating dead code to improve readability.
Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Remove shadow warning and improve consistency

Fix Socket_setnonblocking() for Windows (At least if build with MSVC …

Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Partially revert commit a16cff63 for file
src/Socket.c

- Do not include errno.h for Windows build
- Use errno as a local variable in Socket_setnonblocking() for Windows
Signed-off-by: Juergen Kosel <juergen.kosel@softing.com>

Don't shadow already defined variable (data_size).
In addition, we are consistence with the case statement
above.
Signed-off-by: Niklas Johansson <niklas.johansson@prevas.dk>

Add support for selecting/enforcing a TLS version