Fixes#4040
* The TCP connection is now shutdown after OnBodyFinished
* Any open connections are closed when cNetworkSingleton::Terminate() is called.
* Removed ownership cycles in cUrlClientRequest
* Added a check to the test to ensure there are no leaks.
* Fixes a number of "<function>: not all control paths return a value" warnings on MSVC.
* Introduces the UNREACHABLE global macro and uses it instead of conditionally compiled switch defaults.
* Move cNBTParseErrorCategory from FastNBT.h into FastNBT.cpp to prevent bad calls to message()
* Replaces AppendVPrintf with fmt::sprintf
* fmt::ArgList now used as a type safe alternative to varargs.
* Removed SIZE_T_FMT compatibility macros. fmt::sprintf is fully portable and supports %zu.
* Adds FLOG functions to log with fmt's native formatting style.
Stop using gdb on osx - was breaking the build
Add clang 3.5 build as travis now defaults to 5.0
Fix unknown-warning-option errors on AppleClang
ProtoProxy: Use nullptr
UrlClientTest: add override to callback destructor
Update jsoncpp to use nullptr
* Renaming changes:
* macro prefix "POLARSSL" -> "MBEDTLS"
* functions now prefixed with "mbedtls_"
* rename PolarSSL++ -> mbedTLS++
* rename polarssl submodule
* Use mbedtls' AES-CFB8 implementation.
* Add cSslConfig to wrap mbedtls_ssl_config
* Update cTCPLink and cBlockingSslClientSocket to use cSslConfig
* Use cSslConfig in cHTTPServer
* Use cSslConfig for cMojangAPI::SecureRequest
* CI Fixes
* Set -fomit-frame-pointer on the right target
* Check for intersection between placed blocks and entities.
+ Implemented GetPlacementCollisionBox, to permit custom placement collision boxes for blocks.
* Factored block-entity placement checking into another function in cPlayer.
- Removed vector min/max functions
* Use GetWorld to get the world in DoesPlacingBlocksIntersectEntity.
+ Added block height checks, allow different cEntity subclasses to decide whether they will prevent block placement.
* SelfTests: Added a cEvent stress-test.
* cNetwork: Fixed startup event hang.
The original code used to hang with a ~ 1:50000 chance, because on Linux the cEvent was destroyed before its "Set()" returned.
