X86 is able to deal with
misaligned memory, but it can hurt perf.
Other arch like for example mips
is not able to digest it.
So in order of portability we should get
rid of this UB.
Should fix these memory leaks:
==22737== 14,598,040 (131,472 direct, 14,466,568 indirect) bytes in 2,739 blocks are definitely lost in loss record 3,124 of 3,126
==22737== at 0x4C2F1CA: operator new(unsigned long) (vg_replace_malloc.c:334)
==22737== by 0x90FE4B: LoaderDFF::loadFromMemory(std::shared_ptr<FileContentsInfo>) (LoaderDFF.cpp:443)
==22737== by 0x7BCC86: GameData::loadModel(unsigned short) (GameData.cpp:474)
==22737== by 0x7DF7BC: GameWorld::createInstance(unsigned short, glm::tvec3<float, (glm::precision)0> const&, glm::tquat<float, (glm::precision)0> const&) (GameWorld.cpp:144)
==22737== by 0x7DF44C: GameWorld::placeItems(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) (GameWorld.cpp:120)
==22737== by 0x758D38: RWGame::newGame() (RWGame.cpp:116)
==22737== by 0x786389: LoadingState::enter() (LoadingState.cpp:9)
==22737== by 0x75DC59: void StateManager::enter<LoadingState, RWGame*, RWGame::RWGame(Logger&, int, char**)::{lambda()#1}>(RWGame*&&, RWGame::RWGame(Logger&, int, char**)::{lambda()#1}&&) (StateManager.hpp:40)
==22737== by 0x758484: RWGame::RWGame(Logger&, int, char**) (RWGame.cpp:81)
==22737== by 0x747815: main (main.cpp:13)
This moves the object rendering logic into ObjectRenderer. This makes
GameRenderer a bit smaller.
There are some rendering logic decisions that haven't been brought
back yet since they may be better placed elsewhere.