* Manual modifications to the database caused the map to act unusual
* The server's packet must be created and deleted as a char array
* removed UnloadAll() from several singleton destructors
* added SDL_Delay() to BaseScene::RenderFrame(), to reduce heavy CPU use in menus
The server's changes were easy. The clients means that the constructors
for each scene have one less argument, and each scene has one less member.
The exception to this is LobbyMenu, where the config is used in multiple
places, so it was easier to have the config's reference as a member.
To replace the config's usage, I added this line in most cases:
ConfigUtility& config = ConfigUtility::GetSingleton();
The only requirement is that ConfigUtility::Create() and
ConfigUtility::Delete() are called from the main() function.
The changes were fairly easy, and I was actually able to find and fix an
out-of-sequence bug: The destructors for the AccountManager and the
CharacterManager were using SQL code after the call to SQLite3_close_v2(),
so I added and called UnloadAll() for both of them.
This required that I switch from using a char array for the packet buffers
to using malloc() and free(). They make more sense anyway, and I've
learned (or relearned) something about casting.
I've replaced the BBox class with a pair of inline functions in
check_bounds.hpp. I've also renamed the 'position' variable to 'origin' in
several locations.
These changes are mostly to alleviate ambiguity.
I've also stopped using a separate branch for sharing the CharacterData
structre. This commit won't build, mostly because I need to refactor
InWorld to handle the loss of the PlayerCharacter class.
I should probably rename SQL's tables too.
Here are more tweaks:
* InWorld's quit event now exits to the main menu, bypassing a relayed disconnect message
* InWorld's disconnect hander no longer throws exceptions (dropped creation packet bugfix)
* CombatData's internals now point to std::pair objects
* Enemies are stored in a global list
Kayne Ruse