So many topics! So little time!
I'll start listing them here so i don't forget
i'll also try to say a few words, then cover it in more depth in a separate journal entry.
game state managers
state transitions are already implicitly defined by the natural flow control of a game. the one place i use a state manager "pattern" is in my rigid body modeler/animator. it has 2 states -- model editor, and animation editor. drawscreen calls one of two routines to draw either the model editor screen or the animation editor screen - depending on the "game state" of the modeler/animator module. it then calls one of two routines to process input as either modeler or animation editor input. this is a true "state system" design. and only natural, since you can jump back and forth between model and animation editor with a single mouse click. you see its really two programs in one, a modeler, and an animation editor. the state determines which one is active - sort of non-preemptive multi tasking - ie task switching. as you can see, it doesn't seem to make much sense to slice up a game into separate "programs" (states) each with its own input and draw routines. well actually it does.! <BG>. but a state manager isn't needed unless the relationship between "mini programs" (states) is a flip flop back and forth kind of thing. or jump around between a few. but game states are usually a combo of linear, hierarchy, and loops, not peerless network type things. so the state can usually be handled automatically by call hierarchy and normal flow control methods. Whew! more than a few words! Anyway, more on how you can usually get away without a state manger in a separate post.
writing bug free code.
there are lots of tricks that can help write bug free code. at some point, i'll do an entry on all the ones i know. in a typical large title i do, i've been blessed to average just one non-show stopping code type bug in the release versions. typos in displayed text are another thing however... <g>.
game state managers
there was a recent thread on using game states for menus and such. can't find the post. in it, i said that the natural call hierarchy of games made state driven games overkill. in a later post, i mentioned an application of state driven that i do use: the modeler/animator module. its two programs in one, a modeler, and an animation editor, and you can switch between the two at any time. just now, i realized that caveman is also state driven! its two games in one: rpg, and person sim. so it has two states it runs in: fps/rpg mode, and "The sims" doing an a action mode. each has its own render and input methods. but most games aren't like this, and therefore are not true hybrid multi-state applications. it appears the timeto use state management is when your app is actually two or more apps of equal importance, IE they have a peer 2 peer type relationship, vs a main app and sub/mini app relationship. menus and such are sub-apps of the app that calls them. the main menu is a sub app of main. the game loop is a sub app or the main menu. the in-game menu is a sub-app of the game loop.
sub menus, stats screens, maps, etc.
- not -
main menu <---> game loop <----> in-game menu <---> world map <---> etc <----> etc (for every screen/menu in the game!)
natural call hierarchy handles it all for you. no need to manage states. granted, it can be done that way, but encoding the natural call hierarchy as state transitions can get ugly.
here's examples of state systems that seem to make more sense to me:
model <-----> animate
fps/rpg mode <----> person sim mode