Fixup exceptions in playcontroller #191

This commit adds an "end_play_signal" type. It is meant to be used in play_controller classes to prevent crashes due to bad exception handling / collisions, and should help us to make assumptions more easily about when networking operations will happen It also adds some macros to avoid the proliferation of unreadable try catch blocks. It is expected that many of these can be expanded and simplified after the refactor.

This uses the macros introduced to handle game end exceptions. It should be a strict refactor. Rebased to accomodate conflict here: 516206a Still a strict refactor. Note: If end_turn_exception propogates to the top of playsingle_controller, there is no handler beyond that and the program will terminate. If an end turn signal gets there, we also will terminate. This commit adds an assertion failure with a message clarifying what happened.

Continues trend of previous commit. play_turn becomes no throw with respect to these exceptions, and so we can switch calls to it to use the "PROPOGATE_END_PLAY_SIGNAL" instead of "HANDLE_...". It looks like by progressively doing this and unit testing at each step, we will be able to successfully convert all of the functions in play_controller to use PROPOGATE, which is a cheap if else and not an exception handler, and increase the stability of the engine by avoiding the use of exceptions for control flow. If we continue this way, we could push all the throws into the event handler and the replay module. Perhaps could continue from there, but likely at some point the exceptions might be the simpler solution esp. for breaking recursion in WML / lua or things like this.

After discussion on irc, it seems that this function doesn't throw end level or end turn exceptions, and some play testing suggests that removing the handler macros here is alright.

This function won't throw end level or end turn. game_event pump will not throw these, although the SDL event pump may throw them in response to keyboard / menu events.

This commit starts the process for replay_controller at the play_replay_level function.

Returns a "possible_end_play_signal" now instead. This also required changes to the header in hotkey/command_executor. This seems pretty minor, and could perhaps be refactored later.

This is overrided differently in playsingle, playmp, and replay controllers, and never is it called abstractly, which is what this virtual declaration would suggest.

This changes both the playsingle and playmp implementations, but not the replay_controller implementation which is separate. ( squashed: fixup an oversight in replay_controller::play_side )

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fixup exceptions in playcontroller #191

Fixup exceptions in playcontroller #191

Commits on Jun 8, 2014