Game::AI++ 0.6 Released

[alexjc]

I've made a tentative decision that this seminar will take place at 9 P.M. CET on Saturday (that's 3 P.M. EDT). See this time zone converter.

The topic will be specifically about the implementation of the behavior tree, the framework itself, and an analysis of the ideas generally. Q&A throughout probably.

If you have any requests or comments, now's the time!

Alex

P.S. The event will hopefully be recorded!

[Andrew]

Hope I'm back in time to listen in, but a recording is great, I expect I'll still be on a train at that time.

[Dave Mark]

I should be in. Do I need to wear a tie? (That was meant for you, Andrew!) ;-)

[alexjc]

O.K. Everything is looking good. We'll be starting in 20 minutes. Here's the link:

[EDIT: This meeting already took place and is now over.]


You'll be sent into a waiting room, and I'll let you in ASAP.

Alex

[HoloDoc]

I just downloaded the package and I'm missing the api documentation files. Can I find these somewhere else?

[alexjc]

HoloDoc,

Yes just use Doxygen and regenerate it yourself. All the comments in the code are used to build the API documentation. I'll try and include this with the next release (it was getting too big for little benefit).

Alex

[knizz]

Here is a Makefile for Linux/GCC 4.3.0-users:

# MAKEFILE for Game::AI++
# 5 Files have to be modifed to work with linux and gcc 4.3.0

# tests/unit_Selector.h
#	Add the <cstdlib> header

# alive/tree/Selector.h
#	Add the <cstdlib> header

# alive/core/Core.h
#	Uncomment line 25:
#		// #include "../thirdparty/mmgr.h"

# alive/core/Generic.h
#	Add the <cstdio> header
#	At line 150 replace:
# 		typedef VTable< const Base, Func > VTable;
#  		const VTable* vtable_; // vtable pointer
#	with
# 		typedef VTable< const Base, Func > _VTable;
#  		const _VTable* vtable_; // vtable pointer
#	Replace VTable with _VTable in lines 265, 234

# alive/core/SignalSlot.h
#	At line 67 and 125 replace:
#		typename Delegates::iterator it =
#			std::find(m_Delegates.begin(),m_Delegates.end(), d);
#		if (it != m_Delegates.end())
#		{
#			m_Delegates.erase(it);
#		}
#	with (67)
#		Delegates::iterator it;
#		for (it = m_Delegates.begin(); it != m_Delegates.end(); ++it){
#			m_Delegates.erase(it);
#			return;
#		}
#	with (125)
#		typename Delegates::iterator it;
#		for (it = m_Delegates.begin(); it != m_Delegates.end(); ++it){
#			m_Delegates.erase(it);
#			return;
#		}

# Result of tests/tests:
#	OK (165)

CC=g++ -I. -Ialive

TESTOBJ=tests/doc_OrderedPlanning.o tests/doc_ConditionExecution.o tests/unit_Node.o tests/unit_Task.o tests/unit_Selector.o tests/unit_Filter.o tests/unit_Delegate.o tests/doc_FuturePlanning.o tests/unit_LookupTable.o tests/unit_ControlFlow.o tests/unit_Decorator.o tests/doc_SequenceExecution.o tests/doc_ActionExecution.o tests/doc_ConditionalAction.o tests/unit_Repeat.o tests/func_LookupTable.o tests/func_Stack.o tests/unit_Stack.o tests/unit_SignalSlot.o tests/unit_Memory.o tests/func_Node.o tests/unit_BlackboardCompare.o tests/func_Condition.o tests/func_Tree.o tests/MockNode.o tests/unit_Core.o tests/unit_Sequence.o tests/func_ControlBehavior.o tests/unit_Scheduler.o tests/doc_SelectorExecution.o tests/TestFramework.o tests/unit_Interpreter.o tests/unit_Resolve.o tests/unit_Registry.o tests/func_Visitor.o tests/func_Action.o tests/doc_GreedyPlanning.o tests/stress_Nodes.o tests/doc_ParallelExecution.o tests/unit_Parallel.o tests/unit_Transform.o tests/func_Persistence.o tests/func_PrettyPrinter.o
ALIVEOBJ=alive/Condition.o alive/Brain.o alive/framework/Registry.o alive/framework/Task.o alive/framework/Scheduler.o alive/engine/Interpreter.o alive/engine/Node.o alive/engine/Transform.o alive/thirdparty/mmgr.o alive/tree/Decorator.o alive/tree/CompositeTools.o alive/tree/Filter.o alive/tree/Sequence.o alive/tree/Selector.o alive/tree/ControlFlow.o alive/tree/Parallel.o alive/tree/Repeat.o alive/tree/Resolve.o alive/core/Core.o alive/extensions/ControlBehavior.o alive/extensions/LookupTable.o alive/extensions/Stack.o

all: alive test

alive: libalive.so
alive.a: $(ALIVEOBJ)
	ar q $@ $(ALIVEOBJ)
libalive.so: $(ALIVEOBJ)
	$(CC) -shared -o $@ $(ALIVEOBJ)
alive/%.o: alive/%.cpp
	$(CC) -c -o $@ $?

test: tests/tests
tests/tests: $(TESTOBJ)
	$(CC) -L. -ldl -lcppunit -lalive -o tests/tests $(TESTOBJ)
tests/%.o: tests/%.cpp
	$(CC) -Itest -c -o $@ $?

David

[dgroth]

Hello Alex,

First, beautiful code; it's a pleasure to read. However, it looks like there is a rather huge bug in the 0.6 release of GameAI++

The MockInterpreter is derived from Interpreter, which means that:
- the destructor of MockInterpreter will be called first and then,
- the destructor of Interpreter will be called

The MockInterpreter destructor explicitly deletes m_Scheduler and m_Registry, however, the destructor of Interpreter happily use m_Registry with "deleteAll" call.

This leads 100% segmentation faults. I can't believe no one has experienced this problem since it clearly pops up with the unit tests routines.

Since I don't know what going on yet (still learning the code), could you please advice on what we need to do? What is the intent? Should we delete m_Registry and m_Scheduler in the Interpreter destructor instead (*after* they've been used)?

Oh, and David, your GCC 4.3 suggestion of putting a "return" statement in the middle of a "for" loop is functionally incorrect.

Best regards, Dan.

[jesu37]

hi all I'm looking for a good example of behavior trees, and you can not download, anyone can help?

[alexjc]

I didn't actually manage to reproduce the bug before releasing that code, but yes, you're right of course. The answer is a minor rewrite of the MockInterpreter to use composition instead of inheritance.

I don't recommend using this code in practice anyway, it's very much experimental R&D code, but that's up to you :-)

Alex