/*
==============
jmGameLocal::SpawnEntity
==============
*/
template<class t>
jmEntity *jmGameLocal::SpawnEntity( const char *className, entType_t type ) {
t *entity = NULL;
entity = new t();
strcpy(entity->classname, className);
entity->type = type;
entity->Init();
//VectorSet(entity->angle, 0, 0, 0);
//VectorSet(entity->origin, 0, 0, 0);
memset(entity->angle, 0, sizeof(vec3_t));
memset(entity->origin, 0, sizeof(vec3_t));
//entities.Append( entity );
entities[ numEntities++ ] = entity;
return entity;
}
...
// Allocate player entities at the start of the list.
for(int i = 0; i < MAX_PLAYERS; i++) {
SpawnEntity<jmPlayer>( va("player_%d", i), ENT_PLAYER );
}
[/SOURCE]C++ Superclass memory curroption
Author
I'm having a problem were when I have a class thats derived from another class and the base class is in a array memory gets curropted. In our game we have a base class for all entities called jmEntity. There is a class thats based off of it called jmPlayer.
class jmEntity {
...
}
class jmPlayer : public jmEntity {
...
}
All entities are stored in a array in a class called jmGameLocal.
class jmGameLocal {
public:
jmEntity *entities[MAX_GENTITIES];
...
}
I'm creating each jmEntity based off it is superclass in this function:
The class gets curropted as the game goes on, even though the pointers to jmEntity's are still valid. My question is what am I doing wrong and what should I do to fix it.
Author
Quote:Original post by Sneftel
...congrats!
Sorry I relized I only posted like a sentence my computer was acting up.
I suspect it is your strcpy that is corrupting something (you don't seem to allocate any space for your copy). Instead of wading through it, I'm going to suggest that since you are using classes, you should use C++'s other functionalities. Namely, std::vector and std::string. Also, look into constructors to avoid having to set member variables to sane values in user code (you shouldn't have to memset member variables like you do, it's way too prone to errors).
[edit] I also see that you are doing some sort of home-brew RTTI. It may be a valid approach in your case, but most of the time it means that you shouldn't be using inheritance, or that you are not abstracting enough. In any case, I don't see the point of reinventing the wheel if you must use it. C++ has the typeid operator.
jfl.
[edit] I also see that you are doing some sort of home-brew RTTI. It may be a valid approach in your case, but most of the time it means that you shouldn't be using inheritance, or that you are not abstracting enough. In any case, I don't see the point of reinventing the wheel if you must use it. C++ has the typeid operator.
jfl.
Author
Its not my strcpy classname is defined as char classname[ 100 ]. I hate std::vector its too slow, i have my own dynamic memory class. My problem is the game runs fine for about 35 seconds thats when the entities array starts to become curropted.
so theres nothing wrong with going jmEntity *ent = new jmPlayer, and storing that in the jmEntity array?
so theres nothing wrong with going jmEntity *ent = new jmPlayer, and storing that in the jmEntity array?
I would recommend using std::vectors and std::strings over raw arrays and char*s.
As it is, I'd have to see the entity constructor, but I'm guessing it doesn't allocate enough room for the name, and the strcpy is overwriting memory it shouldn't.
Edit: beaten by jfl.
Edit2:
I bet you it's not. Have you ever, I don't know, profiled one?
As it is, I'd have to see the entity constructor, but I'm guessing it doesn't allocate enough room for the name, and the strcpy is overwriting memory it shouldn't.
Edit: beaten by jfl.
Edit2:
Quote:
I hate std::vector its too slow
I bet you it's not. Have you ever, I don't know, profiled one?
Quote:Original post by crozzbreed23
Its not my strcpy classname is defined as char classname[ 100 ]. I hate std::vector its too slow, i have my own dynamic memory class. My problem is the game runs fine for about 35 seconds thats when the entities array starts to become curropted.
so theres nothing wrong with going jmEntity *ent = new jmPlayer, and storing that in the jmEntity array?
I'm sure it's not your strcpy. It must be someone else's. I suspect, however, that if you tried using std::string/std::vector/modern C++ techniques, you wouldn't even be asking this question.
That said, if you want a more precise response, you will need to provide all relevent code (the smallest amount of code that reproduces the error).
jfl.
Quote:Original post by crozzbreed23
I hate std::vector its too slow, i have my own dynamic memory class.
No, it isn't. It provably isn't. Plenty of people, here and elsewhere, myself included, have done benchmarks and examined object code to show that it isn't. If you think it is, you're wrong, and it's keeping you from being a better programmer.
Quote:My problem is the game runs fine for about 35 seconds thats when the entities array starts to become curropted.
Quite likely it becomes corrupted well beforehand and it just doesn't show up until that point; memory corruption can fester for awhile before causing visible problems, a big reason why such bugs are such a pain to track down. That's one reason why good C++ programmers avoid doing direct memory manipulation stuff whenever feasible.
Author
Quote:I bet you it's not. Have you ever, I don't know, profiled one?
I've had problems with std::vector in the past, mabye cause I was using it wrong whatever, but I created my own dynamic memory class and it works, but in the case i wasn't using it for the entities array I just left it finite. For the sake of argument I did put jmEntities into a std::vector, and it still crashes like I thought.
I agree that this is really hard to track down, do you guys seriouslly think its the chars array? There both null terminated, and they both ahve a finite size.
Heres the jmEntity class.
//// jmEntity//class jmEntity {public: void SetOriginAndAngle( vec3_t origin, vec3_t angle ); virtual void Init( void ); // Run entity thinking. virtual void Think( void ); virtual void ClientThink( void ); // Sends a snapshot for this entity. virtual void SendSnapshot( msg_t &msg ); virtual void ReadSnapshot( msg_t &msg );public: char classname[ 100 ]; char name[ 100 ]; vec3_t origin; vec3_t angle; vec3_t color; vec3_t light_radius; entType_t type; int entNum; bool active;};[/SOURCE]When I wrote a terrain engine awhile back I was using std::vectors I found out later it was my bottleneck, I had to define NDEBUG or build in release mode to speed back up std::vector.
Quote:Original post by crozzbreed23
When I wrote a terrain engine awhile back I was using std::vectors I found out later it was my bottleneck, I had to define NDEBUG or build in release mode to speed back up std::vector.
Which compiler is this? Also, the point of having Debug/Release modes is to have more error catching code while debugging, which will slow things down, and have less error catching code in release mode. As always, you must profile in release mode with optimizations on. With Visual C++ 2005, you may also want to define _SECURE_SCL=0, since it does bounds checking and iterator validation even in release mode (as part of Microsoft's push for more secure code).
As for the code you posted, that gives no relevant information. Is there any way you could post the minimum amount code that compiles and runs, and shows the problem? If it's too much code, then I suggest getting really comfortable with your debugger and start tracking down the error from the point it crashes.
jfl.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
