12 replies to this topic

[gardon]

Quick memory question regarding textures.

I'm coding a memory/image manager in C++ and SDL, and am running into trouble. AFter debugging I"ve figured out that my class is working properly, but it has soemthing to do with the vectors I"m using.

Currently, I have a vector of sTexture structs, that are declared as follows:

struct sTexture
{
SDL_Surface*;
int width;
int height;
}

Where SDL_Surface* is used to store image data.

and in my initialization:

sTexture newTexture;

newTexture.width = w;
newTexture.height = h;
newTexture.surface = textureSurface;

m_Textures.push_back(newTexture);

The following code above was addressed in a function where the width, height, and surface was passed.


My problem is when I'm trying to draw it, I use the surface from the texture I've pushed back onto the vector using m_Textures[whateverNumber].surface

The memory instance is not there. What's wrong? I explicitly call the appropriate functions. It's as if a vector can not store pointer data.

Jason

[.oisyn]

I don't know what the problem is, but I can tell you it has nothing to do with the vector itself.

[SigKILL]

The problem could be that textureSurface has been deleted (may possibly be a local variable somewhere), or that there a no sufficient = operator (or possibly copy constructor) for sTexture (try to create one explicitly if you haven't done so). If you're a destructor abuser you might delete textureSurface in the sTexture destructor (which would cause problems). I could probably continue to give 1000 other possibilities...

[gardon]

No it's not that either, I've checked. I purposely made sTexture a struct (thus the s prefix) so I wouldn't run into this problem.

I dont' know, I"m going nuts.

[Reedbeta]

You said you used a debugger; can't you successively narrow down the section of code in which the bug occurs until you're close enough to see it? Bearing in mind that it's possible part of your code outside this class is stomping on the memory somehow.

[gardon]

Yes. It narrows down to where I shuve the sTexture struct into the vector

[gardon]

I found the direct reason that's causing it.

I've eliminated all other elements besides this one: loading memory through functions.

The only way it works is if I load the surface into memory just before the draw (in the draw function).

The moment I take that: m_surface = SDL_LoadBMP

and put it into another function, even if I call that function right before the draw function, everything goes nuts. Is it my computer perhaps? It seems as if it can't keep track of memory.

[Reedbeta]

How about if you zip up your code and upload it someplace; someone here can download it and see if they encounter the same behavior.

[juhnu]

Might it be you are freeing your SLD (texture) surface twice?

[.oisyn]

Yeah it sounds like heap corruption. Either you're deleting a pointer multiple times like juhnu suggests, you're deleting a pointer you've never allocated, or you're writing past your buffers.

Usually when this happens, you will never notice it at the place and time where it goes wrong, but rather later on when allocating or freeing a pointer in a piece of code that has nothing to do with your actual bug.

[Goz]

Yuor initialisation code block REALLY looks like you are creating a local variable. Once you leave that function that data will no longer exist and the struct will get overwritten ...

[.oisyn]

Nonsense, the vector stores a copy of that variable, which makes perfect sense.

[Goz]

.oisyn said:

Nonsense, the vector stores a copy of that variable, which makes perfect sense.

Ahh yeah ... damn mornings. If it had been

m_Textures.push_back(&newTexture);

I might be talking sense ;)