Code: Select all
ITexture* tex = new CD3D9Texture(this, size, name, format);
if (tex)
{
checkDepthBuffer(tex);
addTexture(tex);
tex->drop();
}
return tex;
Useless checks in Irrlicht source
Useless checks in Irrlicht source
I was going through the source code of the Direct3D9 driver for something I'm working on, and I found this:
the if(tex) is completely useless. Unless the new keyword has been overwritten on CD3D9Texture (which it hasn't been), it is impossible for C++ to return a null pointer from new. If allocation fails, it will throw std::bad_alloc, but tex will never be null. Why check?
That is assuming that you are not compiling with C++ exceptions disabled, or you are not using an older compiler that returns 0 instead of throws std::bad_alloc.
But yeah, it is unneeded. Somebody was probably in the habit of always checking for null pointers when they wrote that. It doesn't really hurt anyways.
But yeah, it is unneeded. Somebody was probably in the habit of always checking for null pointers when they wrote that. It doesn't really hurt anyways.
Checking for null-pointers is never bad habit. In one of engines I used, they had custom new and delete operators which could return 0 in some cases (the reason is that they pre-allocated memory and new and delete were really fast, but you couldn't allocate always).
In the end, computer memory is limited, and who knows what other applications might be active at the time user runs the game (and he really doesn't care about null pointers and exceptions, but only about nasty crash he would get).
In the end, computer memory is limited, and who knows what other applications might be active at the time user runs the game (and he really doesn't care about null pointers and exceptions, but only about nasty crash he would get).
I'm not saying it's awful, or a bad habit, but it is useless in this case. The new operator is not overloaded, and any compiler that follows the C++ standard (which is 7+ years old) will never return NULL from new. If you want to talk about always checking your pointers, there should be a try {} catch(std::bad_alloc ex) {} block. I'm not saying that I think that's necessary, but it would accomplish what the NULL check is trying to do.
-
hybrid
- Admin
- Posts: 14143
- Joined: Wed Apr 19, 2006 9:20 pm
- Location: Oldenburg(Oldb), Germany
- Contact:
The depth texture thing has changed a lot in the last versions. I'd bet that this was just a reminicent to the old code. Something which changed evolutionary. Could be as well that I wrote that with some other thing in mind, just as Nalin said. Moreover, I find these locations indeed easier to skip over. Otherwise, I'd always stop at this location, checking if the pointer could be wrong.
