Is valgrind always right?

I used valgrind to check for memory leaks. When I checked my application with "--leak-check=full --show-reachable=yes -v", it showed tons of errors. However when I looked at the error verbose message, it showed that the ones causing the leak is not my application, but the libraries I was using. Eg, I used SDL for the graphics and CUPS for printing. Valgrind showed a leak on the "SDL_Init()" line from my application. But I am certain that I called "SDL_Quit()" in the destructor of my application. And valgrind not only showed one error from this SDL_Init() line, but many. So valgrind showed approximately 10 message of errors from this line. The same applies to the CUPS library. I made a class to manage the printing. In its constructor, I called cupsGetDests() and store the result to a private member. And in the destructor, I called cupsFreeDests() to free that variable. But valgrind showed a bunch of errors from the "cupsGetDests()" line. Why does this happen?

Thanks in advance.

Have you tried a small test application like

Code:

---

int main()
{
SDL_Init();
SDL_Quit();
return 0;
}

---

??

Some leaks are reported just because you THINK you called the quit - but actually it was not called on some reason...

So make a small test with the call to problematic function only and see if it has leaks...
If it has - you can ignore them in the report build for your own application (while you do not call Init/Quit repeatedly in the lifecycle of the application)

There was this post to the SDL newsgroup last month. I have no idea if that applies to your program or not. (Try the minimal, like vart sugggested.)

Just reminded me of it.

Yeah, it's still there. In fact there's also an error report in SDL_SetVideoMode() line. So, I think it's not my fault after all. Thanks again guys. :D