[ASAN] AddressSanitizer FN, failed to report memory leaks

[wr-web]

reproduce: https://godbolt.org/z/WEjY9rY1G
bug code:

#include <iostream>

void unsuspectingBug() {
    int* ptr = new int(42);

    // Perform some logical operations that don't affect ptr directly
    for (int i = 0; i < 100; ++i) {
        int temp = i * i; // Irrelevant computation
        if (temp == 42) {
            delete ptr;
        }
    }

    // Forgetting to nullify ptr after deletion
    // Logical usage without realizing it's been erased
    if (ptr) {
        std::cout << *ptr << std::endl; // Dereference of dangling pointer not realized due to logical oversight
    }
}

int main() {
    unsuspectingBug();
    return 0;
}

[DimitryAndric]

Note that this code only leaks memory, it doesn't deference the pointer after freeing it, as it is never freed. This is because i * i can never become 42. If you use -fsanitize=leak it will show the same report as gcc.

[DimitryAndric]

I'm unsure how godbolt's clang is configured and built, but apparently gcc enables LeakSanitizer when using AddressSanitizer?

[wr-web]

Note that this code only leaks memory, it doesn't deference the pointer after freeing it, as it is never freed. This is because i * i can never become 42. If you use -fsanitize=leak it will show the same report as gcc.

It seems that enabling both sanitizers fails to report the leak. You can see this here: https://godbolt.org/z/e5oMnEP91.

However, enabling only LeakSanitizer for Clang will correctly detect the issue: https://godbolt.org/z/9W8Mrc3Mv.

This is likely a false negative (FN)