Segfault on delete after istream::read()

[kfmfe04]

Very good.

This is definitely good to know, in case I need to interface with other languages in the future.

#include <iostream>
#include <fstream>

#define WITH_VTABLE

struct A{
        int a;
#ifdef WITH_VTABLE
                virtual void f() { std::cout <<this->a<<std::endl; }
#endif
};

int main()
{
        std::cout <<sizeof(A)<<std::endl;

        A a;
        a.a = 0xA3;

        std::ofstream os;
#ifdef WITH_VTABLE
        os.open( "/tmp/with.bin" );
#else
        os.open( "/tmp/without.bin" );
#endif
        os.write( (char*) &a, sizeof( a )); 
        os.close();
}

hexdump /tmp/with.bin
hexdump /tmp/without.bin

shows the layout. Thank you so much for explaining!

[helios]

Do not do that. Never serialize pointers.

[kfmfe04]

I serialized the object A a

It just so happened that it had a vtbl pointer in it in the case where I left in WITH_VTABLE, which was educational, but shows exactly why we should only serialize PODs (because pointers have no meaning after serialization).

When I assign a.a to be 0xA3, I get:

/tmp$ hexdump with.bin
0000000 8c68 0804 00a3 0000
0000008
/tmp$ hexdump without.bin
0000000 00a3 0000
0000004

on a 32-bit Ubuntu box.

But what if I were serializing over a pipe and the box on the other side were also a 32-bit Ubuntu box with the same executable? I'm guessing that it would still fail because that 8c68 0804 is a pointer to some vtable in the code segment, which could be different even on different runs on the same machine? So that information would be useless.