Thread: bad_alloc

As for your example, your understanding of C++'s evaluation rules is wrong. When dealing with uninitialized variables, undefined behaviour is triggered when the lvalue of the variable is converted to an rvalue. Since the address-of operator involves no such conversion, no undefined behaviour is triggered. See 4.1/1.

The subexpression x merely gets the lvalue, which is a valid operation for uninitialized variables.

Yes, but the argument was that because v[0] is an undefined subexpression of &v[0] that therefore the entire expression &v[0] is undefined.

I would rather use CornedBee's explanation: v[0] accesses an element that does not exist, possibly even causing a null pointer to be dereferenced. From then on we are dealing with undefined behaviour, so &v[0] can be considered as undefined. (What does it mean to take the address of an object that does not exist?)

On the other hand, &x just takes the address of x. x exists and has an address, though its value has not yet been defined.

How about Linux, X86-64, GCC 4.1?

Code:

#include <vector>
#include <iostream>

int main()
{
  std::vector<int> v;
  std::cout << "Address of first: " << &v[0] << '\n';
}

Code:

wasti@watson64 ~/forums $ g++ -D"_GLIBCXX_DEBUG=1" -o crash crash.cpp
wasti@watson64 ~/forums $ ./crash
/usr/lib/gcc/x86_64-pc-linux-gnu/4.1.2/include/g++-v4/debug/vector:194:
    error: attempt to subscript container with out-of-bounds index 0, but
    container only holds 0 elements.

Objects involved in the operation:
sequence "this" @ 0x0x7fff46e9f940 {
  type = N15__gnu_debug_def6vectorIiSaIiEEE;
}
Aborted