C++ :: Dealing With Operator Overload Function Failure

Say I have a class that requires dynamic allocation to implement a few of the operators. Take "+" for example; I need to create a new object to hold the sum of the two parameters whose size is not known at compile time.

I'm pretty sure the standard way to indicate a failure inside the overloading function would be to throw an exception. However I am currently involved in an embedded(ish) project where the spec. says no exceptions are to be used.

I think I have 2 options:

1. Return an "invalid" object (with a flag indicating an error has occurred) and check for this after each operation.

a = b + c
if (a.err)
// handle error
or

2. To forsake operator overloading entirely and think up a new way of doing things where all functions that involve dynamic allocation can return error codes. but this seems rather terrible too as I may end up with something like:

objA a
if (add(&a, b, c) == -1) // assuming b and c are initialized before this snippet starts
// handle error

Is there a number 3 that I haven't thought of? It seems that not allowing exceptions is fairly common even in the non-embedded world [URL] so how is this normally done? or is operator overloading usually avoided when exceptions are not allowed?