the design of inner function _S_oom_malloc in SGI-STL allocator

The code is as follows:

template <int __inst>
void*
__malloc_alloc_template<__inst>::_S_oom_malloc(size_t __n)
{
    void (* __my_malloc_handler)();
    void* __result;

    for (;;) {
        __my_malloc_handler = __malloc_alloc_oom_handler;
        if (0 == __my_malloc_handler) { __THROW_BAD_ALLOC; }
        (*__my_malloc_handler)();
        __result = malloc(__n);
        if (__result) return(__result);
    }
}

I have two questions. 1. why does _S_oom_malloc use infinite loop? 2. as we known, _S_oom_malloc will be called when malloc fails in __malloc_alloc_template::allocate function. And why does it use malloc to allocate space?

Anyone can help me? Thanks a lot.

First, the loop is not truly infinite. There are two exits: by throwing a BAD_ALLOC exception an by allocating the requested amount of memory. The exception will be thrown when the current new-handler is a null pointer.

To understand how that can happen, consult e.g. Item 49 from Effective C++. Essentially any new-handler can either

• Make more memory available
• Install a different new-handler
• Deinstall the new-handler (i.e. passing a null pointer to set_new_handler
• Throw an exception
• abort or exit

Second, the reason that it uses the C library's malloc to allocate space is that malloc on most systems is a well-tested and efficiently implemented function. The Standard Library's new functions are "simply" exception-safe and type-safe wrappers around it (which you as a user can also override, should you want that).