TITLE: results of new

(Newsgroups: comp.lang.c++.moderated, 24 Apr 97)

UNKNOWN:

> > Not to mention that there are operating systems where operator new does
> > not "fail", in the sense of throwing an exception or returning NULL, but
> > your code will core dump if you attempt to access the apparently valid
> > returned pointer.


DEROCCO: "Paul D. DeRocco" <strip_these_words_pderocco@ix.netcom.com>

> What exactly are they featuring? What is the pointer that is returned in
> this case?

KOENIG: ark@research.att.com (Andrew Koenig)

The idea is that when you allocate more memory than the system can handle,
it is conceivable that some other process will free some memory before
you get around to using yours.  So the operating system allocates
appropriate page table entries, gives you back a pointer to where
the memory would be in your address space if the system could have
allocated the memory, and waits.

If you use the memory (or the part of it that the system could not
allocate), it checks again to see whether it can now allocate it.  If
it succeeds, no harm done.  If it fails, the system now has a problem.