On: 17 Oct 03  13:28:46 Pascal Schmidt wrote to Roger Scudder:

 >  RS> because something will probably be  cleaned up when the program ends 
 >  RS> doesn't mean it is a good idea to make it a way of life.  :-)

 > Yeah sure, but there are exceptions. I have a small scene rendering tool which 
 > uses data structures like this:
 > - top is a scene_t
 > - this has a linked list of filter_t and a linked list of element_t
 > - a filter_t includes a linked list of when_t
 > - an element_t includes a linked list of filter_t
 > This stuff is constructed by reading an XML scene description at
startup, with all 
 > the list elements dynamically allocated. After that, the data never
changes while 
 > the program runs. I have not bothered with writing code to traverse
three levels 
 > of linked lists and freeing all the element structures.
 
 To me there is no exception.  I would write the cleanup code.  If you
compiled you code using a memory diagnostic library I am sure it would flag
the lack of matching calls to free.

 Obviously we have differing opinions here.  I don't think that the startup
code, or the OS, or the libraries do cleanup as a service to the programmer
to make things easier on him.  I think they do it because they were written
to be robust.  A robust system makes sure that it has checked for things
like memory leaks and tries to correct them.  Robust code doesn't leave
loose ends hanging for the next piece of code to clean up.  

If that is the way you choose to write code, that's on you.  I just don't
like to do it that way.  In a way it's a philosophical thing.  My
philosophy is this; never make assumptions, and leave things the way you
found them.

-Roger
  

--- Spinone v0.1.79 Win32