Patch: RFA: fix PR

[Andrew Haley]

Tom Tromey writes:
 > >>>>> "Andrew" == Andrew Pinski <pinskia@physics.uc.edu> writes:
 > 
 > >> This patch fixes the bug in the simplest way, by adding a check for
 > >> null_pointer_node when creating the anonymous constructor.  (We can't
 > >> do the check in the caller as that will cause problems when calling an
 > >> ordinary constructor with a `null' argument.)
 > >>> Test case included.  Ok?
 > 
 > Andrew> I would be careful here because INTEGER_CST are shared and I
 > Andrew> assume that null is an INTEGER_CST.  Maybe the better fix
 > Andrew> would be change null_pointer_node to represented a different
 > Andrew> way.
 > 
 > What is happening here is that we have a list of types.  The patch
 > just causes us to replace one type in the list with a different type.
 > I don't think it is possible to accidentally modify null_pointer_node
 > here.
 > 
 > FWIW, I suspect a change to the representation of null_pointer_node
 > would be fairly major.  I think a lot of code knows about how it is
 > represented.
 
It does.  There is only a single object of type (void*), and that is
the null pointer.  It would be unwise to change it at this time.

Andrew.