C++ PATCH: Fix PR 13275

[Gabriel Dos Reis]

Richard Henderson <rth@redhat.com> writes:

| On Wed, Dec 17, 2003 at 05:37:44PM -0800, Mark Mitchell wrote:
| > I didn't think any special handling was required for C, but if special
| > handling *is* required, the same trick should work.
| 
| C also has the concept of integer constant expression.  We don't
| currently implement that notion at all, so the issue hasn't come
| to a head yet.

Just to add a note that C's notion of integer constant expression is
different from that of C++.
(I don't know it affects the implementation of offset of though).

| > Why are we considering reimplementing working functionality that
| > consists of a very few lines of localized changes?
| 
| I guess some object to the exact form of the solution.  I don't
| really care one way or the other, but *definitely* want the same
| solution to work for both C and C++.
| 
| I have interest in this because the C and C++ front ends are
| decomposing &s.x and &a[x] to pointer arithmetic too early, and
| I want to stop that.

I supporte that move. 

-- Gaby