Description of Problem:
This short C++ fragment causes gcc 2.96-98 to print "Internal error: 
Floating point exception" when compiled with -O2 (no FPE with -O0).

(yes, I know this code won't work correctly even if it did compile...)

I've seen other floating point exception bugs in Bugzilla, but they were 
all for earlier versions of gcc, and marked CLOSED.

Steps to Reproduce:
1. g++ bugcase.cpp -c -O2

// bugcase.cpp

template<class T>
class MyArray {
public:
	T* p;
	int nused;
	
	void append(T foo)
	{
		p[nused++] = foo;
	}
};

typedef float v3_t[3];

void test(v3_t v)
{
	MyArray<v3_t> a;
	a.append(v);
}