2.91.66 crashes and hangs

2.91.66 crashes and hangs

[Glen McCluskey]

Below are some internal errors, hangs, and bad assembly output against
egcs-2.91.66 using Win32 NT 4.0.

I write test suites for a living, and these are taken from a larger set
of about 80 similar problems that the suites manifest.

If you're interested in finding out more about the suites, please
contact me or see:

        http://www.glenmccl.com/cppsuite.htm

Thanks very much.

Glen McCluskey

------------------------------------------

namespace A1 {
        typedef short T;
        void f(int);
        class B {};
        int x[59];
        static union {
                int y;
        };
}

namespace A2 {
        typedef short T;
        void f(int);
        class B {};
        int x[59];
        static union {
                int y;
        };
}

using namespace A1;
using namespace A2;

-------------------------

template <class T> struct A {};

A<double> a;

namespace B {
        using ::A;
}

-------------------------

namespace A {
        void f() {}
        struct B {
                friend void f() {}
        };
}

-------------------------

typedef unsigned int size_t;

extern "C" void exit(int);

struct A {
        template <class T> void operator delete(void* p, T i, ...)
        {
                exit(1);
        }

        template <class T> void* operator new(size_t s, T i)
        {
                return new char[s];
        }

        A() {throw -37;}
};

int main()
{
        try {
                new (1234) A;
        }
        catch (int i) {
                if (i == -37)
                        exit(0);
        }
        exit(1);
        return 0;
}

-------------------------

template <class T> struct A {
        int x;
        A() {x = 37;}
        template <class U> A(U u) {x = u.x;}
};
void f()
{
        volatile A<double> a;
        A<double> aa(a);
}

Re: 2.91.66 crashes and hangs

[Martin v. Loewis]

> Below are some internal errors, hangs, and bad assembly output against
> egcs-2.91.66 using Win32 NT 4.0.

Thanks for your good report. Several of these problems have been fixed
in the current development snapshot of egcs; I'll look into the
remaining namespace problem soon.

Regards,
Martin