Problems with dynamic arrays initialization

[Olivier Galibert <Olivier dot Galibert at mines dot u-nancy dot fr>]

cd2 isn't completely clear about whether new Class[42](666);
is a vlid syntax to allocate an array and initialize it
with a non-default constructor. It looks like it is valid
though. egcs (971122) has two problems with that:

1- Abuse of the copy constructor:

-O<-----------
class C {
public:
  void f();
};
 
class A {
public:
  A();
 
private:
  A(const A&);
};
 
class B : public A {
public:
  B(C *);
};
 
void C::f()
{
  new B[10](this);
}
-O<-----------

This gives:
weird.cc: In method `B::B(const class B &)':
weird.cc:11: `A::A(const class A &)' is private
weird.cc:21: within this context

But there is _no_ reason at all to call the copy constructor.

Second problem:
-O<-----------
class A {
public:
  A(int, int);
};

int main()
{
  new A[10](42, 666);
  return 0;
}
-O<-----------
parse.cc: In function `int main()':
parse.cc:8: no matching function for call to `A::A (int)'
parse.cc:4: candidates are: A::A(const A &)
parse.cc:3:                 A::A(int, int)

Looks like the initializer is taken as a comma expression
instead of a parameter list.

For the curious, both of these problems where detected
while compiling kde. Regression from 2.7.2 ? (I don't have
it anymore)

  OG.

PS: I'm on egcs@ but not on egcs-bugs@