On Friday, August 7, 2020, 12:46:56 PM CDT, Walt Karas <wkaras_at_[hidden]> wrote:

On Thursday, August 6, 2020, 5:48:52 PM CDT, Ville Voutilainen <ville.voutilainen_at_[hidden]> wrote:
On Fri, 7 Aug 2020 at 01:36, Walt Karas via Std-Proposals
<std-proposals_at_[hidden]> wrote:

> > WK: Good point, they should be noexcept if the corresponding constructor is noexcept (and the destructor).
> And if it isn't noexcept, what do you do?
> WK: Under the current Standard, if a base or member assignment op is noexcept(false), isn't the default assignment op noexect(false)? That is what I'm thinking, noexcept is inferred for the default.

That's not the point. Look:

C & operator = (C const &rhs)
{
  if (this != &rhs)
  {
    this->~C(); // #1, now the object is dead
    ::new (this) C(rhs); // #2, if this throws, it stays dead

  }
  return *this;

}

void some_automatic_scope() {
  C c1;
  C c2;
  c1 = c2; // if #2 threw, your program is now dead (because there
will be an attempt to destroy c1, and it's already dead), and there's
no recovery from that.
}

You can't do a destroy+placement-new if the placement-new can throw.

WK: OK, also good point. If the constructor can throw, then conceptually the default assign can be like this: https://godbolt.org/z/jvdf53
But then the existence of a default copy assign would be dependent on the existence of both the copy and move constructor. And, this is a lot of overhead.

WK: Oops, should be: https://godbolt.org/z/Eo3n59

Go read Sutter's Exceptional C++
and More Exceptional C++, and try again.