RE: Use of strings in new exception implementation (report libstd c++/1972)

[Greg Bumgardner <bumgard at roguewave dot com>]

Then I guess it follows that the member used to store the what-arg string
(in logic_error and runtime_error) must itself support assignment, so using
std::string is probably the easiest and most obvious way to do it.

-g.b.

> -----Original Message-----
> From: Nathan Myers [mailto:ncm@nospam.cantrip.org]
> Sent: Wednesday, February 14, 2001 7:11 PM
> To: libstdc++@gcc.gnu.org
> Subject: Re: Use of strings in new exception implementation (report
> libstd c++/1972)
> 
> 
> On Wed, Feb 14, 2001 at 04:30:08PM -0800, Greg Bumgardner wrote:
> > I'm also curious about the handling of assignment 
> operators.  The standard
> > shows that exception has an assignment operator, even 
> though it has nothing
> > to assign, but none of the derived classes have them.  If 
> we don't design
> > these classes to correctly support assignment, then the 
> operator should
> > probably be declared private.
> 
> The standard mentions assignment and copy ctor when it has something
> to say about their semantics.  The assumption (as documented in 
> chapter 17) is that if you do something under the hood that would
> make the compiler-generated version do the wrong thing, it's your 
> responsibility (as implementor) to cover up by defining one that 
> works right.
> 
> Nathan Myers
> ncm at cantrip dot org
>