C++ defect reports: how to behave

[Joe Buck]

On Wed, Jun 25, 2003 at 07:23:37PM +0200, Giovanni Bajo wrote:
> I am interested in which is the current policy with respect to implementing
> the C++ defect reports. I noticed that some defects are implemented in the
> code. Others are implemented but "commented away" because they could
> break/modify the semantic of existing code (e.g. DR295). Is there an
> official policy?
>
> If not, I suggest we implement all the defect reports with TC1 status
> into -stdc=c++98 and gnu98, and that we implement all the defect reports
> with WP status into a new dialect, -stdc=c++0x.

It's not clear whether that's quite right.  Clearly if an issue with WP
status conflicts with current use, it can't be enabled by default.
However, it seems that some of the WP issues are just clarifications,
and in cases where our compiler doesn't match there is no language in
the existing standard requiring our current behavior (though of course
we do have to worry about ABI issues).

It might be worth opening up a new dialect as you suggest and growing it
as the committee adds any extensions.  But I'm not sure that the "WP"
designator and -std=c++0x should be the same.