Re: RFC: C++ PATCH to use MUST_NOT_THROW_EXPR for throw() by default

[Jason Merrill <jason at redhat dot com>]

On 05/14/2010 04:53 PM, Mark Mitchell wrote:
> Whether desirable or not, the standard semantics are what untold
> zillions of lines of C++ code expect.  This is a subtle change that will
> not necessarily be visible until run-time when some exception that the
> user expected to propagate does not do so.

There is no change in whether an exception propagates or not; no 
exception can propagate out of a throw() function.  The only difference 
is whether it calls the unexpected handler or the terminate handler 
before killing the program.  So the only difference would be whatever 
diagnostic code a user put in one function and not the other--and I 
can't imagine why anyone would do that.  I am very sympathetic to the 
general point, but this particular difference seems insignificant to me. 
 If you aren't convinced yet, I'll concede the point.

Jason