From: Alexander Terekhov (terekhov_at_[hidden])
Date: 2003-09-29 08:59:15
John Maddock wrote:
>
> >Many modern systems implement longjump via "forced unwinding"
> >("Forced unwinding is a single-phase process (phase 2 of the normal
> >exception-handling process").
>
> They may well do so, but that would require C++ exceptions to be
> supported (or at least the stack unwinding part) wouldn't it?
Yep. Unwinding part for sure.
>
> BOOST_NO_EXCEPTIONS is generally used on systems like WINCE where
> there is no C++ exception support at all.
Wrong approach. throw()-nothing interfaces (and implementations) is
the right approach to address this problem. It's a complete rewrite
in many cases (and it runs much slower ;-) ), however.
regards,
alexander.