[boost] Re: std::expected_exception

From: Alexander Terekhov (terekhov_at_[hidden])
Date: 2003-10-11 09:36:43

Peter Dimov wrote:
[...]
> > ES should have no runtime cost other than some extra info for the
> > search phase. OTOH, enable/disable burns cycles (keystrokes aside
> > for a moment) no matter whether you hit some cancelation point
> > *with cancel request pending* or not. Well, there's a downside...
> > of course. It is just an additional/alternative tool, so to speak.
> 
> I'm not seeing the burned cycles (++fs:[8] looks pretty fast) but I may be
> missing and/or misunderstanding something. :-)

pthread_setcancelstate() "can be" expensive. http://tinyurl.com/qk3p

> 
> throw() specs have a "required stack frame" cost.

It might be better to burden the callers, I think. Consider:

void x() throw() {
 throw 0;
}

void y() throw() {
  x();
  /**/
  x();
}

void z() {
  y();
  throw 0;
}

A "barrier" [for search phase] can be injected in z() only. In a way, 
it's sort of:

void z() {
  try { 
    y(); 
  } 
  phase_one_catch(...) { 
    std::unexpected(); 
  }
  throw 0;
}

Oder?

regards,
alexander.