From kde-core-devel  Tue Nov 02 19:59:36 1999
From: Harri Porten <porten () tu-harburg ! de>
Date: Tue, 02 Nov 1999 19:59:36 +0000
To: kde-core-devel
Subject: Re: KPresenter
X-MARC-Message: https://marc.info/?l=kde-core-devel&m=94157378018526

Waldo Bastian wrote:
> 
> On Tue, 02 Nov 1999, Stephan Kulow wrote:
> > Reginald Stadlbauer wrote:
> > >
> > > On Mon, 01 Nov 1999, Dirk A. Mueller wrote:
> > > >Reginald Stadlbauer <reggie@troll.no> wrote:
> > > >
> > > >> KPresenter uses dynamic_casts. SO does it mean I have to switch all
> > > >> back to C-casts now (would be lots of work :-(( )
> > > >
> > > >errm, if const_casts would work, why don't use them? they're faster! :)
> > >
> > > Who says const_casts would work.
> > >
> > > I mean, you can always live without C++ casts, because
> > >
> > > dynamic_cast<AClass>( aPtr )->doSomething()
> > >
> > > can be replaced by
> > >
> > > ( (AClass*)aPtr )->soSomething().
> > >
> > > Itīs just that the first (C++) solution is the better one.
> > >
> > Hmm, but dynamic_cast will return 0 while (AClass*) will just
> > do _something_. :)
> 
> So if you don't check anything, a dynamic_cast'ed pointer will fail the
> first time after it gets dereferenced and the C-styles casted pointer
> will happily continue to appear valid and will most like result in a
> crash further on in the program.
> 
> Question: Which type of fault is easier to debug?

Additionally the C++ type casts are much stricter, i.e. disallow
accessing private members by casting to a base class that was inherited
"private".

Harri.