[kde-solaris] Crash in kmail (kdepim-3.1.93)

[James F. Hranicky]

On Mon, 2003-11-24 at 07:27, Eva Brucherseifer wrote:
> On Montag, 24. November 2003 12:53, James F. Hranicky wrote:
> > On Mon, 24 Nov 2003 02:22:34 -0400
> >
> > Stefan Teleman <steleman at nyc.rr.com> wrote:
> > > Should _never_ use C-style casts, but dynamic_cast<>(), and check for
> > > returning 0.
> 
> You are casting "this" and  this should never be 0 - strange thing.

Well, I am not, Trolltech is :-> Also, it isn't 0. It's a valid 
address until the C-style cast, then the resulting value of c is an
invalid address.

> >
> > When I try changing the line to this:
> >
> >     const QListViewItem* c = dynamic_cast<QListViewItem * const>(this);
> 
> 
> const QListViewItem* const c = dynamic_cast<QListViewItem * const>(this);
> 
> or 
> 
> const QListViewItem* const c = this;
> 
> should work. The pointer and the object are both const. 

The first line gives this:

   widgets/qlistview.cpp:3815: error: cannot dynamic_cast `this' (of
type `const 
   class QListViewItem* const') to type `class QListViewItem* const' (
   conversion casts away constness)

and both lines give this:

   widgets/qlistview.cpp:3818: error: assignment of read-only variable
`c'

> Still casting away the const shouldn't give any seg faults. I'd expect the 
> problem to be something else. This is only intuition though, since I don't 
> know the code.

Yes, more than likely. Perhaps I should try compiling with gcc's bounds
checking? Has anyone tried this to see how many errors are caught in
QT/KDE?

Jim