Re: g++ 2.95.2 does not provide useful error message when dynamic_cast fails

[<llewelly at dbritsch dot dsl dot xmission dot com>]

>
> "Martin v. Loewis" wrote:
>
> > > According to the code (cp/rtti.c: build_dynamic_cast_1), almost all
> > > errors result in
> > > the same message:
> > >     cannot dynamic_cast `%E' (of type `%#T') to type `%#T
> >
> > > This message is not particularly explanatory. In particular, it took
> > > me a couple of hours to figure out that a dynamic cast was failing
> > > because the type of the expr being cast did not have a vtbl. (I
> > > figured out the problem only by looking at the above gcc source
> > > code.)
>
> > Thanks for your bug report; I fail to see the bug, though. It is not
> > the compiler's job to teach the user C++.
>
> The bug is that the compiler is producing a poor message. Admittedly,
> it is not an error in code production, but quality software handles
error
> conditions as gracefully as it handles correct input. Would you use a
> compiler
> that produced the single message "User Error" on all syntactic and
> semantic
> errors?
>
> Do you still fail to see the bug? If so, perhaps the job of handling
bugs
> should be given with a less narrow view of what is involved in quality
> software.
>
> It is not the bug-reporter's job to teach the bug-receiver proper
software
>
> quality techniques. It is not the bug-receiver's job to insult the
> bug-reporter.
 
I would like to point out that Martin Loewis
  
  (1) Responds to bug reports promptly.
  
  (2) Is almost always  willing to provide clear explanations of why
      something is or is not a bug.

  (3) Is almost always polite.
  
I do not believe Martin intended to insult you.

As a user of gcc, I feel Martin provides a most valuable service, and I   
  greatly appreciate his work.
  
> 
> > Of course, if you can suggest specific patches to improve the quality
> > of the g++ diagnostics, they would be much appreciated. Please send
> > them to gcc-patches@gcc.gnu.org.
> 
> Done.
>     
>     Steven

Of course I will always appreciate better error messages, but I wanted
  to point out that Martin does an excellent job of responding to bug
  reports.