Re: Patch: remove `final' hack in String.toString()

[Corey Minyard <minyard at acm dot org>]

This would be nice, but to maintain bug compatibility with JDK we
probably have to analyze this on a case-by-case basis.  I know, for
instance, that in JDK passing a null pointer as the address to the
DatagramPacket() constructor or to DatagramPacket.setAddress() will
work fine under JDK.  I also know of some code that actually requires
this behaviour to work properly.

Corey

Tom Tromey <tromey@cygnus.com> writes:
> Bryce> Now that we have a fix for PR/2, we don't need this dumb hack
> Bryce> in String.java. Checked in.
> 
> There are other places that we do something similar.
> For instance, in BitSet.java:
> 
>       if (bs == null)
> 	throw new NullPointerException ();
>       int max = Math.min(bits.length, bs.bits.length);
> 
> We never had a very well-defined policy for this stuff.
> Right now I'm inclined to say that we should also explicitly throw a
> NullPointerException when the value is not used,
> e.g. java.io.File.File:
> 
>   public File (String p)
>   {
>     if (p == null)
>       throw new NullPointerException ();
>     path = p;
>   }
> 
> Comments?
> The idea here is that on native platforms we can rely on what we have
> now.  On platforms without an MMU we'll have to modify the compiler to
> do explicit checks everywhere.
> 
> Tom