>Hi,
>if an unsupported family is used, the code might do null pointer dereference
>in m !=0 case (th isn't initialized in that case in the switch(family) code).
>Is it really as it seems or do I miss something ? Futhermore, the code
>"pasifies gcc" by setting ip, ip6 & th to NULL - that is just bogus, it leads
>to NULL pointer dereference later. It should return EAFNOTSUPPORT in that
>case IMHO.
>
>IMHO the code should be changed as in the following patch. Is it correct
>or did I overlook something obvious ?

	I agree with first part of your change.  second and third part of
	your change is, I believe, unnecessary (or cosmetic).
	I'll bring in your change into repository.  Thanks.

itojun