KRB5 1.4 vs. KRB5 1.3.6 on AIX 5.2 (fwd)

[Ken Raeburn]

On Apr 26, 2005, at 13:40, Milton Turley wrote:
> I have done some research on this problem.  The host resolve code does 
> not complete successfully.  I have traced the error to the routine at
> /kerberos/mit/krb5-1.4/src/lib/krb5/os/dnsglue.c.   In the routine a 
> call is made to res_ninit to provide dns information for the kdc.  
> res_ninit updates the structure at statbuf but overlays memory 72 
> bytes past the structure boundary.

If we change statbuf to something like:
   struct { struct __res_state s; char pad[72]; } statbuf;
or perhaps:
   struct { struct __res_state s; char pad[72]; } statbuf1;
   #define statbuf (statbuf1.s)
does it work then?

Do the AIX 5.2 headers declare res_ninit and indicate what type should 
be used for the first argument?

> I have opened a problem with IBM on res_ninit.  The problem is not yet 
> resolved but IBM suggested using a malloc for the structure instead of 
> static storage.  I am trying to get IBM to correct the error in 
> res_ninit.  No resoultion yet.

Well, technically, we're using automatic storage, not static, but it 
seems to me that the problem is not knowing the size and/or type to 
allocate, not stack vs heap allocation.  (Though heap allocation would 
change the failure mode caused by an overrun, it wouldn't by itself get 
rid of the overrun.)

Ken