PATCH rpc server

Jean Francois Micouleau Jean-Francois.Micouleau at
Mon Jan 21 23:30:02 GMT 2002

On Tue, 22 Jan 2002, Martin Pool wrote:

> On 22 Jan 2002, Tim Potter <tpot at> wrote:
> > When the structure is unmarshalled and the pointer is zero, set the
> > struct foo *bar to NULL.  When marshalling, if struct foo *bar is NULL
> > then write out a zero pointer, else write out a one (or 0xdeadbeef
> > (-:).
> I agree with Tim about removing the ptr_ members, but not about
> marshalling them as 0xfeedb0b0.
> I wouldn't be surprised the NT unmarshalling code does actually
> implement the NDR spec for nonunique pointers: if the same opaque
> value is passed for two different pointers then they're meant to be
> combined into a pointer to the same object on the server, and the
> referent of the pointer is only marshalled once.

then that would be a known fact by most of us ;-)

yes NT does implement non unique ptrs, on the paper at least, never
verified the wire version.

But for the common pipes that samba implements, they don't use non unique

> So I think it would be equally easy and more reliable for us to send
> the bottom 32 bits of the local pointer value.

putting a well defined value for ptr is useful when you debug from on wire
traces, you can compare more easily and find alignment errors.


More information about the samba-technical mailing list