python generation problems with pidl

[tridge at samba.org]

Hi Jelmer,

Thanks for all your help recently with the python generation in pidl!

I've found a way to avoid the crashes that we saw recently when
running the testrpc test. The approach is to use a new talloc call
talloc_is_child() to allow is to detect is the mem_ptr element of a
python talloc object is itself a talloc pointer. I've pushed the
changes to my pidl-wip branch and I'd appreciate your comments.

The pidl-wip branch also has some fixes for NULL pointer checks, by
adding some additional wrapper functions in pytalloc.c. I'm not sure
that is the best place for those wrappers though, do you think they
would be better in pyrpc_util.c perhaps?

With these changes the only valgrind errors from testrpc are there
types of errors:

  == Source and destination overlap in memcpy(0x75fb190, 0x75fb190, 16)

this happens when we are copying an object onto itself. We probably
should just skip the copy if the pointers are equal, but I'm not sure
if we should also skip the reference count change.

Cheers, Tridge