Thanks for the review, Jan.

> In general I do not think we can ever fix it all on top of the current
> infrastructure.  These OFFSETs, EMBEDDED_OFFSETs etc. should be dropped, there
> should be support for discontiguos set of memory snapshot associated with
> struct value (so that it supports for example even virtual method tables or
> slices of multidimensional Fortran arrays).

Yeah, I agree. Although, I have been wondering how we could do without
offset, but something along those lines will be necessary (my naive
questions was: why not add offset to address and be done with it, but
of course, address might be a DWARF expression, for instance).

I checked in the patch with the missing function comment (see
attached).  But I have one question:

> > @@ -2310,18 +2309,22 @@ ada_value_primitive_packed_val (struct value *obj, const gdb_byte *valaddr,
> >  
> >    if (obj != NULL)
> >      {
> > -      CORE_ADDR new_addr;
> > +      long new_offset = offset;
> 
> Nitpick - LONGEST (Siddhesh Poyarekar is working on a more thorough fix).

I know you said it's a nitpick, but I don't understand why you think
it should be LONGEST. "offset" itself is a long, and the only use
of that new_offset is for set_value_offset, which takes an int. Are
the changes you are referring to going to make that a LONGEST?

I'll fix that as a followup patch if you confirm the above.

Thanks!
-- 
Joel