screw up PC relative offsets to _GLOBAL_SYMBOL_TABLE_ (caused by me not quite
understanding the inner workings of this phenomenon called `obstack').
Thanks to Thomas Eberhardt for creating rare and unpredictable circumstances.
Changes to obj-aout.c to handle non-absolute `.size' expressions and to set
the `n_other' field of symbols to the symbols type, as given by `.type'
directives.
pk