p_standard_unit seems wrong with 22.0 compilers

As part of a code generator, I am manipulating fixed point types. With more recent compiler wavefronts (supporting 128 bit integers), we have the following definition in system.ads:

 Max_Mantissa          : constant := Standard'Max_Integer_Size - 1;
  Fine_Delta            : constant := 2.0 ** (-Max_Mantissa);

where it used to be

Max_Mantissa          : constant := 63;
Fine_Delta            : constant := 2.0 ** (-Max_Mantissa);

My code generator is chocking at the moment because it doesn't know about the max_integer_size attribute. So I am trying to use p_standard_unit (from any node), which doesn't seem to include that definition either.

(presumably related: call p_eval_as_int on the expression for Fine_Delta raises an exception).

I also noticed that p_standard_unit doesn't include the new Long_Long_Long_Integer (which I do not need at this stage, but that shows it isn't up-to-date with the compiler).

Thanks
Manu

Start by locating the implementation of p_standard_unit and compare its definitions with system.ads, especially Max_Integer_Size and Long_Long_Long_Integer. Reproduce the p_eval_as_int failure on Fine_Delta, then verify that p_standard_unit and expression evaluation handle the compiler definitions described in the issue.