Re: Endian and Ada

[dewar@cs.nyu.edu (Robert Dewar)]

Mike asks

"I'm taking a computer architecture class and my eyes have been opened to the
incredible lack of progress that computer science has made in the area of
architecture-independant programming.  One of the main problems that my
professor likes to point out the endian problem and the fact that so much
software is *not* endian-independant (due mainly to the widespread use of C
and C++).  However, he is not familiar enough with Ada (nor am I) to know if
Ada is endian-independant.  I'm almost positive that it's *possible* to make
endian-dependant code using unchecked programming (necessary for systems
writing).  I was wondering, however, if the day-to-day Ada software out
there is written endian-independant (i.e. the language facilities used in
most day-to-day programming don't depend on whether your architecture is
"big" endian or "little" endian.)"


To a first approximation, NONE of the semantics of Ada depend on the
endianness, you would have to work hard to make an Ada program that
worked BE and did not work LE, or vice versa.

The issue arises when transporting data between targets of different
endianness.

One nice approach is to use the stream facilities of Ada in conjuncction
with a 
target independent representation such as XDR. A future version of GNAT
will provide the option of using XDR for stream representation to
facilitiate this approach (you could program it yourself, but it would
be nice to have it built in).