Re: GNAT 301a with Win95

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

Vadim said:

"  How about -gnat83 option ?"

The -gnat83 switch does provide greater compatibility with Ada 83, for
example:

   new Ada 95 keywords not recognized
   Numeric Error distinct from Constraint_Error
   error checks for later declarations performed
   etc.

but it does not guarantee complete Ada 83 compatibility. There are some
places where obscure differences in overloading etc are simply not practical
to implement.

Also, note that type Character is always 8 bits, since this was permitted
in Ada 83 (by virtue of the Zandvort WG9 decision) in any case.

If you find cases where you think it would be helpful for the -gnat83
switch to do more, then by all means let us know, but we do not promise
to be able to fix all such cases.

We notice some people using -gnat83 routinely because they are compiling
Ada 83 code. That's probably not a good idea for two reasons:

  1. If there are incompatibilities, then they should be found earlier
	rather than later, since most code will be migrated, or at least
	partially reused, to Ada 95 contexts sooner or later.

  2. Since -gnat83 does not provide a guarantee of Ada 83 semantics, you
	are working in a somewhat ill defined semantic framework when you
	use this option.

We retain it for specialized uses, e.g. running old Ada 83 test suites,
rather than general purpose use.