Re: Ada News Brief

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

Dave Wood said

"Er, well, rest assured that 'the company' made no such
claim.  The actual press release contained some suitable
adjective, like "virtually", which somehow was dropped
by the press.  But hey, if they're going to screw up our
press releases, I'm glad it's in our favor.  (At least
they didn't say:  ObjectAda includes the full Ada 95 core
language and will compile no Ada 83 code without changes...)
"


Well I am not sure this is in your favor. I think it is important to educate
people away from the expectation that an Ada 83 switch will work perfectly
on their Ada 83 code. There are two reasons.

1. The switch from an Ada 83 to an Ada 95 compiler may involve changes in
implementatoin dependent choices (e.g. the behavior of representation 
pragmas). This is especially likely to be so if you are switching
front ends (VADS to GNAT, or DEC to Rational, or Alsys to TSP).

2. There are subtle changes in semantics, e.g. of overloading, which are
unlikely to be copied exactly. It makes no sense to have two overloading
algorityhms, where the only function of one of them is to implement
obscure Ada 83 rules that have been judged (a) undesriable and (b)
too obscure to worry about compatibility issues.

In addition, usually it is only worth trying to worry about correct Ada 83
rules, there seems little point (and it would be tough) to diagnose all
possible Ada 83 semantic errors.

One interesting criterion would be to see if a compiler can 100% validate
against 1.11 with its Ada 83 switch. GNAT certainly makes no such claim.

it is certainly true that advertisers and press are likely to be over
enthusiastic in their claims here (I have seem literature from more than
one vendor mislead on this point, not so much deliberately, but simply
as the result of abbreviation, advertising often gives false impressions
by not being able to give details).