In article <36DD54D2.141B@ddre.dk>,
  hm@ddre.dk_nospam wrote:
> I have always found it amazing that the validation suites
> does not care if IO from the keyboard or to the screen
> distorts about half of the available graphic characters.
>
> EG Put('�') to see a Pound_Sign, on the screen.
> American Chauvinism !)


If you have found this amazing, it is because you do not
understand the situation.

The RM has *nothing* at all to say about what characters
look like in the external world. Absolutely NOTHING. A
validated compiler has no requirements at all in this area.

A compiler which used B as the representation of letter-A
and A as the representation of letter-B in the external
world would be annoying and silly, and undoubtedly rejected
by the marketplace, but not necessarily invalid if the
internal codes were correctly represented.

This is a VERY common misconception (that external
characters must correspond to some preconceived conception
of them).

To see just how wrong this is, consider that a compiler can
be fully conformant with the 10646-based requirements for
wide characters, without the ability to output or input a
character in anything like its Han form.

So if the validation process DID complain about the sort
of thing you mention, it would mean that the validation
process was seriously broken.

Of course when you are acquiring a compiler and an
operating system, you must make sure that external handling
of character data meets your requirements.

By the way, in practice this is 100% an operating system
issue, and not at all a compiler issue. The compiler will
handle Latin-1 input and output, and how it got there is
not the compiler's business.

GNAT does support many other character sets, including
multiple PC code pages, and other Latin-n sets, but such
support is in fact a language extension, nothing at all to
do with the RM requirements.


-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own