Re: Emacs and Ada.

[dewar@gnat.com (Robert Dewar)]

Ted Dennison<dennison@telepath.com> wrote in message news:<xQeC7.1496$xS6.2133@www.newsranger.com>...

> Often times a LRM reference wouldn't be particularly 
> helpful to anyone.

Note that GNAT *does* include RM references in those (relatively
unusual) cases where we think it is helpful to do so.

I have often found that at an abstract level people think
it is a good idea to give RM references, but in the past threads I
have always issued the suggestion that people
show an example of a GNAT error message where the RM reference would
be helpful, and so far the number of
examples received = ZERO. 

If you *do* find such a case, by all means let us know, and we will
have a look. We have no objection in principle to
RM references where they are helpful, we just don't find
many such cases.

As Ted very usefully points out it is better to see
  
    1. package x is
    2.    First_Segment : Integer := 3;
    3.    Second_Segment : Integer := Firt_Segment + 1;
                                      |
       >>> "Firt_Segment" is undefined
       >>> possible misspelling of "First_Segment"

    4. end;

(the gnat output) than

      >>> "Firt_Segment" is undefined
      >>> identifiers must be declared before use see RM ..

and this is a perfect case where an RM reference is utterly
and completely gratuitous.

Equally if you write
       
       x := a + b

A message saying missing semicolon is much better than
something like

   incorrect syntax for assignment statement, see RM ...

But I hereby reissue the challenge perhaps this time someone will rise
to it, to give an example of a GNAT
message which does not have an RM reference where an
RM reference would make the message clearer to the user.