Re: Elaborate compile-time error messages (was: GNAT Stream Read ...)

[sb463ba@d250-hrz.uni-duisburg.de (Georg Bauhaus)]

Marin David Condic (condicma@bogon.pwfl.com) wrote:
: Larry Kilgallen wrote:

: > Is there some set of programmers who get
: > really mad at having the compiler be smarter
: > than they are about the nature of a problem.

(Two questions below.)

Given two source files, both plain wrong (a.k.a. nonsense),
admittedly, and differing only in the spelling of one
character, namely the P in these two lines:

<   N : Integer := Private;  -- capital "P" (yes, wrong anyway)
>   N : Integer := private;  -- small "p" (yes, wrong anyway)

----------------
package Emsg is -- not: emsg! (small "e")
   N : Integer := Private;  -- capital "P" (yes, wrong anyway)
end Emsg;
----------------

gnatf -v -ds emsg.ads

GNATF 3.09 (970121) Copyright 1991-1996 Free Software Foundation, Inc.

Checking: emsg.ads (source file time stamp: 1998-12-29 02:42:43)
Source recreated from tree for Emsg (spec)
------------------------------------------


package emsg is
   n : integer := private;
end emsg;


     2.    N : Integer := Private;  -- "P" (yes, wrong anyway)
                          |
        >>> reserved word "private" cannot be used as identifier

 3 lines: 1 error

Compilation exited abnormally with code 1 at Tue Dec 29 03:42:43

To me, this error message is quite satisfactory.  Then

----------------
package Emsg is -- not: emsg! (small "e")
   N : Integer := private;  -- small "p" (yes, wrong anyway)
end Emsg;
----------------

gnatf -v -ds emsg.ads

GNATF 3.09 (970121) Copyright 1991-1996 Free Software Foundation, Inc.

Checking: emsg.ads (source file time stamp: 1998-12-29 02:30:36)
Source recreated from tree for Emsg (spec)
------------------------------------------


package emsg is
   n : integer := <error>;
end emsg;


     2.    N : Integer := private;  -- "P" (yes, wrong anyway)
                         |
        >>> missing operand

 3 lines: 1 error

Compilation exited abnormally with code 1 at Tue Dec 29 03:30:38

So, this time I understand that ":=" is missing an operand.
(right?) I explained to myself that these two very different
error messages might be the result of GNAT checking style.
Is this true?

QUESTION 1:
Why does GNAT behave so very differently when things
are spelled differently? (Try compiling with "emsg"
replacing "Emsg" in the above examples.)

I came across this when brooding over an error message that
informed me of a missing operand and an "is" that should be
":=", which felt right (
   N : Integer is private;
   -- operand missing between "Integer" and "is",
   -- then "is" should be ":=".
Is "is" an operator?  GNAT 1.83 just wants a ";" there.).
But, not knowing  how a compiler works internally and the
formal aspects of languages in general, I was wondering where
the operator was. "is"? So, here is my second question:

QUESTION 2:
Is there a chance that someday GNAT will tell me *which*
operator is missing what? (Or have I overlooked something?)
More generally: Is there a way to have GNAT shout *why* it
was expecting something?

(Let me note (1) that it's still a pleasure working with GNAT
while thinking of C/Perl and what the compilers say when you
forgot to place a ';' somewhere. And (2), I see that I
had better know how compilation is done in order to understand
error messages, much like if I'm driving a car through the
desert and know something about motors and brakes.
But then, some people just want to ride somewhere...)

with best regards,
  Georg