From: Jim Weirich <jweirich@...>
Date: 2002-07-17T14:38:11+09:00
Subject: Re: OT: FORTRAN (Was Re: TkText no -textvariable workaround?)

>>>>> "Hal" == Hal E Fulton <hal9000@hypermetrics.com> writes:

    Hal> I think all or most compilers allowed this
    Hal> once upon a time.

    Hal> Actually, I think this was so long ago it was not 
    Hal> considered a bug. Probably not even a feature. 
    Hal> More of a why-would-you-do-that-anyway.

As I recall, it is because FORTRAN traditionally used
pass-by-reference semantics for passing variables to subroutines.
This allowed you to write ...

      SUBROUTINE SET2(IVAR)
      IVAR = 2
      RETURN
      END

And use it like this ...

      CALL SET2(J)

To set the variable J to the value 2.  Unscrupulous programmers could
also do ...

      CALL SET2(1)

Since the constant 1 was often stored in a particular memory location,
this would change the value stored at that location, so later
references to the literal 1 would really be the value 2.

Not all compilers had this (mis)feature, but it was a common artifact
of a fairly traditional FORTRAN implementation.

This reminds me of the ability in FORTH to define words
(i.e. subroutines to non-FORTHers) named with a literal number.  The
number 1 was often defined as a word that pushes a literal 1 on the
stack (word references take a single cell while literals take two
cells in a tradition threaded code FORTH implementation).  That meant
you could do things like this ...

   : 2 1 ;   ( Define a word named "2" that returns the value 1 )

Of course, just because you could do that, doesn't mean you should.

-- 
-- Jim Weirich     jweirich@one.net    http://w3.one.net/~jweirich
---------------------------------------------------------------------
"Beware of bugs in the above code; I have only proved it correct, 
not tried it." -- Donald Knuth (in a memo to Peter van Emde Boas)