Path: utzoo!attcan!utgpu!jarvis.csri.toronto.edu!rutgers!cs.utexas.edu!uunet!mcvax!hp4nl!botter!star.cs.vu.nl!ceriel
From: ceriel@cs.vu.nl (Ceriel Jacobs)
Newsgroups: comp.lang.modula2
Subject: Re: diff -c pim3 pim4
Message-ID: <2800@ceriel.cs.vu.nl>
Date: 26 Jun 89 07:59:28 GMT
References: <1291@infbs.UUCP>
Reply-To: ceriel@cs.vu.nl (Ceriel Jacobs)
Organization: VU Informatica, Amsterdam
Lines: 91

Some time ago (in March 1989), I posted the following list of
inconsistencies in PIM-4:

... (start of list)
For instance, there an inconsistency between paragraph 9.2 and paragraph
10.1. In paragraph 9.2, page 144, we read:

   ... The types of corresponding actual and formal parameters must be
   *compatible* in the case of variable parameters ... (emphasis mine)

but in paragraph 10.1, page 149, we read:

   ... The type of each formal parameter is specified in the parameter list.
   In the case of variable parameters it must be *identical* with its
   corresponding actual parameter ... (again, emphasis mine)

Also, in chapter 12 of the book, paragraph 12.1, page 52, on variable
parameters, we read:

   ... The types of corresponding formal and actual parameters must be
   the same ...

In paragraph 12 of the Report, page 153/154, we read:

   ... The type ADDRESS .... is compatible with all pointer types ...

   ... If a formal parameter is of type ADDRESS, the corresponding actual
   parameter may be of any pointer type, even if the formal parameter is a VAR
   parameter ...

If the types of the corresponding actual and formal parameter only have to be
compatible, the last sentence quoted is a useless remark.

Apparently, the texts of the 2nd and the 3rd edition got mixed up a bit.
This supposition is supported by the following quote, from
paragraph 14, page 156:

   ... Opaque export is restricted to pointers and to subranges of standard
   types ...

This was in the 2nd edition. The 3rd edition limits opaque export to
pointers only.

Some more support for this supposition is found in Appendix 1.
It contains the syntax of Modula-2, as it appeared in the 2nd edition,
not the 3rd! This obviously is a mistake, because this grammar does not
correspond to the pieces of grammar quoted in the Report. Also, the syntax
diagrams in Appendix 4 are those of the 3rd edition, and thus do not
correspond to the syntax in Appendix 1.

Another problem is caused by the preface to the 4th edition.
On page 5, we read:

   ... the definition of the language has also undergone a few minor adaptations
   in the area of Standard Functions:

	FLOAT(i) and CHR(i) accept an argument of type INTEGER
	TRUNC(x), HIGH(a), ORD(ch), and SIZE(T) are of type INTEGER

However, in paragraph 10.2 of the report, page 150, we read:

   ... ORD(x)          ordinal number (of type CARDINAL) ...

The text in the first 5 lines of paragraph 8.2, page 141, got mangled.

I assume that the intention for variable parameters is, that the types of the
corresponding formal and actual parameters must be identical, like in the
3rd edition. For opaque types, however, it is not clear to me whether the change back to
the 2nd edition was intentional or accidental.

... (end of list)

I also wrote a letter about this to Prof. Wirth. His reaction contained
the following paragraph (copied verbatim):

  You are of course right in assuming that while editing parts of the 2nd
  edition slipped in. The types of VAR parameters must not only be compatible,
  but identical. Opaque type continue to be restricted to pointers. I am
  particularly dismayed that the Syntax (App. 2) is wrong; it should be
  replaced by that of the 3rd Edition.

So, don't believe the 4th edition!

--
Ceriel Jacobs, Dept. of Mathematics and Computer Science, Vrije Universiteit,
De Boelelaan 1081, 1081 HV Amsterdam, The Netherlands
Voice: +31 20 5485577 			Email:	ceriel@cs.vu.nl
-- 
Ceriel Jacobs, Dept. of Mathematics and Computer Science, Vrije Universiteit,
De Boelelaan 1081, 1081 HV Amsterdam, The Netherlands
Voice: +31 20 5485577 			Email:	ceriel@cs.vu.nl