Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!wuarchive!uunet!munnari.oz.au!metro!seagoon.newcastle.edu.au!cc.newcastle.edu.au!eepjm
From: eepjm@cc.newcastle.edu.au
Newsgroups: comp.lang.modula2
Subject: Re: Compiler reviews
Message-ID: <1991May7.144056.10082@cc.newcastle.edu.au>
Date: 7 May 91 04:40:56 GMT
References: <2934.2821B3F2@puddle.fidonet.org>
Organization: University of Newcastle, AUSTRALIA
Lines: 44

In article <2934.2821B3F2@puddle.fidonet.org>,
   Cees.van.Kessel@p4465.f14.n512.z2.fidonet.org (Cees van Kessel) writes:
> During a recent cleanup of my paper stack I found several article copies on 
> compiler reviews.  Another trip to the Uni library completed the list and here 
> its:

 [and then follows a list of references to review articles for various
  Modula-2 compilers]

This is undoubtedly useful stuff, but I'm feeling lazy, so it's worth
asking the question: has anyone read these reviews (and/or other sources
of information) to the point of being to reach some general conclusions
- for example, is it possible to make statements of the form "almost
everyone agrees that Brand X is far superior to its competitors"?  Or does
it simply turn out that different reviewers have different opinions?

For the record, here is what I know about the subject:
1. I've used several different Modula-2 compilers for the VAX, and they were
   all of unacceptable quality.  (Sorry, I can't remember the names of the
   suppliers, though I could ask our computing centre if anyone really wants
   to know).
2. I've briefly used a Logitech compiler for the IBM PC before giving up
   in disgust - though I've since been told that that was an old version and
   that some of the problems with it have been fixed.
3. I've used FTL for the IBM PC very extensively over the last few years,
   and am generally satisfied although I can see some room for improvement.
   I have a rather long list of bugs for this one (but this is not a
   criticism, I believe that I would accumulate an equally long list for
   *any* compiler that I had used that heavily).  It has only two really
   severe faults: (a) all the built-in floating point operations are badly
   designed (slow, and not re-entrant so that programs crash when you try
   to use multitasking); (b) it can only produce Real Address Mode code,
   i.e. it ignores the segmentation hardware of the 80286 and up and
   therefore you are limited to program size < 640K.  There are a few other
   aspects of the compiler design with which I disagree, but I didn't plan
   to make this posting a full review (Could produce a full review, I
   suppose, if enough people want it).
4. I've seen lots of commentary, in this group and elsewhere, on various
   other compilers but very little on *comparative* judgements - most
   people, like me, probably know only one compiler really well and
   therefore can't answer the interesting question of "which is better than
   which".

Peter Moylan                        eepjm@cc.newcastle.edu.au