NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/comp/lang/pascal/6412

Internet Message Format | 1992-11-09 | 5.1 KB

Path: sparky!uunet!think.com!yale.edu!ira.uka.de!uka!uka!news
From: [e-mail] (|S| Norbert Juffa)
Newsgroups: comp.lang.pascal
Subject: Re: What we need is a TPU to PAS convertor (Re: tpu files)
Date: 9 Nov 1992 10:09:00 GMT
Organization: University of Karlsruhe (FRG) - Informatik Rechnerabt.
Lines: 75
Distribution: world
Message-ID: <[e-mail]>
References: <[e-mail]> <[e-mail]> <[e-mail]>
NNTP-Posting-Host: irav1.ira.uka.de
X-News-Reader: VMS NEWS 1.23
In-Reply-To: [e-mail]'s message of Thu, 5 Nov 1992 22:56:31 GMT

In <[e-mail]> [e-mail] writes:

> In article <[e-mail]> [e-mail] (Jim Nicholson) writes:
> >Anyway, the original post was in error; what's needed is not a TPU
> >disassembler.  What we need is to force Borland to either stabilize the TPU
> >definition (make it release independent), publish the interface with each
> >release (making TPU upgrader programs easier to write), or provide their own
> >upgrade utility.
>
> None of those would help.  The interface definition has been published for
> TP 6 for at least a year now (see TWU1.ZIP on garbo for the most up to date
> version).   I believe the same author published the description for 5.5
> before that.  Certainly I've distributed source code that uses the .TPU
> format for all versions of TP from 4.0 onwards (see INTRFCxx.ZIP on
> garbo).  However, there are technical reasons why an upgrade utility would
> probably be unreliable - see my earlier post on this.
>
> A reliable solution that Borland could implement would be a compiler option
> to produce .OBJ output instead of .TPU output. If you want Borland to do
> this, start buying compilers that do it instead of Borland compilers.  Stony
> Brook's, for instance.  Borland is market driven.  If they sell just as many
> compilers without that option, why put it in?
>
> Duncan Murdoch
> [e-mail]

I went to Borland's German product launch for Turbo Pascal/Borland Pascal 7.0
held in Frankfurt/Main on November 4th. Zack Urlocker and Anders Hejlsberg
(inventor of Turbo Pascal) did the presentation. I even had the chance to
talk to Mr. Hejlsberg for about 30 minutes before the presentation. The
presentation was attended by about 1200 people. There was a question&answer
session at the end in which Mr. Hejlsberg was asked why they kept on changing
the TPU format from one TP version to the other and why they wouldn't provide
a converter tool for TPU files. His answer was that the TPU format is designed
for maximum speed, as it is basically a copy of the compiler's internal symbol
table. He also said that he thinks it is impossible to do a TPU converter that
does a 100% conversion, as some of the information present in the original PAS
file has already been lost by the compilation process. He instead encouraged
software writers to always provide the source code where it would be helpful.
Borland has taken this path as well, as they include the library and
Turbo-Vision source code with TP/BP 7.0, whereas before one had to pay an
additional US$200 to get that.

Mr. Hejlsberg also asserted that development of the Pascal compiler was purely
customer driven. Borland constantly asks software developers working with TP
what features they want to see in TP. They keep a TOP-50 (or 20?) list of these
items. Keep in mind that about 50% of all TP are professionals, which do
projects of up to a few 100 KLOC (kilo lines of code) with TP. Their needs may
be very much different from the needs of a student learning Pascal using TP,
for example. He gave the same answer when asked if their would be an OS/2
version of TP. If there was enough demand voiced by customers, they would do
it.

In my private communication with Mr. Hejlsberg, I asked him why certain bugs in
TP (that have been reported here before) haven't been fixed yet. His reply was
twofold: 1) Some of the bugs have been worked around by customers (I assume
he ment *important* customers here) and a bug fix would break their code. 2) In
trying to fix some bugs they discovered that it was very difficult to find a
'clean' fix and that they were likely to introduce a new bug by removing the
old one (this actually happened on some occasions). So they decided it would be
better to stay with the old and well known bug. BTW, Borland will shortly
introduce a CD-ROM which has a complete bug-list for TP/BP including patches
and workarounds on it. No date was given, though, at which this will be
available.

I also asked Mr. Hejlsberg why he didn't use some of my fast library routines
for TP (available as TPL60N17.ZIP from [e-mail]) for the TP/BP 7.0
library. He told me something about untested code, QA,..To me it sounded
rather like that "not invented here" phenomenon. Anders Hejlsberg does the
RTL all by himself, with another programmer being in charge of TV.


Norbert Juffa
-------------------------------------------------------------------------------
Norbert Juffa         email: [e-mail]       Live and let live!