Programmer 7500/MAX_PROGRAMMERS.iso/INFO/TURBOPAS/TURBO30.ZIP/TURBO30.MSG

Text File | 1985-12-28 | 5.9 KB | 145 lines

The following are messages on the SMUG RBBS about Turbo Pascal 3.0.
P. L. Olympia, Sysop, 301-963-5249   03/31/85

Msg #  1439 Dated 03-22-85 05:38:59
From: PHILIP BURNS
  To: All
  Re: TURBO 3.0

I received my copy of Turbo 3.0 today.
I have only worked with it for a few hours,
but my initial impressions are extremely positive:
 (1)  All the bugs I encountered in release 2.0
      have been fixed.
 (2)  The compiler is AT LEAST twice as fast as
      before.  For example, a 10000+ line program
      I have which used to take 9+ minutes to
      compile now compiles in less than four minutes.
 (3)  The generated code is substantially smaller --
      as much as 35% for some of my programs --
      and runs considerably faster.
 (4)  The editor appears to run much faster --
      nearly instantaneous screen updates, etc.
 (5)  The PC/DOS version (which I am using) includes
      standard procedures for directory access and
      manipulation.  In addition, file names may now
      contain drive/subdirectory information:
      a file name like 'C:\mydirec\myfile.dat' is now legal.  (Cont...)


Msg #  1440 Dated 03-22-85 05:40:05
From: PHILIP BURNS
  To: All
  Re: TURBO 3.0 (CONT.)

 (6)  The installation program allows you to define
      the drive/directory for the message file.
      No more having to have multiple copies strewn
      all over your disks!
 (7)  The are a number of system-interface
      procedures for getting command line arguments,
      defining a path for TURBO overlays, etc.
 (8)  There are a number of advanced graphics
      functions -- for generating circles,
      fill patterns, etc. -- as well a very nice
      implementation of Turtle graphics.
 (9)  Full I/O redirection on standard files is
      provided.  The run-time package also uses
      standard MS DOS file handles.
      NOTE:  THIS MEANS THAT TURBO 3.0 WILL ONLY RUN
             UNDER MSDOS/PCDOS RELEASE 2.0 OR LATER!
(10)  You can change buffer sizes for files.
(11)  You can open a file for append.
(...Cont.)


Msg #  1441 Dated 03-22-85 05:41:37
From: PHILIP BURNS
  To: All
  Re: TURBO 3.0 (CONT.)

(12)  Text I/O is considerably faster -- especially
      with larger buffer sizes -- and a number of new
      procedures are available for manipulati~s+$H  text files.
(13)  External assembler procedures appear to work
      correctly now.
(14)  The BCD version provides nice editing facilities
      for numbers, which should ease writing
      business-oriented programs.

On the negative side, there are a few things still missing:
 (1)  Still no integer type > 16 bits.
 (2)  Still no out-of-block GOTOs.  (This hinders writing
      parsers using recursive descent.)
 (3)  The 8087 version still only supports the
      64-bit real -- no 32 bit reals, no 80 bit reals,
      no long integers, etc.
 (4)  There are quite a few errors in the published
      documentation. Hence, you MUST READ CAREFULLY the
      README file enclosed on the distribution disk.


Msg #  1442 Dated 03-22-85 05:44:25
From: PHILIP BURNS
  To: All
  Re: TURBO 3.0

One word of caution about Turbo 3.0 -- in fixing the EOLN and
related functions, Borland changed their function a little when
applied to Kbd.  If you apply EOLN to KBD, you'll find things
get a little behind -- there's a one-character delay.  This means
that if you wrote your own procs to process/edit input lines,
and you used EOLN(Kbd), then THEY WON'T WORK ANYMORE.  The easiest
fix is to check for a carriage return character and not bother
with the EOLN -- not exactly standard Pascal, but what the heck,
who cares anyway.


Msg #  1455 Dated 03-24-85 22:31:09
From: JEFF BASS
  To: PHILIP BURNS
  Re: (R)TURBO 3.0

I too have been using verion 3.0 for a few days and am very impressed.
I wish they had come out with it sooner as I have written various things
like the cmdlin routine in assembler....  One other difference I noted
from version 2.0:  The 8087 version now checks ALL arithmetic for over-
flows, etc before handing it over to the 8087 -- this means that NAN's
do not propogate through expressions as per the IEEE standard and there
is not a way to say that you don't want overflows etc to use the IEEE
standard response (as there is in MS Pascal).  This means that you can't
use the NAN as a statistical missing value in TURBO (as I like to do)
but this is a relatively minor problem...  In version 2.0, NAN's propo-
gated as per the IEEE standard for all but divide by 0 and sqrt(neg
num).  Now you get kicked out with an error.  Also, Phil, with respect
to the 8087, there is no easy way to get at the 80 bit format -- it is
intended to be internal only (thus providing the guard bits required
by the IEEE standard).  You can copy the whole 8087 stack, but that
seems like overkill just to get an 80 bit representation into memory.
Comments?


Msg #  1457 Dated 03-25-85 02:14:35
From: PHILIP BURNS
  To: JEFF BASS
  Re: (R)TURBO 3.0

I wuld have preferred the 8087 implementation in Turbo to allow as
direct an access to the entire instruction repertory as possible.
Version 3.0 isolates us further than ever, as you point out.
The 80 bit reals are useful for accumulating inner products, among
other things, and direct access to them is useful for that reason,
among others.I currently have to access them via INLINE code
in Turbo -- I had hoped that 3.0 would eliminate that need.  Perhaps
in version 4.0 ....

Yes, I agree, NANs are quite useful for missing values, among other
things.  I've used CDmainframes for many years, and one feature
of CDC architecture which has always proved invaluablhas been
the infinite/indefinite NANs provided by the floating-point
instructions.  They really grow on you, to the point where programming
machines without them seems like a really redious chore.  That's one
reason that I like the 8087.

* END OF MESSAGES *