Mathematica 5.0

["Georg Ch. Mellau" <Georg.Mellau@XXXXXXX.de>]

Liebe Mathematica Freunde,


Ich habe einen Email drangehängt die ich WR geschickt habe mit meine
Erfahrungen mit Mathematica 5.0.
Sicher ist die Anbingung an .NET C# ein Meilenstein in der Entwicklung von
Mathematica und - vorausgestezt das es so funktioniert wie es verkauft
wurde-
, damit ist auch eine update auf 5.0 empfehlenswert. Leider ist was mich
betrifft nicht der Fall.
Da ich insgesamt mehr als 30 MB code und über 100 MB daten is Kernel lade
könnte sein das im "Normallfall" von mir beschriebene Probleme
nicht auftreten.

MfG

Georg



Dear Mathematica developer team,

I am using Mathematica very extensively, I have a big set of mathematica
functions (10.000 pages of written code, half of them manipulating the
FrontEnd for input/output) which I use for the analysis of high resolution
spectra.
>From the version 3.0 I am using daily Mathematica for several hours.

I am telling this to you with the hope that you may take the following very
seriously:

MATHEMATICA version 5.0 IS THE MOST UNSTABLE VERSIONS OF ALL MATHEMATICA
versions !!!!!!!!!!

I used in the last years the version 4.1.2 which was very stable, both of
them the Kernel and FrontEnd.
I passed the version 4.2, there was a big error in the FrontEnd (in version
5.0 fixed) so that it was for me useless.

I started for some weeks to use my code with the version 5.0. It crashes at
about every 30-40 minutes with different error messages.
What I found very bizarr is the fact that not as usually the FrontEnd
crashes, now the Kernel
is the one that crashes. It happens very often and there is a total crash:
one gets only a OS message that the system was doing
something wrong.

Much worse, in all versions happend time to time that the FrontEnd lost the
link to the Kernel. In the version 5.0 ist this very often. It takes me up
to 10 minutes to load all the code and Data and start the analysis. After 15
minutes I have once 2 Kernels running with all my data in the "lost kernel".
It is impossible.

The system writes error messages as

Mathematica has detected a possible internal error. If possible, report the
error to support@XXXXXXX.com, quoting "Assertion '!"Expression written to
link when FrontEnd was not expecting any expressions"' failed at
KernelBackground.c:435", and describe in as much detail as possible what you
were doing when the error occurred

with a relative high frequency.

This Mathematematica "gold" release is not what I am ususually used to get
from Wolfram Research.




With best regards

 Dr.Georg Mellau




P.S.

I found one of the possible reasons (But I am not sure if it is not only
side effect): "Compiled" functions seems not to be stable in Version 5.0.
After I changed the Compile functions with normal ones the crash frequency
decreased. The function- quite simple-  :


Symathreplacearraypart = Compile[
      {
        {x, _Real, 1},
        {y, _Real, 1},
        {i, _Integer}
        },

        If[(i + Length[y] <= Length[x]),
        (

          Join[
            Take[x, {1, i - 1}],
            y,
            Take[x, {i + Length[y], -1}]]

          ),
        (

          If[(i + Length[y] == Length[x] + 1),
            (
              Join[
                Take[x, {1, i - 1}],
                y]
              ),
            (
              x
              )]

          )]

      ];