Scilab 5.1 starts with "error 4 Undefined variable: exec " message. Bug?
nick.ko...@gmail.com
I install and run Scilab 5.1 on 32 bits Windows Vista. I does not
work, buttons are unavailable and error message has appeared:
!--error 4
Undefined variable exec
Moreover, closing console window does not stop SCilex.exe and I have
to kill it from the task manager.
Does anybody know how to make Scilab 5.1 working correctly?
Thanks in advance.
Nick
Roy
seems more stable than the scilab 5.x series
- Roy
Tim Wescott
(top posting fixed)
> On Mar 11, 2:18 pm, nick.kostro...@gmail.com wrote:
>> Hi Everybody!
>>
>> I install and run Scilab 5.1 on 32 bits Windows Vista. I does not work,
>> buttons are unavailable and error message has appeared:
>>
>> !--error 4
>>
>> Undefined variable exec
>>
>> Moreover, closing console window does not stop SCilex.exe and I have to
>> kill it from the task manager.
>>
>> Does anybody know how to make Scilab 5.1 working correctly? Thanks in
>> advance.
>>
> more stable than the scilab 5.x series
>
and reinstalling Scilab -- I have had trouble with 5.x, but nothing close
to this severe.
(And I sure wish the Scilab team _listens_ to all this chatter about 5.x
-- Scilab has always been a very good application, but it lacks a beta
program and it's "production" releases always have some bug or another.
And no, nightly builds do _not_ count as beta releases.)
(While I'm ranting, if there are any Scicoslab developers out there, this
is _your_ chance to one-up the Scilab team by establishing a beta program
of your own. I'll participate if you do, and try to send nice helpful
bug reports with more than "it no work, make work good!").
ycollet
> -- Scilab has always been a very good application, but it lacks a beta
> program and it's "production" releases always have some bug or another.
> And no, nightly builds do _not_ count as beta releases.)
You still can install side by side scilab-5.1 and scilab-master so as
to test the new modifications and report ASAP bug related to scilab-
master.
By doing so, you can be pretty sure that scilab will work on your
platform.
I know that sometimes it's hard (uninstalling scilab master,
reinstalling scilab master) but it worths the effort.
Yoou can also have a look at http://gitweb.scilab.org to follow what
kind of changes have been comited to see if it worth a reinstallation
of scilab-master.
And a message to Roy: what will you do when you will find a bug in
Scicoslab ? Stop using *lab* at all ?
If you find a bug in Scilab, report this bug on http://bugzilla.scilab.org.
Switching to scicoslab each time you find a bug is not really a
solution (I am not judging scicoslab, but the fact that you give up
each time you find a bug).
YC
Delebecque Francois
.
>
> And a message to Roy: what will you do when you will find a bug in
> Scicoslab ? Stop using *lab* at all ?
>
sup...@scicoslab.org
We try our best to find quickly a workaround and fix the problem for the
next release.
FD
Roy
> And a message to Roy: what will you do when you will find a bug in
> Scicoslab ? Stop using *lab* at all ?
> Switching to scicoslab each time you find a bug is not really a
> solution (I am not judging scicoslab, but the fact that you give up
> each time you find a bug).
>
> YC
Short answer: Yes, if the bug is onerous enough and I can't find
anyone to explain to me how to fix it or work around it, I will quit
using the tool or switch to one that works. I don't have a stake in
either fork, nor do I intend to ruffle any feathers - I just want a
tool that is usable.
I tried two of the 5.x scilab releases on windows XP. Both of them
only displayed a small portion of a scicos window (even for small
drawings) and I couldn't figure out why. They would also crash on me
sometimes for no obvious reason. I read around on this group and saw
that other people were also having issues with the 5.x series,
especially with graphic card drivers, so I assumed I didn't find
anything new. I didn't even bother reporting the specific issues here.
If you think it will help, I will file a bugzilla report on the scicos
rendering "bug", now that I know about that service.
I even tried the latest release in the 4.x series, but couldn't get
the C compiler to work with it even though it seemed to function
properly otherwise. The scicoslab version 4.3 worked, and the people
here in this group were very helpful in getting my questions answered
& bugs resolved. So that's the version I'm using. I hope that the two
versions merge into a single one again in the future, so we won't have
to have discussions like this one. :-)
- Roy
Tim Wescott
I know I'm being a pain in the behind, but this is important. I like
Scilab and I want it to succeed, or I'd just be polite and let you all
slide off of a cliff.
Your release process is broken, the way that it's broken is exactly what
having a properly administered beta program fixes and your comments
indicate that you simply don't understand the value of or the reason for
a beta program.
The reason I say that your release process is broken is because I've been
with you since just before you released 4.0, and every major
release has been accompanied by bugs that just shouldn't be there.
They're mostly bugs that couldn't easily be caught by automatic testing
(although some of them are), but they're all bugs that could be caught
with a good beta program, assuming that you've got a good community of
beta testers out there.
In fact, most of what you allow to escape as "production releases" are of
beta quality. Were I releasing code of that quality for public
consumption I would be profoundly embarrassed. In fact, I cannot
recommend Scilab to clients and acquaintances with a clear conscience,
unless I heavily qualify my recommendation by pointing to a version that
I know works. That should not be my job -- that should be your job.
The consequence of this behavior on your part is that your quality
problems drive away a large portion of your potential user base. Most
potential users don't want to report bugs, or participate in newsgroups.
Most potential users _certainly_ don't want to struggle with broken
software that require advanced computer skills to work around all of the
problems inherent in its use.
By releasing software of the quality that you are releasing, you are
shooting yourselves in the feet. Whether you are writing free software
or purchased software, the success and longevity of your product is
determined by the size and satisfaction level of your user base.
Releasing a shoddy product (and I hate to say it, but 5.x has proven to
be pretty shoddy) without calling it 'beta' tells the world "we don't
care about quality". And most users don't have the time, the
inclination, or the skill to work around the problems in known-buggy
software.
Since you don't seem to know what a beta program is, let me outline it
here:
When you get ready for a new version (5.0.4, say), you don't release it
as 'regular' software. Instead you release a beta version (5.0.4.b0,
say). The beta version is your way of saying "I think this is good
enough, but I know that I'm human and fallible, and I'd like you to take
a look". People like me, who love what you're doing and want to help
download it, knowing that (a) they're going to find bugs, (b) they're
going to report the bugs they find, (c) they're going to continue to be
safe in recommending the most recent _real_ release to 'civilians', and
(d) you're going to address the bugs that they find.
Once you get all necessary feedback on the bugs, you fix them, _without_
adding any new features. If you have to do extensive fixes you release a
second beta, and if you're having a really bad time with that release you
do it as many times as necessary to get things ready to go. Then, when
you're _sure_ that you have a reasonable product worked out, you release
the 'real' version (hopefully to acclaim).
A beta program isn't going to keep bugs from happening -- we're human,
after all. What a beta program _will_ do is keep new users from
downloading 5.0.3 and running into a stupid bug that would have been
caught by a beta program (and they're there). Because when a new user
_does_ do something like print a number that's really close to 1.0 and
see 2.0, or use the program for half a day and have it crash or bog down,
or have it play any of the other tricks that Scilab 5.0.x plays on my
Windows box, THEY WON'T STICK AROUND. Instead of becoming and _old_
Scilab user, they'll stick to pencil and paper, or they'll buy Matlab, or
otherwise leave the Scilab community before they even start.
I just don't want to see that happening, because I want to see Scilab be
successful. And if Scilab is going to be successful, then it needs to
follow the same path that OpenOffice, Debian, and all the other
successful open (and closed) software efforts follow, and use some
effective quality management, of which beta testing is not the least
element.
Pierre
Hi Nick,
Your problem is probably due to the SCI/etc/modules.xml file which has
been badly generated during the installation (with innosetup).
In the past, we had problems with Innosetup and antivirus, if
possible, can you tell us your antivirus name ?
A workaround : Replace the content of the file "C:\Program Files
\scilab-5.1\etc\modules.xml" with the following lines :
<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<!DOCTYPE modules SYSTEM "../modules/core/xml/modules.dtd">
<modules>
<module name="core" activate="yes"/>
<module name="helptools" activate="yes"/>
<module name="demo_tools" activate="yes"/>
<module name="fileio" activate="yes"/>
<module name="io" activate="yes"/>
<module name="output_stream" activate="yes"/>
<module name="shell" activate="yes"/>
<module name="action_binding" activate="yes"/>
<module name="console" activate="yes"/>
<module name="history_manager" activate="yes"/>
<module name="completion" activate="yes"/>
<module name="data_structures" activate="yes"/>
<module name="boolean" activate="yes"/>
<module name="elementary_functions" activate="yes"/>
<module name="functions" activate="yes"/>
<module name="localization" activate="yes"/>
<module name="time" activate="yes"/>
<module name="dynamic_link" activate="yes"/>
<module name="integer" activate="yes"/>
<module name="jvm" activate="yes"/>
<module name="gui" activate="yes"/>
<module name="double" activate="yes"/>
<module name="overloading" activate="yes"/>
<module name="windows_tools" activate="yes"/>
<module name="special_functions" activate="yes"/>
<module name="string" activate="yes"/>
<module name="genetic_algorithms" activate="yes"/>
<module name="simulated_annealing" activate="yes"/>
<module name="parameters" activate="yes"/>
<!-- comments -->
<!-- comments -->
<module name="graphics" activate="yes"/>
<module name="renderer" activate="yes"/>
<module name="graphic_export" activate="yes"/>
<module name="linear_algebra" activate="yes"/>
<module name="statistics" activate="yes"/>
<module name="signal_processing" activate="yes"/>
<module name="cacsd" activate="yes"/>
<module name="optimization" activate="yes"/>
<module name="polynomials" activate="yes"/>
<module name="symbolic" activate="yes"/>
<module name="interpolation" activate="yes"/>
<module name="sparse" activate="yes"/>
<module name="differential_equations" activate="yes"/>
<!-- comments -->
<module name="scicos" activate="yes"/>
<module name="scicos_blocks" activate="yes"/>
<module name="metanet" activate="yes"/>
<module name="arnoldi" activate="yes"/>
<module name="tclsci" activate="yes"/>
<module name="scipad" activate="yes"/>
<module name="sound" activate="yes"/>
<module name="randlib" activate="yes"/>
<module name="mexlib" activate="yes"/>
<module name="texmacs" activate="yes"/>
<module name="m2sci" activate="yes"/>
<module name="matio" activate="yes"/>
<module name="maple2scilab" activate="yes"/>
<module name="scilab2fortran" activate="yes"/>
<module name="javasci" activate="yes"/>
<module name="intersci" activate="yes"/>
<module name="spreadsheet" activate="yes"/>
<module name="umfpack" activate="yes"/>
<module name="compatibility_functions" activate="yes"/>
</modules>
Pierre
Hazem
I have to agree with Tim.
I love Scilab and consider it the best alternative to MATLAB, but it
gets me frustrated alot, and I cannot recommend it to many people,
including my boss and colleagues at work. It frustrates me because of
stupid bugs, because the Scilab and Scicos development have been
split, and because of the support system around it which is nowhere
near as simple, sustainable, and convenient as the one used by the
Mathworks (for example FileExchange). All these are factors in a
successful product. And while I admire the dedication to new features
and technical innovation, the overall quality is determined my many
other factors, as Tim has explained.
The Scilab development process needs review and improvement, and while
this may take time, the alternative is far worse.
I just want a Scilab that is usable, I don't want to have to struggle
with it, I want to do science with it.
Best wishes to all involved,
Hazem
On Mar 12, 11:29 pm, Tim Wescott <t...@seemywebsite.com> wrote:
> On Thu, 12 Mar 2009 00:25:46 -0700, ycollet wrote:
> >> (And I sure wish the Scilab team _listens_ to all this chatter about
> >> 5.x -- Scilab has always been a very good application, but it lacks a
> >> beta program and it's "production" releases always have some bug or
> >> another. And no, nightly builds do _not_ count as beta releases.)
>
> > You still can install side by side scilab-5.1 and scilab-master so as to
> > test the new modifications and report ASAP bug related to scilab-
> > master.
> > By doing so, you can be pretty sure that scilab will work on your
> > platform.
> > I know that sometimes it's hard (uninstalling scilab master,
> > reinstalling scilab master) but it worths the effort. Yoou can also have
> > a look athttp://gitweb.scilab.orgto follow what kind of changes have
> > been comited to see if it worth a reinstallation of scilab-master.
>
> > And a message to Roy: what will you do when you will find a bug in
> > Scicoslab ? Stop using *lab* at all ? If you find a bug in Scilab,
> > report this bug onhttp://bugzilla.scilab.org. Switching to scicoslab
> --http://www.wescottdesign.com- Hide quoted text -
>
> - Show quoted text -
nick.ko...@gmail.com
Thanks a lot for the workaround! It cured Scilab in no time!
I use Kaspersky Anti-Virus 2009 Version 8.0.0.506(a) by Kaspersky Lab.
(http://www.kaspersky.com)
It seems that Scilab 5.1 differs from Scilab 4.1.2 so much that my
macros made for 4.1.2 do not work correctly. The most annoying things
are event handler, graphic window and predef function as well as
need to load code as 'toolbox'.
Thanks again.
Nick
pradeep
I was facing the similar issue as Nick with Error 4. finally disabled
Kapersky and reinstalled Scilab to come around this problem
Thanks,
pradeep
sugan....@gmail.com
i already ran program and got output but now i run the same program but it will show undefined variable error.........what is the problem? how can i rectify it?????????