MathGroup Archive 2012

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Sending an interrupt to the frontend?

On 7/19/12 at 3:51 AM, leigh at (leigh pascoe) wrote:

>I would also agree with that. Mathematica seems to hang for no
>apparent reason

There is always a reason. But determining the true reason for a
crash is often very time consuming and far from easy.

>and the only solution is to kill the process and restart, often losing
>work in progress.

Losing work in process as a result of killing a program is the
usual consequence of killing a program.

>I posted a query about this several times on the forum, but got no
>replies apart from someone who confirmed that they experienced the same
>problem. It is difficult to contact Wolfram technical help without a
>specific sequence that crashes the program.

Without a specific sequence that leads to a crash, there isn't
much anyone can do. Even if I seem to get a similar crash
running code that seems to match your description, there can be
no assurance whatever the root cause is the same. And if I
cannot duplicate the crash, clearly there is nothing I can do to
resolve the fundamental issue.

Note, I personally don't work for Wolfram and have no access to
the source code for Mathematica. But this really changes
nothing. For anything that is not working correctly be it
Mathematica, another program, you car whatever, the first
essential step is to be able to duplicate the undesired behavior.

Also, even if Mathematica were totally bug free (as in
everything works precisely as documented, not necessarily as you
might expect), Mathematic would still not be crash free.
Mathematica give you a tool set that allows you to do
arbitrarily complex problems. Any tool set that gives you this
much power will give you the power to write code that crashes
your system whether or not that is your intent. Additionally, it
is simply not feasible for Mathematica to check code of
arbitrary complexity for problems that lead to crashes.
Algorithms for checking sufficiently complex code probably don't
exist or don't execute in reasonable time.

I am not suggesting Wolfram should not do everything they can to
make Mathematic robust and fail in graceful ways rather than
simply crash. I am saying there are very real limits to what
Wolfram can achieve in this regard.

I've been using Mathematica on an almost daily basis for the
past 10-20 years. And yes, I too manage to crash Mathematica now
and then (for some reason far more often under Windows than
under Mac OS X). And yes, I've lost work and am often quite
annoyed when this happens.

But, invariably, the root cause turns out not to be a bug in
Mathematica. Rather it is something I did that doesn't make
sense. Often this means, syntax errors or typos on my part that
get interpreted as valid Mathematica code but far from what I
intended. Or setting up a problem that needs far more memory
than I expected or have installed in my system. Or a myriad of
other things along these lines.

  • Prev by Date: can one make local symbol in a pure function?
  • Next by Date: Re: What is best trick to eliminate flicker in simulation animation?
  • Previous by thread: Re: Sending an interrupt to the frontend?
  • Next by thread: Re: Sending an interrupt to the frontend?