Re: [femlisp-user] Equations/laplace/electromagnetic potential demo

[Nicolas Neuss]

[Alexander, sorry for the double-posting, but this should got to the
list, of course]

Alexander Shendi <address@hidden> writes:

> Dear List,
>
> I have installed femlisp from quicklisp. 
>
> I am using sbcl 1.3.10 on openbsd 6.1 (snapshot).
>
> The only configuration I did was unzipping triangle.zip in:
> ~/quicklisp/dists/quicklisp/software/femlisp-20170227-git/external/triangle
> and typing "make" in the this directory to build triangle
> and showme. 
>
> I tried to run the Equations/Lapalace/Electromagnetic Potential/ Demo.
>
> The demo seems to run up to a point and then stops with the error message:
> "argument A given to GESV! is singular to working machine precision".
>
> I have attached a more detailed log.
>
> I have looked at the meshes generated by triangle with the 
> "showme" program and the mesh looks OK to me.
>
> Do I need to install more external software to run this demo?
>
> Thanks in advance for your help,
>
> Alexander

Hi Alexander,

thank you for the bug report.

Sorry for this late answer, but this error did not occur for me at
first, so it took me a little time to reproduce it.

It looks as if a preliminary remedy is to set in your .sbclrc
initialization file:

  (setq *READ-DEFAULT-FLOAT-FORMAT* 'double-float)

(or alternatively you could execute this command before loading Femlisp).

[Of course, Femlisp should ideally also work in a "single-float
context", and there are some provisions inside femlisp.asd which should
make it work.  But somehow, this fails in your case, and I will have to
look more closely why the above provisions do not work.]

Nicolas

P.S.: For numerical work, double-float is usually considered to be the
      right precision, so making it the default read format is not
      considered bad.  Also, most other languages usually use doubles
      instead of floats.