[Dune] Compiler.

[Oliver Sander]

Hi guys!

> Sounds good. But also sounds like much work. I think we should stay focus on 
> the real problem, wirting numerical code.

I think we should have automated build testings _because_ we can then
focus on the 'real' problems.  Whatever those are.  Just trying to build
Dune by hand once in a while on some odd machine takes much more effort
in the long run and never really yields sound results.  We will never
truly be able to claim that 'Dune runs on A, B, C with compilers D, E'.

I must say, however, that I don't think that it's of paramount importance
to have Dune compile with every compiler out there.  If a feature is
standard-conforming, and a compiler does not like it, I propose we do
not hack our code but suggest people use a better compiler.
Well, that's the purist view.  Let the flame war begin!

It is not known to me whether the above said applys to Roberts codim-
example.  Another little detail comes to my mind, though.  I think
that, because the struct codim is a, well, struct, it should be called
Codim, because that's what people expect from our naming conventions.

Have a good day,
Oliver


> The problems, which caused that 
> icc7.0 didn't compile the code anymore, was the template parameter in the 
> entity class, code like this:
>
> template <int codim,....>
> class Entity
> {
>  template <int cc>
>  struct codim
>  {
>
>  };
>
>  typedef codim<codim>::.....
>
> };
>
> The problem is, that icc inherits typedefs, which means even if the codim 
> struct is only a typedef in a base class, this conflicts but also is not as 
> clear as naming the template parameter of the dune grid interfaces different 
> then internal typedefs, i.e. no int template parameters named codim anymore. 
> I just cleaned the interface classes in dune/common/entity.hh, alu3dgrid.hh 
> and onedgridentity.hh. I think in all other grid implementation this problem 
> also occurs and should be fixed in the next time.
>
> I think for the moment we should just test from time to time that the code 
> compiles. Having a automatic code compile testing system is nice, but we 
> should focus our strength on the real problems.
>
> Cheers
>
> R
>
> -- 
>
>  Robert Klöfkorn           <robertk at mathematik.uni-freiburg.de>
>
>  Mathematisches Institut              Tel: +49 (0) 761 203 5631
>  Abt. für Angewandte Mathematik       Fax: +49 (0) 761 203 5632
>  Universität Freiburg
>  Hermann-Herder-Str. 10
>  79104 Freiburg
>
>  http://www.mathematik.uni-freiburg.de/IAM/homepages/robertk
>
> _______________________________________________
> Dune mailing list
> Dune at hal.iwr.uni-heidelberg.de
> http://hal.iwr.uni-heidelberg.de/cgi-bin/mailman/listinfo/dune
>