[[Imported from SVN: r6886]]

[[Imported from SVN: r6801]]

With the exception of mpicollectivecommunication.hh
and propertymap.hh, which I didn't understand.

[[Imported from SVN: r6784]]

[[Imported from SVN: r6683]]

[[Imported from SVN: r6682]]

This patch removes the constructor from the overload set.
Before the constructor was there but failed to compile
due to the static assertion.

If you have methods f(K) and f(FieldVector<K,n>) with
n>1 the overload was ambigous when calling f() with
a FieldVector<K,1> because there is a conversion
from this to FieldVector<K,n> - even if it does not
compile due to the failing static assertion.

Now this is avoided using SFINAE. This fixes fs#1024. 

[[Imported from SVN: r6680]]

[[Imported from SVN: r6559]]

[[Imported from SVN: r6494]]

to FieldVector::size() as a method.

With this patch applied FieldVector::size remains an enum, hence all your
code compiles as before.  However you get a compiler warning explaining
the issue and suggesting to configure with --enable-fieldvector-size-is-method.
If you do so, then FieldVector::size becomes a method, and you may have
to go out and adjust your code.  The configure flag is transitional and
will be removed after dune 2.2 is out.

Remember that there is no smoother way to do the transition, as you can't
have FieldVector::size to be both an enum value and a method at the same
time.

This fixes FlySpray issue 819.


[[Imported from SVN: r6486]]

[[Imported from SVN: r6386]]

[[Imported from SVN: r6385]]

How difficult is this. I just don't get it. 



[[Imported from SVN: r6384]]

[[Imported from SVN: r6377]]

[[Imported from SVN: r6375]]

Merged r6371 from the trunk, that ensures the correct size of the FieldVector given to the copy constructor.


[[Imported from SVN: r6372]]

different size!

[[Imported from SVN: r6371]]

[[Imported from SVN: r6366]]

FieldVector< complex<T> > = FieldVector<T>

[[Imported from SVN: r6324]]

[[Imported from SVN: r6272]]

[[Imported from SVN: r6179]]

  valued data

[[Imported from SVN: r6177]]

[[Imported from SVN: r6175]]

DynamicVector. Inheritance doesn't work for template specialization
lookup... not that this specialization isn't necessary, if you are
using FieldVector<T,n> via the interface class DenseVector< FieldVector<T,n> >

[[Imported from SVN: r6174]]

  few cases, where CRTP really makes sense!)
* type infroamtion is provided via DenseMatVecTraits

[[Imported from SVN: r6171]]

- FielfVector uses int, not unsigned int

[[Imported from SVN: r6168]]

that std::array defines these operators aswell and c++ prefers to
complain about an inaccesible base class, rather than using the cast
operator :-(

[[Imported from SVN: r6167]]

duplicated code, we only have to add the scalar operators.

[[Imported from SVN: r6165]]

[[Imported from SVN: r6157]]

[[Imported from SVN: r6135]]

[[Imported from SVN: r6129]]

vectors using different storage containers

[[Imported from SVN: r6128]]

[[Imported from SVN: r6105]]

[[Imported from SVN: r5926]]

[[Imported from SVN: r5923]]

[[Imported from SVN: r5922]]

merge to release

[[Imported from SVN: r5877]]

* add a specialization of sqrt to fvmeta, this should fix the
  ambiguous sqrt(int)

merge to release

[[Imported from SVN: r5874]]

[[Imported from SVN: r5860]]

* hide specializations from doxygen
* use ROWS and SOLS instead of n and m for the FieldMatrix
* use conjugateComplex instead of fm_ck
* document istl_assign_to_fmatrix

[[Imported from SVN: r5659]]

  type
* make *norm* in fvector and fmatrix use this real type

[[Imported from SVN: r5656]]