[bugfix] correct DUNE_VERSION_NEWER alias

See merge request !353

The semantics of the old macro DUNE_VERSION_NEWER is DUNE_VERSION_GTE,
not the opposite. While we're at it, also fix the deprecation message to
suggest DUNE_VERSION_GTE instead of DUNE_VERSION_GE.

[bugfix] Fix compile error with NDEBUG

See merge request !352

It seems that the `#if ...` was originally introduced to avoid
a compiler warning in case later `assert()` are removed due to
NDEBUG. When changing to `TestSuite` in e7b592a2 this
was not adjusted.

**Surprisingly no one compiled our tests with NDEBUG since June 2017 otherwise this would have triggered a compile error!**

Replace DUNE_VERSION_NEWER with DUNE_VERSION_GTE.

See merge request !329

[bugfix] silence signed/unsigned comparison warning

See merge request !349

cmake: fix regex for hidden files

See merge request !330

[Vc] Deal with the missing swap() on Vc proxies.

Closes #59

See merge request !339

This is needed to pull in the fix for #88.

[Vc] Work around missing std::for_each_n() in g++ 7.2.

Closes #88

See merge request !341

`Vc/common/algorithm.hh` uses `std::for_each_n()` when it detects C++17, but
g++ 7.2 does not seem to provide it (yet).  Work around this by disabling the
macro that signals C++17-support in Vc.

Closes: #88.

This is a problem with Vc 1.3.0.  Vc 1.3.2 seems to have implemented
`swap()`.

We work around it using copying and a temporary to implement `swap()` on our
own wrapping proxies.  Any member type of a Vc SIMD type should be a
fundamental type, so we don't bother using moves.

Closes: #59.

Deal with Vc access proxies

Closes #59

See merge request !338

This is needed by the `luDecomposition()` in densematrix.hh

This is needed because newer versions of Vc introduce their own proxies as the
result of expression like `v[0]`.  These proxies cannot be constructed outside
of Vc, not even by copying or moving, so we cannot pass them out of our
`lane()` function.  We work around this by introducing our own proxy class,
that we can copy.

That proxy class was originally introduced for !193 (Proper SIMD Interface)
and has been backported for the 2.6 Dune release.

Closes: #59

bump version to 2.7-git [ci skip]

See merge request core/dune-common!337

parallel index sets: correct typo; return value of AllSet::contains() [ci skip]

See merge request core/dune-common!335

Previously hidden directories were also matched and stripped. This
resulted in the build failing if the build dir contained one hidden
directory (e.g. '/tmp/.abc/dune/...')

Fix module name for dunecontrol update

Closes #78

See merge request core/dune-common!332

Fix status message

See merge request core/dune-common!331

When dunecontrol update fails the module name no longer has
underscores instead of hyphens.

This fixes #78

We are not 'running' the doxygen documentation,
we are building it.

and added variants DUNE_VERSION_GT, DUNE_VERSION_LT, and DUNE_VERSION_LTE.

Replace throw by terminate in destructor

See merge request core/dune-common!327

re-enable transformTuple test fails with debian 8 gcc

See merge request core/dune-common!273

added a range function for range based for loops over consecutive integers

See merge request core/dune-common!325

[std] add fallback implementation for std::conjunction / std::disjunction

See merge request core/dune-common!322

Disbable a deprecation warning in a test

See merge request core/dune-common!328

Since the destructor is no longer noexcept(false) it
will never throw an exception to its caller but always
call std::terminate().

In C++11 destructors are implcity `noexcept` such
that a `throw` leads to a call of `std::terminate()`
anyway. Hence this commit does not change the behaviour
but only avoids a warning.

BTW: Destructors are `noexcept` for good reason, because
it is in general very hard to recover from a failing destructor.
This also applies here.