Merge branch 'get-rid-of-uggrid-heapsize' into 'master'

ref:core/dune-grid \[This is WIP because it needs some minor fixes in
dune-uggrid.\]

The hand-coded memory heap of UG has been removed from dune-uggrid about a
year ago. Hence, the heapSize parameter of UGGrid does not have any purpose
anymore. This patch removes it completely. It only leaves the public method
setDefaultHeapSize and marks it as deprecated, in order not to break people's
codes.

See merge request [!345]

  [!345]: gitlab.dune-project.org/core/dune-grid/merge_requests/345

Merge branch 'bugfix/find-parmetis-broken' into 'master'

See merge request [!355]

  [!355]: Nonecore/dune-grid/merge_requests/355

Merge branch 'feature/gridptr-can-read-gmsh' into 'master'

ref:core/dune-grid This MR adds the ability of reading gmsh files to the
GridPtr class. This needs testing by some gmsh users.

See merge request [!312]

  [!312]: gitlab.dune-project.org/core/dune-grid/merge_requests/312

Merge branch 'feature/shared-ptr-in-meta-grids' into 'master'

ref:core/dune-grid Previously, raw pointer were used to pass the host grid to
geometry grid and ownership of the object was transferred. This commit
switches this to the more modern way of using a shared_ptr for this case. The
old constructors are deprecated.

This fixes [#86]

See merge request [!322]

  [#86]: gitlab.dune-project.org/NoneNone/issues/86
  [!322]: gitlab.dune-project.org/core/dune-grid/merge_requests/322


Closes #86

The hand-coded memory heap of UG has been removed from dune-uggrid
about a year ago.  Hence, the heapSize parameter of UGGrid does
not have any purpose anymore.  This patch removes it completely.
It only leaves the public method setDefaultHeapSize and marks it
as deprecated, in order not to break people's codes.

Merge branch 'feature/uggrid-avoid-warnings' into 'master'

ref:core/dune-grid Avoid a few compiler warnings in UGGrid.

See merge request [!362]

  [!362]: gitlab.dune-project.org/core/dune-grid/merge_requests/362

Initialize grid-impl pointer to nullptr in default
constructor to avoid compiler warning.

Avoid warning in assertion by explicitly converting to int.

element and boundary parameters.

Merge branch 'feature/referenceelement' into 'master'

ref:core/dune-grid This MR adds a convenience method referenceElement for the
entity class. The following will now be possible:

    const auto& refElem = referenceElement( entity );

    // forwarding the call the referenceElement< ctype, dim > (entity.type() );

A similar method already exists for the Geometry, however, like the method
type on Entity, building the geometry maybe expensive and therefore using the
entities' type method is the appropriate choice here.

See merge request [!349]

  [!349]: gitlab.dune-project.org/core/dune-grid/merge_requests/349

Merge branch 'fix-vector-leaf-index' into 'master'

ref:core/dune-grid It is used to store certain leaf indices, so let's call it
'leafIndex'.

See merge request [!361]

  [!361]: gitlab.dune-project.org/core/dune-grid/merge_requests/361

It is used to store certain leaf indices, so let's call it
'leafIndex'.

Merge branch 'feature/update-authors' into 'master'

See merge request [!359]

  [!359]: Nonecore/dune-grid/merge_requests/359

Merge branch 'i386-tolerance' into 'master'

ref:core/dune-grid This fixes a test failure on i386 with recent versions of
GCC where the numbers looked identical, even when printed with increased
precision or in binary (hexfloat) format, for the test-dgf-yasp-offset test:

    InvalidStateException [checkEntityLifetimeForCodim:.../dune-grid-2.6.0/dune/grid/test/checkentitylifetime.hh:77]:
    ERROR! inconsistent corner(0) coordinate for entity 27
      (0 0.30000000000000004 != 0 0.30000000000000004)
      (0x0p+0 0x1.3333333333334p-2 != 0x0p+0 0x1.3333333333334p-2)

But the test still failed as i386's x87 FPU uses a higher precision
internally.

Reference:
[https://buildd.debian.org/status/fetch.php?pkg=dune-grid&arch=i386&ver=2.6.0-4&stamp=1570661057&raw=0]

See merge request [!356]

  [https://buildd.debian.org/status/fetch.php?pkg=dune-grid&arch=i386&ver=2.6.0-4&stamp=1570661057&raw=0]:
    https://buildd.debian.org/status/fetch.php?pkg=dune-grid&arch=i386&ver=2.6.0-4&stamp=1570661057&raw=0
  [!356]: gitlab.dune-project.org/core/dune-grid/merge_requests/356

This fixes a test failure on i386 with recent versions of GCC where
the numbers looked identical, even when printed with increased
precision or in binary (hexfloat) format, for the
`test-dgf-yasp-offset` test:

```
InvalidStateException [checkEntityLifetimeForCodim:.../dune-grid-2.6.0/dune/grid/test/checkentitylifetime.hh:77]:
ERROR! inconsistent corner(0) coordinate for entity 27
  (0 0.30000000000000004 != 0 0.30000000000000004)
  (0x0p+0 0x1.3333333333334p-2 != 0x0p+0 0x1.3333333333334p-2)
```

But the test still failed as i386's x87 FPU uses a higher precision
internally.

Reference: https://buildd.debian.org/status/fetch.php?pkg=dune-grid&arch=i386&ver=2.6.0-4&stamp=1570661057&raw=0

Merge branch 'issue/test_tolerances' into 'master'

ref:core/dune-grid

### Summary

Change the fixed hard coded tolerances in grid tests with something related to
numeric_limits<ctype>::epsilon().

### Details

Many tolerances in the grid tests are hard coded to something like 1.e-8.
Those tests will fail if ctype=float. Thus, all those comparisons are changed
to either sqrt(numeric_limits<ctype>::epsilon()) or
numeric_limits<ctype>::epsilon(), depending on the previous value.

### Discussion

The tolerance values seem quite arbitrary. Sometimes it is compared against
1.e-8, sometimes 1.e-7 or 1.e-6 or 1e3*numeric_limits<ctype>::epsilon(). Maybe
it would be cleaner to use the FloatCmp utilities and clearly define what's
the expected accuracy.

See merge request [!352]

  [!352]: gitlab.dune-project.org/core/dune-grid/merge_requests/352

Merge branch 'gmsh_reader_fix' into 'master'

ref:core/dune-grid

-   otherwise on windows it is opened in text mode
    -   ftell/fseek on windows only work correctly on binary files
    -   see e.g.
        [https://stackoverflow.com/questions/47256223/why-does-fseek-0-seek-cur-fail-on-windows]
    -   this makes no difference on posix systems

also: use ftello/fseeko with type off_t instead of ftell/fseek with type long

-   [https://www.gnu.org/software/libc/manual/html_node/File-Positioning.html]
-   "using ftello is preferable whenever it is available"

See merge request [!353]

  [https://stackoverflow.com/questions/47256223/why-does-fseek-0-seek-cur-fail-on-windows]:
    https://stackoverflow.com/questions/47256223/why-does-fseek-0-seek-cur-fail-on-windows
  [https://www.gnu.org/software/libc/manual/html_node/File-Positioning.html]: https://www.gnu.org/software/libc/manual/html_node/File-Positioning.html
  [!353]: gitlab.dune-project.org/core/dune-grid/merge_requests/353

Merge branch 'use-SOURCES-in-dune-add-library' into 'master'

ref:core/dune-grid dune_add_library implicitly treats all unclaimed trailing
arguments as source file names, but explicitly marking those as SOURCE makes
the code more readable.

See merge request [!354]

  [!354]: gitlab.dune-project.org/core/dune-grid/merge_requests/354

dune_add_library implicitly treats all unclaimed trailing
arguments as source file names, but explicitly marking
those as SOURCE makes the code more readable.

  - otherwise on windows it is opened in text mode
  - ftell/fseek on windows only work correctly on binary files
  - see e.g. https://stackoverflow.com/questions/47256223/why-does-fseek-0-seek-cur-fail-on-windows
  - this makes no difference on posix systems

also use ftello/fseeko with type off_t instead of ftell/fseek with type long
  - https://www.gnu.org/software/libc/manual/html_node/File-Positioning.html
  - "using ftello is preferable whenever it is available"

Merge branch 'bug/do-not-crash-paraview-on-macos' into 'master'

ref:core/dune-grid Paraview (well, probably VTK) on macOS crashes when reading
ASCII files with denormalized float values. This seems to be due to libc++'s
IO stream implementation, which sets the fail bit when reading a denormalized
value. On top of that, Paraview just segfaults when encountering a fail bit
while loading VTK files.

libstdc++ does not set the failt bit, and I don't know what is the correct
behavior, but having Paraview mysteriously crash on DUNE output files is not
good.

So this patch truncates subnormal floating point values to 0 when writing
ASCII. This should not seriously influence any visualization results, but it
might trip up people who use exact floating point comparisons for tests (but
then you shouldn't do that anyway\...).

See merge request [!350]

  [!350]: gitlab.dune-project.org/core/dune-grid/merge_requests/350

Paraview (well, probably VTK) on macOS crashes when reading ASCII files with
denormalized float values. This seems to be due to libc++'s IO stream
implementation, which sets the fail bit when reading a denormalized value.
On top of that, Paraview just segfaults when encountering a fail bit while
loading VTK files.

libstdc++ does not set the failt bit, and I don't know what is the correct
behavior, but having Paraview mysteriously crash on DUNE output files is not
good.

So this patch truncates subnormal floating point values to 0 when writing ASCII.
This should not seriously influence any visualization results, but it might trip
up people who use exact floating point comparisons for tests (but then you
shouldn't do that anyway...).

Merge branch 'bugfix/gridtype-uggrid' into 'master'

ref:core/dune-grid Small fix to make grid type definition in config.h work for
UGGrid.

See merge request [!333]

  [!333]: gitlab.dune-project.org/core/dune-grid/merge_requests/333

for an entity by calling referenceElement( entity ).

Merge branch 'feature/gmsh-reader-boundarysegments-1d-grids' into 'master'

ref:core/dune-grid Boundary segments were not inserted for 1d grids (where
boundary segments are single nodes). However this is needed to e.g. attach
boundary data to a `Dune::FoamGrid<1, 3>` or Dune::OneDGrid.

-   Fixes [#93]
-   Adds a test for 1d gmsh boundary data with OneDGrid

See merge request [!348]

  [#93]: gitlab.dune-project.org/NoneNone/issues/93
  [!348]: gitlab.dune-project.org/core/dune-grid/merge_requests/348


Closes #93

Merge branch 'feature/sumouternormalerror-warning' into 'master'

ref:core/dune-grid The check is only true for affine elements and if the
element is not part of a network where intersections can overlap. The 'error'
will still be propagated as a warning in these cases.

With this change the grid check passes for all tested dune-foamgrid
configurations which previously failed due to this one check.

See merge request [!332]

  [!332]: gitlab.dune-project.org/core/dune-grid/merge_requests/332

Merge branch 'cleanup-1' into 'master'

See merge request [!347]

  [!347]: Nonecore/dune-grid/merge_requests/347