The dynamic analysis tests have uncovered a number of floating point
exceptions, mainly divide by zero. This MR enables floating point
exceptions if tests are built with a debug build type and the
compiler is a gnu compiler.

Safety guards have been added to classes that exhibited exceptions.

TestAdjustLowerBoundForLogPlot.cxx - Tried to take log10 of negative number

TestlinePlotDouble - inf and nan used in test. Disable exceptions

vtkChart.cxx - guards to protect against zero divide

vtkChartBox.cxx - guards to protect against zero divide

vtkInteractiveArea.cxx - trying to take log10 of negative number

vtkPiecewisePointHandleItem.cxx - protect against degenerate lines

vtkAMRSliceFilter.cxx - protect against zero length vector

vtkLagrangianBasicIntegrationModel.cxx - guards to protect against zero divide

UnitTestMultiThreshold - disable exceptions for one test that uses Nan()

vtkIntersectionPolyDataFilter.cxx - guard acos range

vtkPCAStatistics.cxx - guards to protect against zero divide

vtkCubeAxesActor.cxx - guard against log10(0.0)

vtkColorTransferFunction.cxx - guard to protect against zero divide

TestGPURayCastIsosurface.cxx - disable exceptions, possible driver issue

vtkDendrogramItem.cxx - check for degenrate line

vtkDendrogramItem.h - comparison doesn't meet irreflexive requirements

Marked as legacy in 8.2.

These new baselines are largely for new text rendering from the newer
freetype.

Same speed but much better end result than the old
transparency code.

Also fix a few classes that were not passing properties
to their sub props.

Fix medical example to make opaque objects opaque.

Found via `codespell`

Ken Martin authored 6 years ago and Brad King committed 6 years ago

make image difference symmetric to catch some
failures that were slipping through. This exposed
a number of valid images that were no longer
correct and needed to be replaced.

Kitware Robot authored 6 years ago and Brad King committed 6 years ago

Populate a script with `map_<md5>=<sha512>` variable settings and
then run the following to convert the content link files:

    git ls-files -- '*.md5' | while read f; do
      md5="$(cat "$f")"
      sha512="$(eval echo \${map_$md5})"
      rm "$f"
      echo $sha512 > ${f%.md5}.sha512
    done

Found via `codespell` and `grep`

154 occurances of ;;\n

This required some clean up, because the automatic changes borked a couple of files, but mostly resulted in stupid whitespacing. A few find/replace fixed things up.

There are now some duplicate semis because it did changes like:

  ~AMRIndexIterator() override{};
to
  ~AMRIndexIterator() override= default;;

Note there was a pointless semi before, which clang-tidy didn’t expect / account for.

My next commit will remove them.

The old MapDataArrayTGoMultiTexture relied on
being able to lock dowwn a specific texture unit
which cannot be guaranteed as external code
or render passes may bne uising that texture unit.

Instead we support the method having the name of a texture
along with the matching ethod in vtkProperty to name
a texture. Between these two users can assign texture
coordinate arrays to specific textures.

Deprecated the old index and unit based methods.
Removed a no longer needed vtkTexture.h include
which may bite some folks who are using vtkTexture
but not including the header (which they shoudl do)

found via `codespell` 

https://stackoverflow.com/questions/13816385/what-are-the-advantages-of-using-nullptr
C++11 introduces nullptr, it is known as the Null pointer constant and
It improves type safety and resolves ambiguous situations unlike the
existing implementation dependent null pointer constant NULL.

https://msdn.microsoft.com/en-us/library/4ex65770.aspx
The nullptr keyword can be used to test if a pointer or handle reference
is null before the reference is used. Function calls among languages
that use null pointer values for error checking should be interpreted
correctly. You cannot initialize a handle to zero; only nullptr can be
used.

Found with regexes:

vtkBooleanMacro.*int\)
vtkBooleanMacro.*int \)

This commit removes double whitespace from source comments + fixes some other typos.

Typos fixes for documentation and source code comments. This commit is a follow up to:   !3424

This reverts commit 7458e28b.

because they are not VTK objects. So this example does not work
with the current wrapping framework.

This topic catches another batch of NULL conversions that
clang did not handle for various reasons. It also includes
a number of changes to comments and strings where NULL was
used. It also catches some recent code changes where NULL
was used instead of nullptr.

Now that VTK requires a C++11 compiler we can ditch using custom typedefs to
wrap =delete;