Sometimes we need more decimal digits of precision when converting
double values to ASCII than what is afforded by %g, so use %.*g with
std::numeric_limits<double>::max_digits10 to specify the precision.

Hat tip to Simon Praetorius <Simon.Praetorius@tu-dresden.de> for
pointing out a regression in the XML readers and test data to use
in this test. See
https://www.vtk.org/pipermail/vtkusers/2016-September/096663.html.

There were surprisingly few.  The benefit is that when additional enums are added, a warning will point you to consider every switch statement.

In 2 cases, made trivial use of NewInstance, so that it’s not unused.

attempt to fix issues from
vtk/vtk!4916

ffmpegvideo source recognized stereo video now

vtkSkybox has a mode to properly display it

The previous implementation os vtkSEPReader had a use-after-free.

The function `vtkSEPReader::ReplaceFileName(..)` started with:

```c++
  delete[] this->FileName;
```

And that means that the string pointed by `FileName` was
destroyed. Even if the value of `FileName` was stored in the local
variable `fileName`, in `vtkSEPReader::RequestData`, the string
pointed by it was destroyed!

The patch of this commit is a lot simpler: both the string pointed by
`FileName` and the one that will replace it already exist in memory,
and will not be modified during the run of `RequestData`. So we can
just swap the pointers, without any need to memory
allocation/deallocation to copy the strings.

Fixes compilation error on MSVC 2015.

remove use of vtkJson namspace that should have been just Json

error: ‘unique_ptr’ is not a member of ‘std’

The const qualifier for the return type of
vtkGLTFExporter::WriteToString() is not needed. The return value is a
copy of a string, so const isn't needed to protect internal state of
the vtkGLTFExporter object.

vtkImageReader can read raw binary files. vtkImageReader2, which is
vtkImageReader's superclass, provides a `CanReadFile` that always
returns 0. Hence we need to provide this method otherwise vtkImageReader
will not say it reads raw files.

Also adds corresponding mappings in the various backend Context2D
devices.  The new stipple pattern is mapped similarly to the DOT_LINE
pattern, but with twice the number of dots.  This has the effect of
making it look a little less "sparse".

You can now inline the buffer data (base64) as
well as write to a string

Its automated fix was to use std::to_string(), but in fact I think everything it found was a false positive really, so I used an explicit cast instead.

and a simple test for it

Jerome Dubois authored 6 years ago and Sebastien Jourdain committed 6 years ago

This work was done by the CEA:
- HyperTreeGrid class now inherits from vtkDataObject instead of vtkDataSet
- Improved HyperTree (HT) class while keeping HT flexible.
- Complete rewrite of (Super)Cursors, more robust and efficient implementation.
- Propagate HT/Cursors refactoring to filters/readers/sources.
- Further optimizations of some filters.

Our thanks go to Los Alamos National Laboratory and Kitware for their support
in the integration of our developments.

Thanks @sebastien.jourdain, @demarle and @patchett2002.

ToDo (@CEA), review comments in code.

Co-authored-by: Jerome Dubois <Jerome.Dubois@cea.fr>
Co-authored-by: Guenole Harel <Guenole.Harel@cea.fr>
Co-authored-by: Jacques-Bernard Lekien <Jacques-Bernard.Lekien@cea.fr>

FFMPEG 2.2.3 provides libswscale, so we can just require that it exist.

Our minimum version of 2.3.3 provides:

  - libavcodec 55.69.100
  - libavformat 55.48.100
  - libavutil 52.92.100

which means we can remove all of our conditional code checks since
they're all guaranteed to be satisfied by the overall minimum FFMPEG
version.

With the new code, we shouldn't be using deprecated code anymore.