On Fri, Dec 10, 2021 at 2:40 PM Andreas Tille <[email protected]> wrote: > > Hi Mathieu, > > Am Fri, Dec 10, 2021 at 02:31:20PM +0100 schrieb Mathieu Malaterre: > > On Fri, Dec 10, 2021 at 12:22 PM Étienne Mollier <[email protected]> > > wrote: > > [...] > > > My impression by looking at it to unroll is that something seems > > > caught in a recursive loop. > > > > The bug is upstream. One cannot do: > > > > image = gdcm.Image() > > > > This will not work with the simple shared pointer implementation in > > upstream gdcm. > > I'f volunteer to open an upstream issue about this - but could you > give some more verbose information what will work instead?
You can copy/paste the following [...] GDCM is a C++ project, that is wrapped to python using swig. It use a simple reference counting mechanism (c++98 style), where application programmer are required to write: gdcm::SmartPointer<gdcm::Image> im = new gdcm::Image; Since SmartPointer is a class template it is not accessible from python. A predefined set of class have been explicitly wrapped to python using something like: %template(SmartPtrScan) gdcm::SmartPointer<gdcm::Scanner>; this is not the case for gdcm.Image as of release 3.0.10. --- A potential solution would be to use an image instance from an ImageReader/ImageWriter . writer = gdcm.ImageWriter() proper_image_ref_count = writer.GetImage() [...] > > I would just remove the gdcm test suite from pydicom. > > I'll do so - but it would be great to get this working in the next > release. Agreed. Thanks
