[Debian-med-packaging] Bug#841959: Problems with your ismrmrd/1.3.2-4.1 NMU

[Ghislain Vaillant]

On 01/11/16 17:50, Gilles Filippini wrote:
>> Could you give me more insight of the compatibility problem with HDF5 1.10?
>
> The most important change in the HDF5 1.10 release is that the hid_t
> type was changed from 'int' to 'long long' [1]. Assuming that hid_t is
> int isn't true anymore. Thus the change to the ISMRMRD_Dataset
> structure, because the fileid field is actually used as an hid_t. See
> for example the ismrmrd_open_dataset() function in libsrc/dataset.c:
>
> int ismrmrd_open_dataset(ISMRMRD_Dataset *dset, const bool
> create_if_needed) {
>     /* TODO add a mode for clobbering the dataset if it exists. */
>     hid_t fileid;
>     ...
>     fileid = H5Fopen(dset->filename, H5F_ACC_RDWR, H5P_DEFAULT);
>
>     if (fileid > 0) {
>         dset->fileid = fileid;
>     }
>     ...
>
> I can see two solutions:
> 1- make libismrmrd-dev depend on libhdf5-dev
> 2- set fileid as type long long to avoid including hdf5.h
>
> What do you think?

Considering the fix results in a change of API, I prefer polling 
upstream for an acknowledgement of the proposed solution:

   https://github.com/ismrmrd/ismrmrd/issues/73

Many thanks for the summary of the problem.

Ghis