FTBFS with HDF5 1.10 due to a change of fileid type

[ghisvail]

Just forwarding an issue we had with compilation of ismrmrd on Debian following the transition of the archive to hdf5 v1.10.

Essentially, it all comes down to a change in the HDF5 API regarding the returned type of H5Fopen, which is no longer an int but a long long in v1.10. Actually, HDF5 encapsulates this return type to a hid_t, and the patch proposed by Gilles uses this type for ISMRMRD_Dataset.fileid which eventually fixes the problem.

Are you happy with this change? If so, I can forward Gilles' patch in a subsequent PR.

[hansenms]

@ghisvail sorry for the slow response here. I also need to deal with the other PR you put up.

But yes, sounds like a reasonable change. Would that be backwards compatible with 1.8?

[ghisvail]

@ghisvail sorry for the slow response here. I also need to deal with the other PR you put up.

No problem.

Would that be backwards compatible with 1.8?

Gilles already pushed the change in the archive which still has 1.8 and the build and test suite ran fine. Since hid_t was 32-bit on HDF5 1.8, the change remains compatible with the old signature of ISMRMRD_Dataset.

[hansenms]

Great, thanks! I will watch for the PR then.