fully_sampled_recon_three_chains fails

[KrisThielemans]

At present, image_array has shape [0,0,0]. complex_image_data seems alright though (its shape after as_array is [2,256,256]).

I haven't checked the corresponding SIRF demo.

[evgueni-ovtchinnikov]

fully_sampled_recon_three_chains.ipynb works fine on my laptop

[KrisThielemans]

@evgueni-ovtchinnikov I've moved this issue here. I have checked that on my VM both the notebook and actual demo work fine. However, on an archlinux install both fail. This could be a gadgetron server issue but Gadgetron's unit-tests and integration tests work. I have no clue what to do about it. would it help if I see if I can give you access to the archlinux system?

[evgueni-ovtchinnikov]

@KrisThielemans: access to archlinux would certainly help - how else can i investigate what goes wrong there :)
what is archlinux anyway?

[KrisThielemans]

we've seen problems with the other Gadgetron demos returning only 1 image as opposed to both repetitions. Seems to be related to reading the ISMRMRD file.

[KrisThielemans]

@evgueni-ovtchinnikov I've tried rebuilding with our version of HDF5 as opposed to the system one. Same problem unfortunately.

[KrisThielemans]

we have this same problem now on the new VM. Everything is fine until here https://github.com/CCPPETMR/SIRF/blob/ea4a56f9f07ae5a17d74887fbca002dc69313b5c/examples/Python/MR/Gadgetron/fully_sampled_recon_three_chains.py#L109

c=complex_image_data.as_array()
c.shape

returns (2, 256, 256), and a display of the numpy.abs(c) looks fine.

However, the extracted array is not https://github.com/CCPPETMR/SIRF/blob/ea4a56f9f07ae5a17d74887fbca002dc69313b5c/examples/Python/MR/Gadgetron/fully_sampled_recon_three_chains.py#L114-L118

In [44]: real_image_data.number()
Out[44]: 0

In [45]: real_image_data.as_array().shape
Out[45]: (0, 0, 0)

[KrisThielemans]

This could be a Gadgetron change/bug I guess.

[evgueni-ovtchinnikov]

caused by bugs in new ExtractGadget, Gadgetron issue gadgetron/gadgetron#743 created

[KrisThielemans]

They fixed one bug in master quite a while ago. We need to retest this with current Gadgetron master

[paskino]

If you want to try to build Gadgetron master there's a SuperBuild branch and also a related draft PR SyneRBI/SIRF-SuperBuild#438