Summary

Actual behavior

Passing a white matter segmentation with the 'wmseg' input to a Node using the EpiReg interface raises the following Error:

TraitError: The 'seg' trait of an EpiRegOutputSpec instance must be a pathlike object or string representing an existing file but a value of "/Users/olivercontier/bigfri/scratch/preproc_bbr_workdir/sub-01/sub_wf/ses_wf/_ses_i_0/bbr/epi2struct_fast_seg.nii.gz" <class 'str'> was specified.

Expected behavior

FSL's epi_reg script does not perform segmentation when a white matter segmented file is already provided. Hence, the interface should not check whether the 'seg' output exists in that case.

I think this should be fixable by changing the behavior of _list_outputs() method of the EpiReg class.

How to replicate the behavior

I tried to provide more details in my post on neurostars.

Platform details:

{'commit_hash': '07af08f98',
 'commit_source': 'repository',
 'networkx_version': '2.4',
 'nibabel_version': '3.1.1',
 'nipype_version': '1.6.0-dev+g07af08f98',
 'numpy_version': '1.18.5',
 'pkg_path': '/Users/olivercontier/bigfri/scratch/bids/code/sandbox/nipype_epireg/nipype/nipype',
 'scipy_version': '1.5.2',
 'sys_executable': '/opt/anaconda3/envs/bigfri/bin/python',
 'sys_platform': 'darwin',
 'sys_version': '3.8.5 | packaged by conda-forge | (default, Aug 20 2020, '
                '20:51:28) \n'
                '[Clang 10.0.1 ]',
 'traits_version': '6.1.0'}

Execution environment

My python environment outside container