FSL (+6.0) BET chokes with long filenames

[oesteban]

Would adding the copyfile metadata just fix this, @satra?

I think I've seen this issue reported before, but I can't seem to find it. Feel free to link and close this if it is a duplicate

[satra]

is it a path length issue or a filename issue or a command path length issue?

we could use copyfile=False and change the command to use a relative path. but that interface (one of our oldest), also returns absolute path for the output file name, which bet needs to run.

[oesteban]

is it a path length issue or a filename issue or a command path length issue?

Not sure, it could definitely be that it doesn't like nipype's parameterized folders (meaning, a filename issue). What is the difference between path length issue and command path length issue?

we could use copyfile=False and change the command to use a relative path. but that interface (one of our oldest), also returns absolute path for the output file name, which bet needs to run.

The output file is the first thing I tried shortening, so you're probably right on this one too. Which means there's not an immediate fix for this, right?

[oesteban]

Okay, this also happens with the output:

terminate called after throwing an instance of 'NiftiIO::NiftiException'
  what():  Error: cant open file /home/oesteban/tmp/fmriprep-ds210/work/fmriprep_wf/single_subject_02_wf/func_preproc_task_cuedSGT_run_03_echo_1_wf/skullstrip_bold_wf/_bold_file_..oak..stanford..groups..russpold..data..openfmri..ds000210..sub-02..func..sub-02_task-cuedSGT_run-03_echo-3_bold/skullstrip_first_pass/vol0000_xform-00000_merged_brain.nii.gz

[satra]

is it possible that just this part ..oak..stanford..groups..russpold..data..openfmri..ds000210..sub-02..func..sub-02_task is throwing things off?

[oesteban]

Doesn't seem so:

bet derivatives/smriprep-0.6/sub-02/anat/sub-02_desc-preproc_T1w.nii.gz _bold_file_..oak..stanford..groups..russpold..data..openfmri..ds000210..sub-02..func..sub-02_task-cuedSGT_run-03_echo-3_bold/skullstrip_first_pass//some.file.nii.gz

works okay

[satra]

so the easiest fix would be:

1. use copyfile=False for the input file to link locally
2. make the output filename be local (no absolute path) for generation, but not for list_outputs.

bet allows the user to send an output location, so that would have to remain as is.

[oesteban]

Yup, I imagined this was the solution.