In an example dataset applyxfm4D was not setting the TR correctly. The TR was reported as 0.7 before any processing, and after applyxfm4D is was 1.0
This is related to debugging this list quesion:
https://www.jiscmail.ac.uk/cgi-bin/wa-jisc.exe?A2=FSL;b548c010.2009
I looked at how flirt was doing this, and flirt.cc sets the out.tdim from the input.tdim, so I added this same functionality to applyxfm4D.cc.
In my tests (on my mac) the TR is correctly set to the input TR (0.7) with this change.
I think I shared the preproc data with you guys on Oxfile just now (not clear if it worked). I can upload the example data to jalapeno if needed. It's from the person on the list.