It is the first error that occurs in the log. I tried running feat on the
same data using my MacBookPro (with the precompiled release from the
download page) and it went just fine. I didn't let it finish, but it was
processing for a few minutes with no problems at all.
I was able to run fslmaths on the inputs as you specified in your reply on
my MacBook.
After seeing that the data does indeed work somewhere else, I suspected it
might be the compilation we used (the precompiled programs fail immediately
with floating point exception errors on our machines). The FSLMACHTYPE
returned by fslmachtype.sh is linux_64-gcc4.2, so I copied linux_64-gcc4.1
and used that for the compilation. Unfortunately, I still got the same errors.
As for your suggestion about the 32/64 bit file system problem, are you
referring to the actual address length used to denote the blocks on the disk?
I'm also attaching the report_log.html from a run of feat on the same input
data using the newly compiled FSL installation using the computer's local
file system (not the nfs mounted /ifs/tmp referred to in my earlier
postings). Unfortunately, I'm still getting the same error.
Before I actually try it, do you think it's worth it/recommended to hard
code fslmachtype.sh to return a 32-bit architecture, and recompile a 32-bit
version of FSL?
Thank you very much for your help Mark.
Best,
Arash and Liberty
|