bf: resolve bashism with minimal impact

[octomike]

We found a bashism in bedpostx_mgh when running trac-all and I used the tool checkbashisms in the debian devscripts package to

identify more possible issues within the scripts/ folder only:

for f in $(find . -type f) ; do head -n1 $f | grep -q '/bin/sh' && (checkbashisms $f || echo -e "\n\n$f\n\n") ; done

These are the results:

possible bashism in ./mri_motion_correct line 65 (should be >word 2>&1):
  mri_convert --in_info --read_only $1 &> $transcript_file
possible bashism in ./mri_motion_correct line 215 (should be >word 2>&1):
which which &> /dev/null
possible bashism in ./mri_motion_correct line 222 (should be >word 2>&1):
  which basename &> /dev/null
possible bashism in ./mri_motion_correct line 296 (should be >word 2>&1):
  which mri_convert &> /dev/null
possible bashism in ./mri_motion_correct line 299 (should be >word 2>&1):
  mri_convert &> /dev/null
possible bashism in ./mri_motion_correct line 323 (should be >word 2>&1):
    $p &> $transcript_file
possible bashism in ./mri_motion_correct line 347 (should be >word 2>&1):
  mri_convert $infile $outfile &> $transcript_file
possible bashism in ./mri_motion_correct line 372 (should be >word 2>&1):
    mri_convert $f $tmpdir_use/$i.mnc &> $transcript_file
possible bashism in ./mri_motion_correct line 392 (should be >word 2>&1):
      minctracc -lsq6 $tmpdir_use/$i.mnc $tmpdir_use/0.mnc $tmpdir_use/${i}_to_0.xfm &> $transcript_file
possible bashism in ./mri_motion_correct line 407 (should be >word 2>&1):
      mincresample -transformation $tmpdir_use/${i}_to_0.xfm -like $tmpdir_use/0.mnc $tmpdir_use/${i}.mnc $tmpdir_use/${i}_to_0.mnc &> $transcript_file
possible bashism in ./mri_motion_correct line 433 (should be >word 2>&1):
  mincaverage $volumes_to_average $tmpdir_use/avg.mnc &> $transcript_file
possible bashism in ./mri_motion_correct line 448 (should be >word 2>&1):
  mri_convert $tmpdir_use/avg.mnc $outfile &> $transcript_file


./mri_motion_correct


possible bashism in ./bedpostx_mgh line 129 (bash arrays, ${name[0|*|@]}):
if [ -z ${fslver[1]} ]; then fslver[1]=0; fi
possible bashism in ./bedpostx_mgh line 129 (bash arrays, H[0]):
if [ -z ${fslver[1]} ]; then fslver[1]=0; fi
possible bashism in ./bedpostx_mgh line 130 (bash arrays, ${name[0|*|@]}):
if [ -z ${fslver[2]} ]; then fslver[2]=0; fi
possible bashism in ./bedpostx_mgh line 130 (bash arrays, H[0]):
if [ -z ${fslver[2]} ]; then fslver[2]=0; fi
possible bashism in ./bedpostx_mgh line 131 (bash arrays, ${name[0|*|@]}):
if [ ${fslver[0]} -lt 6 ]; then
possible bashism in ./bedpostx_mgh line 132 (bash arrays, ${name[0|*|@]}):
    echo "Outdated FSL version (${fslver[0]}.${fslver[1]}.${fslver[2]})"


./bedpostx_mgh


possible bashism in ./map_to_base line 62 (should be 'b = a'):
if [ "$rt" == "surface" ] ; then ldir=surf ; fi


./map_to_base

The pull requests fixed the bashisms by translating the == comparison to ['s = and the bash redirection shortcut &> x to > x 2>&1 . The bedpostx_mgh script however relies on bash arrays and so I decided to switch to interpreter to bash.

No testing was done, but the changes are very trivial.