As Jim said, comm needs both input files to be sorted. After case-shifting both of your sample input files to lowercase, they happen to be in sorted order. If that is not the case with your real data files, you will also need to sort them after shifting to lowercase.

I don't think you want the tr -s option. That suppresses repeated adjacent occurrences of the same character in the output. (For example:

Code:

echo HELLO | tr -s '[:upper:]' '[:lower:]'

would produce the output:

Code:

helo

note the LL in the input and the single l in the output.)

The process substitution feature Jim suggested is available in bash, some recent versions of ksh, and a few other shells; but it is not in the standards and is not available in many other shells. If you're using a shell that just supports POSIX standard features (and your input might need to be sorted), you could try something more like:

Code:

tr '[:upper:]' '[:lower:]' < x2.txt | sort > $$.2.txt
tr '[:upper:]' '[:lower:]' < x3.txt | sort > $$.3.txt
comm $$.[23].txt
rm -f $$.[23].txt

which creates copies of your input files that have been converted to lowercase and sorted, runs comm on the copies, and then removes the copies.