Remember : there is no need to use '-r' option under HP-UX. Those tests have been done on HP-UX. You must use '-r' with Linux.

Here is the content of my file : 

- 1 line with 14 x '1', one space, one newline
- 1 line with 14 x '2', one tabulation, one newline

Controled with od :

#od -c MyFile                             

0000000   1   1   1   1   1   1   1   1   1   1   1   1   1   1      \n
0000020   2   2   2   2   2   2   2   2   2   2   2   2   2   2  \t  \n
0000040

Now we try with sed and the syntax that should work. Line one is correctly edited, line 2 not. It's normal as last character is a tabulation :

#sed 's/[ ]*$//' <MyFile | od -c                             

0000000   1   1   1   1   1   1   1   1   1   1   1   1   1   1  \n   2
0000020   2   2   2   2   2   2   2   2   2   2   2   2   2  \t  \n
0000037

Now we try with Perl. Great, it works, but why ? Because the '\s' means blankspace and not only space. Blank space is characters producing white-space in displayed text. It includes spaces and tabulations.

#perl -pi -e 's/\s+$/\n/g' <MyFile | od -c

0000000   1   1   1   1   1   1   1   1   1   1   1   1   1   1  \n   2
0000020   2   2   2   2   2   2   2   2   2   2   2   2   2  \n
0000036

So we can adapt now sed so that it works as well as perl. Two way for that :

--> Use special naming convention [:space:] wich means blankspace as '\s' in sed.

#sed 's/[[:space:]]*$//' <MyFile | od -c  

0000000   1   1   1   1   1   1   1   1   1   1   1   1   1   1  \n   2
0000020   2   2   2   2   2   2   2   2   2   2   2   2   2  \n
0000036

--> If Linux implementation of sed don't know [:space:] (well, it should ;-) it is a litle bit more verbose :

#SPACE=`echo " \t\c"`
#sed 's/['"$SPACE"']*$//' <MyFile | od -c