Resolve the link before passing it to perltidy.

for q in * ; do perltidy $(readlink -e "$q") ; done
[download]

You could also modify perltidy to resolve links using Cwd's realpath.

In both cases, the backup will appear next to the linked file.

plus the documentation says -b modifies in place (not quite true in this case.)

Truly modifying in place can require lots of memory (since you can't insert into a file, just overwrite), and it fails bad (you lose data if an error occurs halfway through). You should expect "in place" to mean the final result is stored in a file by the same name as the original.

This affects symlinks, and it can also affect permissions. Perhaps a better solution would be to copy the original, then read from the backup and overwrite the original. I know some were interested in changing Perl's -i, but I don't know if that's one of the changes they had in mind. It would slow things down.

Something like

perltidy script.pl && cat script.pl.tdy >script.pl && rm script.pl.tdy
[download]

Note, that it doesn't create backup.

Update: another approach is to dereference symlink:

perltidy -b `readlink script.pl`
[download]

Thanks, no backup isn't much of an issue since the file is under version control anyway.