What use rm -rf "`pwd`/folder instead of rm -rf ./folder

Question

Why would I want to use the command

rm -rf "`pwd`/folder"

instead of

rm -rf ./folder

What reasons are there to prefer one over the other?

Answer 1

rm -rf "`pwd`/folder" is bad and you shouldn't use it. It's only bad due to a corner case, but this corner case could turn out to be a security issue leading to deleting the wrong directory. The corner case is if the current directory ends with a newline character: the command substitution strips trailing newlines, so if the current directory is /some/where/strange⏎ (where ⏎ stands for a newline), this command would delete /some/where/strange instead.

The variant rm -rf "$PWD/folder" has no downside over using the pwd command, and has the advantage that it doesn't suffer from the corner case of a trailing newline. Every Bourne/POSIX shell updates the PWD variable after each directory-changing command (cd, pushd, etc.) and it has the same value that the pwd built-in command outputs. (The external pwd command might have a different output, but 1. pwd calls the built-in if there is one and all common shells have one, and 2. there is no advantage to using the external command here anyway.)

rm -rf "$PWD/folder" has no advantage over rm -rf ./folder, however. It has the theoretical downside that rm -rf "$PWD/folder" might go over the maximum command line length while rm -rf ./folder fits. However the maximum command line length is almost always significantly larger than the maximum file path length, so a command with a single path pretty much always fits.

Using "$PWD/$foo" instead of "$foo" has the advantage that it's guaranteed not to start with a dash, so it won't look like an option. That's not an issue if the path is spelled out explicitly in the script, only when it's a variable. And "./$foo" offers the same protection. Both "$PWD/$foo" and "./$foo" are correct only when $foo is a relative path. So rm -rf "$PWD/$foo" has no advantage over rm -rf "./$foo". And as seen above, if there are many paths involved, using the shorter relative form can help to stay under the command line length limit.

Using "$PWD/folder" has a clear benefit over "./folder" if you're storing it in a variable for later use: it keeps designating the same file even if the script changes to another directory in the meantime. While rm -rf "./folder" is perfectly fine, code like this requires an absolute path:

output_directory="$PWD/output"
# If interrupted, delete the partial output.
trap 'rm -rf "$output_directory"; exit 1' HUP INT TERM
do_stuff >"$output_directory/file1"
do_more_stuff >"$output_directory/file2"
cd "$output_directory"
further_stuff file1 file2 >file3

If $output_directory was a relative path, the cleanup code could work before the cd command or after, but not both.