If you’ve ever ended up with a directory structure like this

dir1/dir1/...

you know the struggle is real.

What is the potential pitfall with scp -r?

If you wrote something like

scp -r dir1 "${SERVER}:dir1"

I see a red flag. What you are trying to do is copying dir1 from your machine to a server. This will work well the first time:

scp -r dir1 "${SERVER}:dir1"
ssh "${SERVER}" ls dir1 // output: hello.txt

But try this a second time, and you’ll get:

scp -r dir1 "${SERVER}:dir1"
ssh "${SERVER}" ls dir1 // output: dir1, hello.txt

How can you make scp -r work in an idempotent fashion?

The answer here is simple: Just omit the name of the target directory. It will then automatically use the name of the source directory:

scp -r dir1 "${SERVER}:~"
ssh "${SERVER}" ls dir1 // output: hello.txt

And testing for idempotency:

scp -r dir1 "${SERVER}:~"
ssh "${SERVER}" ls dir1 // output: hello.txt

Conclusion

Having something like scp -r MYDIR ...MYDIR is a red flag. You should probably simply drop the second MYDIR and you are good to go!