Neel Sharma
/about /projects /writing
Back to writing

Running UNSW CSE commands when working locally

7 June 2026

I’m currently taking COMP6991, UNSW’s Rust course. The course is great, however, due to the way Rust is set up on UNSW CSE’s servers, you cannot use any remote development tools, like SSHFS or Zed’s Remote Development feature, to complete coursework on their servers. This means that you lose the ability to run autotest, the test suite provided by the course, unless you do your work entirely inside SSH or VNC.

To bypass this restriction, my tutor recommended CSERun - it’s a simple tool that copies your current working directory from your computer to CSE’s servers and then runs a given command inside the remote copy - so you can copy your local project to CSE’s server and then run autotest in the server.

…but there is absolutely no need for this to be a 4 MB Rust binary with 96 crate dependencies! (Yes, I built it myself and checked!)

So, here’s the same thing, implemented in a 20-line shell script:

#!/usr/bin/env sh
set -euo pipefail

if [ $# -eq 0 ]; then
    echo "Usage: $0 <command> [command to run...]" >&2
    exit 1
fi

CWD_NAME=$(basename "$PWD")
REMOTE_HOST="<INSERT ZID HERE>@login.cse.unsw.edu.au"
REMOTE_PATH="~/.runsw"

# copy cwd to remote, ignoring .gitignore and .git
rsync -azP --delete \
    --exclude=.git \
    --filter="dir-merge,- .gitignore" \
    ./ "$REMOTE_HOST:$REMOTE_PATH/$CWD_NAME/"

# run provided command
ssh -t "$REMOTE_HOST" "cd $REMOTE_PATH/$CWD_NAME && $*"

Now, credit where credit is due, this version is less user-friendly than CSERun. Specifically, you need to create the ~/.runsw directory in the CSE server before using the script, otherwise rsync will fail.

Additionally, for a more frictionless experience, it’s best if you set up SSH keys for logging onto the CSE servers. If you don’t, you’ll have to type in your zPass a lot. To create SSH keys and register them with CSE, follow these steps:

  1. Use ssh-keygen -t ed25519 to generate a keypair.
  2. Go to ~/.ssh and copy the contents of the generated file that ends in .pub - this is your public key.
  3. ssh into the CSE server and append the contents of the .pub file to ~/.ssh/authorized_keys (if the file isn’t there, create it and chmod 600 it so only you can view and modify it)
  4. In your local machine, add this to your ~/.ssh/config, replacing the values as necessary:
Host unsw
  HostName login.cse.unsw.edu.au
  IdentityFile <INSERT PATH TO THE NON-.pub GENERATED KEY FILE HERE>
  User <INSERT ZID HERE>

Now, you can use ssh unsw instead of ssh [email protected] for a secure, passwordless SSH login experience!

To make the script use your newly created keys, simply set REMOTE_HOST to "unsw":

- REMOTE_HOST="<INSERT ZID HERE>@login.cse.unsw.edu.au"
+ REMOTE_HOST="unsw"

After that, the script won’t prompt you for your zPass a million times!

Here’s an example of me running the autotests for the first lab of 6991:

$ runsw 6991 autotest
sending incremental file list
./
.gitignore
              7 100%    0.00kB/s    0:00:00 (xfr#1, to-chk=4/6)
Cargo.lock
            153 100%  149.41kB/s    0:00:00 (xfr#2, to-chk=3/6)
Cargo.toml
            178 100%  173.83kB/s    0:00:00 (xfr#3, to-chk=2/6)
src/
src/main.rs
            435 100%  141.60kB/s    0:00:00 (xfr#4, to-chk=0/6)
Found cargo project: mini_grep
Located autotests for mini_grep
6991 cargo build --target-dir target # crate.tar
   Compiling mini_grep v0.1.0 (/tmp/tmpilav8beq/autotest)
    Finished `dev` profile [unoptimized + debuginfo] target(s) in 0.33s
Test 0 (./target/debug/mini_grep foo) - passed
Test 1 (./target/debug/mini_grep hi) - passed
Test 2 (./target/debug/mini_grep HI) - passed
Test 3 (./target/debug/mini_grep '😂') - passed
4 tests passed 0 tests failed
Connection to login.cse.unsw.edu.au closed.

Until next time!

Back to writing