A dotfiles test script can verify file diffs, not just existence

When using dotfiles across multiple machines, I kept running into "updated the repo but forgot to deploy" and "manually tweaked a config and now it's drifted." Adding a test.sh next to install.sh with three levels of verification solved this.

# File existence check
check_file() {
  [[ -f "$1" ]] && ok "$2: $1" || ng "$2 not found: $1"
}
 
# Permission check (verify 700/600 for SSH)
check_permission() {
  local actual=$(stat -c "%a" "$1" 2>/dev/null)
  [[ "$actual" == "$2" ]] && ok "$3: $2" || ng "$3: $actual (expected: $2)"
}
 
# Content diff check (repo vs deployed)
check_file_diff() {
  local src_content dest_content
  if [[ "$4" == "user_section" ]]; then
    src_content=$(sed '/^\[user\]/,/^$/d' "$1")
    dest_content=$(sed '/^\[user\]/,/^$/d' "$2")
  else
    src_content=$(cat "$1")
    dest_content=$(cat "$2")
  fi
  diff -q <(echo "$src_content") <(echo "$dest_content") &>/dev/null \
    && ok "$3 content matches" \
    || ng "$3 content differs"
}

The key trick is the exclusion pattern in check_file_diff. Git config's [user] section (name, email, signing key) differs per machine, so sed strips the entire section before comparing. Without this, every machine reports a diff and the test becomes useless.

Running ./test.sh gives color-coded pass/fail output, making forgotten deploys and unintended drift immediately visible.