Vagrant - Provisioning script not changing directory - shell

Question

New to vagrant, please help!
Vagrantfile
Vagrant.configure("2") do |config|
config.vm.box = "laravel/homestead"
config.vm.provision "shell", path: "vm-setup/provision.sh"
end
vm-setup/provision.sh
# Update apt-get
apt-get -y update
# Install tree
apt-get install tree
# Create .bash_aliases
sudo echo 'alias cls="clear"' >> ~/.bash_aliases
sudo chsh -s $(which zsh) vagrant
cd /vagrant
provision.sh file runs fine. When I run "vagrant provision" it updates apt-get, installs tree and even changes the shell to ZSH.
But sudo echo 'alias cls="clear"' >> ~/.bash_aliases and cd /vagrant lines do not work, not sure why. When I vagrant ssh into the machine, I am being taken to root directory (/home/vagrant). I would like to start in /vagrant folder.

Answer 1

Vagrant's shell provisioner by default runs with privileged = true:
privileged (boolean) - Specifies whether to execute the shell script
as a privileged user or not (sudo). By default this is "true".
When you perform vagrant ssh you login to a VM as vagrant user.
That's why:
1.
# Create .bash_aliases
sudo echo 'alias cls="clear"' >> ~/.bash_aliases
It writes to root's ~/.bash_aliases and it is really there:
root#vagrant:~# id
uid=0(root) gid=0(root) groups=0(root)
root#vagrant:~# cat .bash_aliases
alias cls="clear"
Solution: write to vagrant's home folder:
# Create .bash_aliases
echo 'alias cls="clear"' >> /home/vagrant/.bash_aliases
chown vagrant:vagrant /home/vagrant/.bash_aliases
2.
cd /vagrant
This means that folder was changed in provision script, nothing else.
Solution: add this statement to vagrant's .bash_aliases as well:
echo 'cd vagrant' >> /home/vagrant/.bash_aliases
Your final vm-setup/provision.sh is:
# Update apt-get
apt-get -y update
# Install tree
apt-get install tree
# Create .bash_aliases
echo 'alias cls="clear"' >> /home/vagrant/.bash_aliases
echo 'cd /vagrant' >> /home/vagrant/.bash_aliases
chown vagrant:vagrant /home/vagrant/.bash_aliases
chsh -s $(which zsh) vagrant

Answer 2

Even not being the case, just for the sake of completion:
I had struggled many times while trying to use Vagrant as testing tool for setup scripts and just now I realized the underlying reason:
Using this Vagrantfile statement:
config.vm.provision "shell", path: "myScript.sh"
myScript.sh is inlined to the virtual machine standard input. This is good in the sense you don't need access to the actual script from inside the virtual machine (typically through /vagrant path).
...but it comes with the drawback that any relative path won't work properly.
Of course: We can adjust it to absolute path based on /vagrant. But this requires to modify the script we are trying to test.
So in this case (and in my opinion in any case we are not going to disable /vagrant share), it is a better solution to use "inline:" option with the "machine-internal" path:
config.vm.provision "shell", inline: "/vagrant/myScript.sh"
...This will inline this statement instead of the contents of the file and relative paths (or even "script path based" ones such as $(dirname "${0}")/relative/path) are going to work properly.
Additionally, if the setup script you are going to test is intended to be executed by non privileged users (for example if it is going to set up some user configuration we will expect to work just after a vagrant ssh -with vagrant user-) it is also a good idea to add the privileged: false option, pointed out by #Nickolay:
config.vm.provision "shell", inline: "/vagrant/myScript.sh", privileged: false