I was able to reproduce both issues, thanks for reporting!
The proxy auth cookie secret is not persisted. It's stored in platform runtime data. I have migrated this into the database.
Redirection logic was incorrect when the token is invalid. This is also fixed.
As you probably have access to the IP of the server, you could simply display a link once the setup is complete in the CLI:
https://<IP>?<token>
So not even copy/paste needed for most.^^
@mehdi this was indeed a bug in the restore code path. the setupToken, as you guessed, should be sent in the top level object and not as part of the backupConfig . Fixed now.
@savity you have to reboot the server using your VPS control panel or "reboot" via SSH. Do not use the reboot button in Cloudron dashboard, it doesn't work. We will make a new release most likely end of this month (6.4) where it will get fixed.
I learnt a few things about the python world as part of fixing this:
There is now something called poetry. Projects now contain a poetry.toml file which can be converted into a requirements.txt using poetry export --format requirements.txt > requirements.txt .
There is something called an "editable install" - https://www.python.org/dev/peps/pep-0660/ . Idea being you can just run the code and the code just "installs" away packages as needed. See https://www.python.org/dev/peps/pep-0660/#id5 .
pyproject.toml is some new project settings file that possibly replaces setup.py
@nebulon Oh yeah, I had not thought to try the right-click, good catch. Right-clicking on a webpage never feels intuitive to me, except when I want the browser's context menu
@girish I don't know if it was installed before the upgrade, but it wasn't enabled, because unbound managed to run without problem
I'm not sure how it got there. It was supposed to be a clean Ubuntu 16.04. The only things I installed manually besides cloudron are basically ZSH and screen.
In any case, I marked this as solved, thanks !
@girish Yes, it does appear to pass the correct credentials, and the function in question seems to give no error. I'll try to debug further on the app side, but for now I think we can just file this as an unexplained weird thing