entropy-tss application data is stored in current working directory

[ameba23]

The directory where the key-value database is stored for entropy-tss is determined by the function entropy_kvdb::get_db_path:

entropy-core/crates/kvdb/src/lib.rs

Line 21 in 3df5d39

pub fn get_db_path(testing: bool) -> String {

It calls std::env::current_dir to get the current working directory, then adds .entropy/production/db or .entropy/testing/db.

Usually when i run entropy-tss from the command line this is fine - it creates a .entropy in whatever directory i am in, and i can easily clean it up afterwards.

However recently when setting it up to run continuously in a VM enclave using a systemd service, it took me a while to figure out where the db was being stored as the process was being started by systemd, not my personal user.

The files end up in the root directory (/.entropy). This is not a conventional place to put a hidden directory containing application data.

I don't know about other platforms but i think for linux the convention is to put writable system-wide application data in /var/lib/<application name> or /etc/<application name> and user application data should follow the XDG base directory specification (eg: with a crate such as microxdg).

My proposal would be that if the production feature flag is present, we should attempt to use either /etc/entropy or /var/lib/entropy, and if we get a permissions error fall back to either the current working directory, or XDG directory (~/.local/share/entropy or ~/.local/state/entropy).