This is primarily a cosmetic change.
- Increases code clarity because all system test blocks now start with `switch_to_system`
- Optimizes dependency ordering because `switch_to_system` has no
dependency on the server sshd
These scripts previously failed when called with syscalls like
`execve` (used by, e.g., Python's `subprocess.run`) that use no default
interpreter for scripts without a shebang.
Advantages:
- Pure test evaluations
- The test framework can now be used by flakes that extend nix-bitcoin
- Most features of `run-tests.sh` are now accessible via `nix build`/`nix run`.
We keep `run-tests.sh` for advanced features like `scenarioOverridesFile` and adhoc scenarios.
Other changes:
- `run-tests.sh` now builds aggregate VM tests like `basic` or
`buildable` by creating all VMs in a single evaluation.
This speeds up the tests and eases debugging by separating the eval and build steps.
- Use the new `nix` CLI which has improved build output logging
by prefixing output lines with the origin drv name.
Don't auto-switch to root when executing make-container.sh, because
auto root switching is also implemented in extra-container.
Besides simplifying the code, this is useful for a later commit that
introduces flakes-based container building.
With this change, the container is built under the regular user
instead of root, thereby utilizing the user's regular fetcher and
evaluation caches.
Avoid adding flake resource paths to the store (via string
interpolation).
This reduces performance and can lead to modules getting imported
twice, once through a local path and once through a store path.
This might not be needed in a future Nix release, in which case we can
revert this.
Previously, dhcpcd and bitcoind starting up in parallel could lead to
the following error in bitcoind:
```
bitcoind: libevent: getaddrinfo: address family for nodename not supported
bitcoind: Binding RPC on address 127.0.0.1 port 8332 failed.
bitcoind: Unable to bind any endpoint for
```
After the initial failure, the bitcoind service would always restart successfully.
This race condition, where both applications were simultaneously
manipulating network resources, was only triggered under specific
hardware conditions.
Fix it by running bitcoind after dhcp has started (by running after
`network-online.target`).
This bug and the fix only affect the default NixOS scripted
networking backend.
