Run latest upstream master instead of pinning to a known-good version

[hacklschorsch]

This is the test grid and we like to tun whatever is most current :)

[btlogy]

I would not do this as it would breaks the reproducibility offered by Nix (I'm even surprised it is actually possible).

I would rather propose to handle this dependency as an input of the flake which is about to land in this repository (#33). And then automate the nix flake update to bump this input on regular basis.

But I might be wrong.

[hacklschorsch]

I would not do this as it would breaks the reproducibility offered by Nix (I'm even surprised it is actually possible).

Nix isn't reproducible.

I would rather propose to handle this dependency as an input of the flake which is about to land in this repository (#33). And then automate the nix flake update to bump this input on regular basis.

The system runs on auto-update and installs whatever gets into the 'stable' channel anyway - why shouldn't we do that with our own software?

[btlogy]

The system runs on auto-update and installs whatever gets into the 'stable' channel anyway - why shouldn't we do that with our own software?

l was naively thinking we were using NixOS to have some reproducibility when deploying environments.
If was wrong, forget it then.

Does this means the nixosConfiguration of testgrid should not end up build and deployed from the flake we're working on?

[hacklschorsch]

l was naively thinking we were using NixOS to have some reproducibility when deploying environments.

I'd like the test grid to always be running the latest Tahoe-LAFS master branch.

If flakes make that impossible, then I would please not have them.

Maybe we have a different understanding of what the test grid or this repo is for?

[hacklschorsch]

Let's talk about what you imagine this repository to look like when we're done vs. how I imagine it would look like.

[btlogy]

l was naively thinking we were using NixOS to have some reproducibility when deploying environments.

I'd like the test grid to always be running the latest Tahoe-LAFS master branch.

If flakes make that impossible, then I would please not have them.

Maybe we have a different understanding of what the test grid or this repo is for?

From the roadmap, the idea was:

Refactor the Nix code to integrate the nixosConfiguration (of testgrid) with the top-level flake, which will be used to manage and automated the deployment of all NixOS configurations

But we've also stated that this was a WiP. So let's not flake it then.

[btlogy]

If we do not want or can not have deployment reproducibility, then why not.