Manage multiple F Prime deployments under a single repository

[kubiak-jpl]

Pattern

Many projects use more than one deployment within their source tree. For example, a flight deployment alongside multiple test deployments for different testing venues.

Some examples of questions that could be documented:

1. Are multiple deployments in the same git repository supported?
2. When multiple deployments are used, what configuration items are shared? What configurations can be independent per deployment?
3. What do root and deployment CMakeLists files look like for a multi-deployment setup?
4. Are there other known limitations to multi-deployment setups that should be made clear up front?

Rationale

Many projects use more than one deployment within their source tree. For example, a flight deployment alongside multiple test deployments for different testing venues.

A majority of the code may be identical within these configurations. For example, different configurations could setup devices with different io pins or allocated memory limits for camera images. Other past use cases involve simulation deployments, where parts of the original deployment are mocked or replaced with software simulations.

In either case it is desirable to co-locate these deployments, instead of placing them in separate sub-modules and build trees. However, issues like shared config/ directories tend to complicate and add nuance to these setups. Documenting how to setup multiple deployments, along with the limitations of this setup, would help when architecting how F Prime repositories should be structured.

[ljreder]

Good questions. For a testbed I am working on we need two deployments. One is for a hardware simulator written in F prime and the other is for the FSW written in F prime. This is a bit of problem if not supported. I always thought multiple deployments should be possible but have not done this yet? Well are multiple deployments within a single project doable?

[kubiak-jpl]

They are doable but there are some caveats that should be documented.

They work best when the two deployments target different platforms and are therefore built independently from one another. For example, a hardware deployment and a software-in-the-loop deployment for a PC. Otherwise you start to run into issues with the config/ directory being shared between deployments in the same build hierarchy

[SterlingPeet]

It seems like this should be doable for multiple flight deployments within the same mission. If it is not already available within the stock cmake tooling, I would imagine having a base configuration folder and some cmake magic to grab tweaked config values out of a deployment-overrides config folder. This did come up as an issue at GT, but I think that project ended using multiple git repositories as their workaround. It wasn't as desireable from a data product signoff and release standpoint.

[zimri-leisher]

We have had up to five deployments in our git repo, and the FPrime team can correct me if I'm wrong but I believe they make extensive (10+) use of deployments in their own project code.

I'm curious to know what issues you all have been having with a shared "config" directory. So far, we haven't encountered any issues that have been because of a shared FpConfig.h or similar. Most deployment-specific configuration we do is in the topology or instances files, with a small amount in the topology.cpp file.