curiosity_rover_ogma: Introduce Ogma demonstration (#159)

[ivanperez-keera]

Introduce demonstration of code generated automatically using Ogma to protect against a (fictitious) unwanted situation in the Curiosity rover demo.

Fixes #159.

[asimonov]

Nice addition — having a worked Ogma/Copilot example against an existing demo makes the tool much easier to approach than the standalone examples.

I built and ran this locally on aarch64, and it behaves as documented: build.sh completes, the generated copilotrv node comes up, and driving the rover past |x| = 2.5 starts the stream of Empty messages on /copilot/handlerKeepRoverInCheck, which stops again when the rover returns inside the bound. Good to have confirmation beyond the x86 CI run.

A few things I'd like to see addressed before merge:

1. Leftover debug command in the Dockerfile

Dockerfile, in the generation step:

RUN cd /tmp/space-ros && \
    ogma ros --project manifest/project.ogma && \
    find && \
    cd /tmp/space-ros/monitor/copilot/src/ && \
    runhaskell Copilot.hs

The bare find dumps the whole tree into the build log — looks like it was meant to be removed.

2. monitor/ should be in .gitignore

build.sh writes the entire generated ROS package into ./monitor via docker build --output type=local,dest=./monitor. After one build, git status shows:

?? curiosity_rover_ogma/monitor/

The .gitignore added here is generic Haskell boilerplate (dist, .stack-work/, *.hi, …), none of which matches anything in this directory — the one entry it actually needs is monitor/. A .dockerignore for the same path would also help, since once monitor/ exists it gets shipped as build context on every subsequent build.

3. Unpinned dependencies

RUN cabal install --lib copilot copilot-c99 ...
RUN git clone https://github.com/nasa/ogma
RUN cd ogma && cabal install ogma-cli:ogma

Neither the Copilot libraries nor the Ogma checkout are pinned. Because the clone sits in its own layer it's also cached indefinitely, so in practice this is both irreproducible and not tracking upstream. Since .github/workflows/docker-image.yml auto-discovers every */build.sh and runs it on every PR to this repo, an upstream change to Ogma or a Hackage release can break CI for unrelated PRs. A pinned tag/commit for Ogma plus a cabal.project.freeze (or --constraint flags) would make this stable.

---

Minor / optional:

• build.sh passes --build-arg VCS_REF= and --build-arg VERSION= to all three builds, but neither this Dockerfile nor Ogma's generated one declares those ARGs, so Docker warns about unconsumed build args. Looks like copy-paste from the other demos.
• The first docker build --target builder -t nasa/ogma:latest . appears redundant — the following docker build --output ... builds that stage anyway. It also claims the nasa/ogma name in the local image store, which could be confusing if that image is ever published.
• README: # Compilation and # Execution should probably be ## under the top-level heading; some fenced blocks carry $ prompts and others don't, which makes copy-paste inconsistent; and document.json is referenced without its manifest/ prefix.
• README says "we need at least 4 terminals", but Terminal 1 is docker compose up -d, which returns immediately — so it's really 3.
• Worth adding this demo to the list in the root README.md (that list is already missing a few others, so no strong feelings).

Things I checked that are fine: the RMW/network wiring works out (the generated image inherits RMW_IMPLEMENTATION=rmw_cyclonedds_cpp from osrf/space-ros:curiosity_demo, and both containers use host networking); the container name in the README matches what run.sh derives via tr '/' '_'; and /model/curiosity_mars_rover/odometry and /move_forward both exist in the rover demo as referenced.