Default (no argument): run all three phases in order.
phase1: no-cameras, no-subprocess only.
phase2: with-cameras only.
phase3: with-subprocess only.
The three phases run mutually exclusive sets of tests (selected via different pytest marker filters) and must be invoked separately. For a full pre-PR run, every phase must pass before moving on to the next.
All commands run inside the already-running container via docker exec. Each clone of the repo on the host system has its own container, so at the start of a session resolve the one mounting this clone into ARENA_CONTAINER (empty = none running — use the dev-container skill first):
ARENA_CONTAINER=$(docker ps --filter "volume=$(git rev-parse --show-toplevel)" --format '{{.Names}}' | head -1)
Run the phase commands below in that same shell so $ARENA_CONTAINER stays set (or substitute the literal name).
Phase 1 — no cameras, no subprocess
The fastest of the three. Useful on its own when iterating on changes that don't touch rendering or subprocess paths.
docker exec "$ARENA_CONTAINER" su $(id -un) -c \
"cd /workspaces/isaaclab_arena && \
/isaac-sim/python.sh -m pytest -sv \
-m 'not with_cameras and not with_subprocess' isaaclab_arena/tests"
Phase 2 — with cameras
docker exec "$ARENA_CONTAINER" su $(id -un) -c \
"cd /workspaces/isaaclab_arena && \
/isaac-sim/python.sh -m pytest -sv \
-m 'with_cameras and not with_subprocess' isaaclab_arena/tests"
Phase 3 — with subprocess
docker exec "$ARENA_CONTAINER" su $(id -un) -c \
"cd /workspaces/isaaclab_arena && \
/isaac-sim/python.sh -m pytest -sv \
-m 'with_subprocess' isaaclab_arena/tests"
Run a single test file or function
# single file
docker exec "$ARENA_CONTAINER" su $(id -un) -c \
"cd /workspaces/isaaclab_arena && \
/isaac-sim/python.sh -m pytest isaaclab_arena/tests/test_asset_registry.py"
# single function
docker exec "$ARENA_CONTAINER" su $(id -un) -c \
"cd /workspaces/isaaclab_arena && \
/isaac-sim/python.sh -m pytest isaaclab_arena/tests/test_asset_registry.py::test_default_assets_registered"
On failure
- Report the failing phase and the first failing test (file, function, and one-line error).
- Stop. Do not run later phases until the failure is investigated.
- If the failure looks unrelated to recent changes, run the same test in isolation to confirm.
A run is "all green" only when all three phases pass with no failures or errors.