toolchain: fix KeyError('num_fluids') in case-optimization fpp generation

[sbryngelson]

Summary

Case.__get_sim_fpp (toolchain/mfc/case.py) reads self.params["num_fluids"] by direct subscript when emitting the case-optimization Fypp header, while every other parameter in that same block uses self.params.get(key, default).

The packer constructs a dummy case with empty params to enumerate configured targets for the golden metadata file:

# toolchain/mfc/packer/pack.py
case = input.load(None, {}, {})
for target in get_configured_targets(case):
    ...

Under --case-optimization, that empty-param case flows through get_fpp → __get_sim_fpp, and the bare subscript raises KeyError: 'num_fluids'.

Impact

./mfc.sh test --case-optimization fails for every golden case with Error: 'num_fluids' — even though the case-optimized simulation itself builds and runs correctly. The failure is in the post-run golden-packing/metadata step, not the compiled binary (the run completes with exit 0 and correct output before the packer is invoked).

CI's case-optimization job runs ./mfc.sh run --case-optimization + a custom output check (.github/scripts/run_case_optimization.sh), not the golden harness, so it never hits this path — which is why it went unnoticed.

Fix

One line, matching the .get(..., default) pattern of every sibling in the same block (nb, weno_order, muscl_order, …):

-#:set num_fluids            = {int(self.params["num_fluids"])}
+#:set num_fluids            = {int(self.params.get("num_fluids", 1))}

For the dummy empty case the default is only used to compute a throwaway staging slug (which won't match any real build and is skipped); for real cases num_fluids is present, so behavior is unchanged.

Testing

• Reproduced: ./mfc.sh test --only <AMR golden> --case-optimization → KeyError: 'num_fluids' before the change; passes after.
• Verified on GPU (V100) with both OpenACC and OpenMP offload: two AMR multi-level goldens pass under --case-optimization (2/2 on each backend) after the fix.

[Copilot]

Pull request overview

Fixes a toolchain crash during case-optimization Fypp header generation when a “dummy” case is created with empty parameters (used by the golden packer to enumerate targets). This prevents KeyError('num_fluids') in Case.__get_sim_fpp while keeping behavior unchanged for real cases where num_fluids is set.

Changes:

• Replace direct self.params["num_fluids"] access with self.params.get("num_fluids", 1) to avoid KeyError for empty-param dummy cases.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 60.95%. Comparing base (8285ba0) to head (46d8b62).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@           Coverage Diff           @@
##           master    #1641   +/-   ##
=======================================
  Coverage   60.95%   60.95%           
=======================================
  Files          83       83           
  Lines       20003    20003           
  Branches     2983     2983           
=======================================
  Hits        12193    12193           
  Misses       5782     5782           
  Partials     2028     2028           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[sbryngelson]

apropos of little

The fix itself → closing #1641 does have a (small, real) consequence: it abandons a correct 1-line bug fix and leaves ./mfc.sh test --case-optimization broken (KeyError: 'num_fluids') on master.

Recall what #1641 actually fixes: case.py:394 reads self.params["num_fluids"] by bracket while every sibling uses .get(...). The packer builds an empty-params dummy case, so under --case-optimization that path KeyErrors — I hit it live earlier this session. It's a genuine latent bug, present on master right now.

The nuance:

• It's niche. CI's case-opt (and the normal workflow) use ./mfc.sh run --case-optimization + a custom validator, which never touches the golden-packer path. So CI doesn't exercise the bug — which is exactly why it went unnoticed.
• But it's a real dev-facing break. Anyone who runs the golden test harness with --case-optimization (a legitimate way to check case-opt against goldens) hits the KeyError immediately.