FIX: Add safety check for zero reference area (Fixes #2435)

[ayush4874]

Fixes #2435.

If REF_AREA is set to 0.0 (auto-calculation) but the geometry projection fails (e.g. flat plate aligned with projection plane), the area remains 0.0. This causes immediate divergence/NaNs later in the solver.

Added a check in SetPositive_ZArea to catch this case and error out cleanly via SU2_MPI::Error instead of crashing.

Verified using the NACA0012 testcase:

• Validated that a standard run with REF_AREA=0.0 still works correctly (calculates ~1.0), ensuring no regression for valid cases.

[pcarruscag]

Reference area is not necessary for all physics in SU2

[ayush4874]

Should I change this to a SU2_MPI::Warning, or should I move the check to where the coefficients are actually calculated?

[ayush4874]

@pcarruscag can you please address this?

[pcarruscag]

Error would be better, try to find a good place for it.

[ayush4874]

Done. I reverted the changes in CPhysicalGeometry and moved the check to CEulerSolver::SetReferenceValues .

This ensures we only flag the error when the solver actually needs the reference area, without blocking other physics.

[pcarruscag]

Still failing the tests and nota good place since that solver is not the only one that may use the reference area

[ayush4874]

I think, this might be right, added the check to CIncEulerSolver (Incompressible) also. This will cover all aero solvers without blocking other physics.

[pcarruscag]

sure just iterate on it until all the tests pass and let me know once you're done

[ayush4874]

@pcarruscag can you have a look now?

[pcarruscag]

Doesn't do what I asked before

[ayush4874]

@pcarruscag is it alright now?

[pcarruscag]

Error if something tries to do an invalid operation due to 0 ref area was the requirement

[ayush4874]

@pcarruscag can you have a look again?

[pcarruscag]

It still doesn't look like it does what we want.

[ayush4874]

I switched CFlowOutput to a soft return just to get the CI green, since the MMS regression tests were crashing without a REF_AREA.

Do you want me to keep this (maybe add a Warning?), or should I revert to the hard Error and just add a dummy REF_AREA to those failing test configs?

[ayush4874]

@pcarruscag I agree that Error is the right way, but it breaks the MMS regression tests. I tried to fix the test configs, but I found that TestCases is not a submodule in this branch, so I physically cannot commit changes to them in this PR.

To unblock the CI, I’ve switched CFlowOutput to a Warning for now.

Once this is merged, I will open a PR to su2code/TestCases to add the missing REF_AREA. After that is merged, we can switch this back to a hard Error.

[ayush4874]

@pcarruscag I tried switching this to a print/warning, but it caused the thermal_beam_3d regression test to fail. It seems that test script compares the output log against a golden file, so the extra warning text triggers a diff error.

Since I can't update the TestCases repo within this PR (it's not a submodule here), I'm stuck between Error (Crashes legacy MMS tests (missing REF_AREA)) and Warning (Fails regression output comparison).

I've reverted to the silent return for now just to get the CI green. Once this is merged, I can open a PR to su2code/TestCases to add the missing reference area, and then we can come back and enforce the hard Error.

[pcarruscag]

thanks.
to be honest this issue is just not a high enough priority to justify the amount of guidance that I think you need to make an acceptable solution.

[ayush4874]

Understood. Thanks for your time in reviewing this. I didn't realize initially how much the legacy test dependencies would complicate what should have been a simple check. I'll look for a more self-contained issue to tackle next.