[sbryngelson]

Summary

Regenerates the goldens of the two IBM reacting-surface thermostat tests (F8D466BD Thermostat, C02C399E Thermostat -> Wall Temperature) that miss them on master.

What failed and why

#2200 added these tests with goldens made before #2205 (second-order face composition in the chemistry diffusion correction flux) landed, and #2200's CI ran on a base without #2205. Both PRs were green on their own bases, so the combination was never tested. On master the two tests miss their goldens by up to 7.7e-6 relative (abs 1.1 on the energy field, tolerance 1e-6), on CPU, NVIDIA and AMD lanes alike.

It is not #2213: on 86eeb34 (#2200 alone) the tests fail with identical numbers, and reverting only src/common/m_chemistry.fpp to its pre-#2205 state on that commit makes both pass against the old goldens.

The change is expected: the cases have multicomponent diffusion gradients, which #2205 changes. Largest change of any saved field, relative to that field's maximum: 2.6e-5 (Thermostat) and 3.8e-5 (Wall Temperature), in one species; the conserved-energy-sized fields move by 4e-6 and the rest by 7e-7 or less.

What changed

Only tests/F8D466BD and tests/C02C399E (golden.txt and golden-metadata.txt, as generated). No source or tolerance change.

Testing

• NVIDIA A100, NVHPC 25.11, OpenACC: both tests fail on master and on Mean gas temperature source: mean_temp_wall (thermally developed flow) and immersed-boundary support #2200 alone before the change, pass after regeneration.
• Chemistry label on the same A100 build with the regenerated goldens: 45 of 45 pass (no other chemistry golden is stale).
• AMD MI300A (amdflang AFAR, OpenMP offload): both regenerated tests pass.

This PR was made with Claude Code (AI).

[github-actions]

Claude Code Review

Head SHA: 767b357

Files changed:

• 0

Findings:
No issues found. The prepared diff and changed-files list were empty, so no files were checked in step 4.