[priyanshusky0]

Summary

• What: checkBalance() now exits with EXIT_CODE_INTERNAL_ERROR when the C or N balance residual is >= EPS (1e-8) in magnitude, or is NaN/inf. Before it only logged a warning and kept going.
• Motivation: See Make carbon and nitrogen balance failures fatal #411. Also noticed while doing this that a NaN residual was passing silently before, since both fabs(NaN) < EPS and fabs(NaN) > 0 are false.

Residuals below EPS are still zeroed and treated as noise, same as before. I didn't touch the balance equations or the clamping accounting.

On failure it now logs the element, residual, year/day/time, pre/post-update/final pool totals, inputs, outputs and the clamped amount. I used %.17g for these so you can actually see a ~1e-8 difference when the pool totals are large. If both C and N fail in the same step, both get reported before exiting.

How was this change tested?

• Added tests/sipnet/test_modeling/testBalanceCheck.c. It sets up the balance tracker directly and checks, for both C and N: zero, +/- 0.5*EPS and the value just below EPS pass; +/- EPS and +/- 1e-3 exit; NaN, +inf and -inf exit. There's also a balanced case with a nonzero clamping adjustment.
• To make sure the test actually catches things, I temporarily changed >= to > and removed the isfinite check in the source; the test failed in both cases.
• make unit passes.
• make smoke passes, with no changes to any sipnet.out. I also ran each smoke case with quiet off on master first, and none of them were hitting the balance warning, so nothing new should start failing. russell_4 has a skip file so I ran it by hand too, it exits 0.

Related issues

• Fixes Make carbon and nitrogen balance failures fatal #411

Checklist

• Related issues are listed above. PRs without an approved, related issue may not get reviewed.
• PR title has the issue number in it ("[#] <concise description of proposed change>")
• Tests added/updated for new features (if applicable)
• Documentation updated (if applicable)
• docs/CHANGELOG.md updated with noteworthy changes
• Code formatted with clang-format (run git clang-format if needed)

[Alomir]

Looks good, thanks for contributing! One minor change suggested.

[Alomir]

This is too much detail. How about just:

Suggested change

	- SIPNET now exits with `EXIT_CODE_INTERNAL_ERROR` when a timestep fails the carbon or nitrogen balance check (residual magnitude >= 1e-8, or a NaN/infinite residual), instead of logging a warning and continuing. The error reports the residual, timestep, pool totals, inputs, outputs, and clamping adjustment at full precision. (#411)
	- SIPNET now errors and exits when a mass balance check failure occurs, instead of logging a warning and continuing (#414)

Note also that we have been putting the PR numbers in CHANGELOG instead of the issue numbers

[Alomir]

Looks great, thanks