Stabilize DVS polarity sign inference

[FlorianPfaff]

Bug

infer_polarity_contrast_sign(...) accumulated signed normal-flow evidence directly in an unscaled floating-point scalar.

For large but finite, cancelling flow batches, the partial sum can overflow to infinity even though the final mathematical score is finite. Once the accumulator reaches inf, later opposing evidence cannot recover it, so the inferred polarity-contrast sign can be wrong.

A concrete finite batch proportional to [0.75, 0.75, -0.75, -0.9] * finfo(float).max has a negative total score, while the previous sequential accumulation overflowed after the first two entries and returned +1.

Fix

• collect the informative signed polarity contributions;
• scale them by the largest absolute contribution before summation;
• use math.fsum for accurate cancellation;
• compare the scaled score against the correspondingly scaled zero tolerance;
• preserve the existing +1 fallback when all evidence is zero or numerically inconclusive.

Regression coverage

Add a focused extreme-value test under strict NumPy overflow/invalid handling. It verifies that the finite cancelling batch is inferred as -1 rather than being dominated by an overflowed positive partial sum.

Validation

• reproduced the previous accumulator result as inf for a batch whose scaled sum is -1/6;
• verified the scaled summation returns the negative sign;
• branch comparison: 2 commits ahead and 0 behind main;
• diff is limited to 15 additions / 4 deletions in the implementation and one 11-line regression test.

GitHub Actions is authoritative for the full test, lint, packaging, and integration matrix.

[github-actions]

✅MegaLinter analysis: Success

Descriptor	Linter	Files	Fixed	Errors	Warnings	Elapsed time
✅ COPYPASTE	jscpd	yes		no	no	31.73s
✅ JSON	prettier	7	0	0	0	1.09s
✅ JSON	v8r	7		0	0	4.07s
✅ MARKDOWN	markdownlint	68	0	0	0	1.94s
✅ MARKDOWN	markdown-table-formatter	68	0	0	0	0.6s
✅ PYTHON	black	1865	88	0	0	99.79s
✅ PYTHON	isort	1865	140	0	0	2.91s
✅ REPOSITORY	betterleaks	yes		no	no	2.88s
✅ REPOSITORY	checkov	yes		no	no	51.99s
✅ REPOSITORY	gitleaks	yes		no	no	17.77s
✅ REPOSITORY	git_diff	yes		no	no	0.42s
✅ REPOSITORY	secretlint	yes		no	no	83.76s
✅ REPOSITORY	syft	yes		no	no	5.47s
✅ REPOSITORY	trivy-sbom	yes		no	no	7.28s
✅ REPOSITORY	trufflehog	yes		no	no	33.18s
✅ YAML	prettier	11	0	0	0	0.8s
✅ YAML	v8r	11		0	0	12.39s
✅ YAML	yamllint	11		0	0	0.7s

Notices

📣 MegaLinter 9.5.0 is out! Discover the new features and security recommendations in the release announcement. (Skip this info by defining SECURITY_SUGGESTIONS: false)

See detailed reports in MegaLinter artifacts

Your project could benefit from a custom flavor, which would allow you to run only the linters you need, and thus improve runtime performances. (Skip this info by defining FLAVOR_SUGGESTIONS: false)

• Documentation: Custom Flavors
• Command: npx mega-linter-runner@9.6.0 --custom-flavor-setup --custom-flavor-linters PYTHON_BLACK,PYTHON_ISORT,COPYPASTE_JSCPD,JSON_V8R,JSON_PRETTIER,MARKDOWN_MARKDOWNLINT,MARKDOWN_MARKDOWN_TABLE_FORMATTER,REPOSITORY_CHECKOV,REPOSITORY_GIT_DIFF,REPOSITORY_GITLEAKS,REPOSITORY_BETTERLEAKS,REPOSITORY_SECRETLINT,REPOSITORY_SYFT,REPOSITORY_TRIVY_SBOM,REPOSITORY_TRUFFLEHOG,YAML_PRETTIER,YAML_YAMLLINT,YAML_V8R

Show us your support by starring ⭐ the repository