test(x509): guard against multi-curve signature verification fallback

[schenkty]

Problem

Issue #28: remove the weak multi-curve try-all certificate signature verifier. The curve is always known from the issuer SPKI.

Fixes #28.

Status

The weak helper was already removed on main in #30 (verify_ecdsa_declared_curve only). This PR adds a regression test so a try-all fallback cannot return.

Tests

test_no_multi_curve_verification_fallback — correct curve verifies; wrong-curve declaration and other-curve keys are rejected with no fallback.

---

Note

Low Risk
Adds tests only; no changes to certificate verification logic in application code.

Overview

Summary

Issue #28: ECDSA cert verification must use only the curve in the issuer SPKI, not a try-all-curves fallback. That weak path was removed on main (#30). This PR adds a regression test so that behavior cannot return unnoticed.

Related Issues

Fixes #28

Changes Made

• No change to runtime certificate verification; behavior already rejects wrong-curve SPKI and unrelated keys without multi-curve fallback
• New integration test covers secp256r1 and secp256k1 self-signed certs: correct SPKI verifies; mismatched curve declaration and genuine other-curve keys do not

Testing

From repo root, run the x509 builder tests (or test_no_multi_curve_verification_fallback specifically). Expect pass: own-curve verification succeeds; cross-curve and mis-declared SPKI cases fail verification.

Breaking Changes

None

^{Reviewed by Cursor Bugbot for commit 1b6751b. Configure here.}

[larseidsvoll]

Verdict: APPROVE

Test-only regression for #28 after #30 removed the multi-curve try-all verifier. Locks in known-curve-only verification:

• Happy path: each ECDSA cert verifies under its declared curve
• Wrong-curve OID on real key bytes must not verify (both directions) — the case the removed helper would accept
• Unrelated other-curve SPKI rejected (no fallback)

!matches!(..., Ok(true)) is fail-closed (Ok(false) or Err both pass).

Nit (non-blocking): overlaps somewhat with test_ecdsa_verification_requires_declared_curve above; still worth the explicit #28 guard.

Still draft. CI in progress.

Review only — do not bot-merge (Ty lock on node-rs). Humans merge. No @ humans.

[larseidsvoll]

Security look @ 06e9c0db (draft — #28 regression: no multi-curve ECDSA verify fallback).

No security issues found.

• Diff is test-only; no production change.
• On main, Certificate::verify_signature still calls verify_ecdsa_declared_curve (OID dispatch secp256r1/secp256k1 only). No multi-try / try-all-curves helper remains.
• New test locks the invariant: correct-curve OK; wrong-curve declaration (both directions) and unrelated other-curve SPKI must not verify as Ok(true).

No HIGH+. No human @.

[larseidsvoll]

Verdict: APPROVE (re-stamp after #45)

Tip moved for main merge; #44 undrafted. Diff unchanged: test-only #28 multi-curve fallback regression.

Security Audit green. Lint green; Tests still running.

Review only — do not bot-merge. Humans merge. No @ humans.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[larseidsvoll]

Thanks for the APPROVE. Nit acknowledged — some overlap with test_ecdsa_verification_requires_declared_curve is fine; keeping the explicit #28 guard as-is.