Cross-Node Test Results: Core v30.2, Knots v29.3, Libre v30.2

[8144225309]

Cross-Node Test Results: Core v30.2, Knots v29.3, Libre v30.2

Full btcaaron core_test suite run against 3 Bitcoin node implementations on the same dedicated VPS (16GB RAM, 4-core Xeon, Debian). This is the second run — first was v29-era nodes. Upgrading to v30 surfaced one new divergence and resolved another.

Node	Version	Pass Rate
Core	v30.2.0	22/22
Knots	v29.3.knots20260210	19/22 (3 musig() failures)
Libre	libre-relay-v30.2-2	22/22

0 consensus divergences. 0 string-level divergences.

---

Regtest Results — 16/16 PASS (all 3 nodes)

Test	Cases	Core	Knots	Libre
tutorial_tx_policy_consensus	3	ALL PASS	ALL PASS	ALL PASS
task2_controlblock_mutation_matrix	5	ALL PASS	ALL PASS	ALL PASS
task3_sig_output_correctness_matrix	4	ALL PASS	ALL PASS	ALL PASS
task4_policy_consensus_split_matrix	4	ALL PASS	ALL PASS	ALL PASS

All three nodes fully synced (initialblockdownload: false), each ran its own isolated regtest instance. Zero differences in transaction acceptance/rejection, Taproot validation, control block verification, Schnorr signatures, fee policy enforcement, or consensus mining.

---

Mainnet Descriptor Results — musig() Divergence

Tests 5 & 6 (per node)

Case	Core	Knots	Libre
valid_plain_pk_timelock	PASS	PASS	PASS
invalid_nested_pk_musig	PASS	FAIL*	PASS
nonstandard_candidate_policy_deferred	PASS	PASS	PASS
control_musig_keypath	PASS	FAIL*	PASS
control_miniscript_plain_pk	PASS	PASS	PASS
target_pk_musig_nested	PASS	FAIL*	PASS

Test 7: Cross-Instance Consistency (all 3 simultaneously)

Case	Core	Knots	Libre	Consistent?
control/musig_keypath	OK	ERR_OTHER	OK	DIVERGENT
control/plain_pk_timelock	OK	OK	OK	YES
target/nested_pk_musig	ERR_INVALID_MUSIG	ERR_OTHER	ERR_INVALID_MUSIG	DIVERGENT

---

Divergence Analysis

BIP-390 musig() Descriptors — Genuine Gap

Knots' 3 failures are all musig() descriptor cases. This is not a parsing quirk — Core PR #31244 ("descriptors: MuSig2") merged into master on 2025-07-31, after v29.0 shipped (2025-04-15), and landed in v30.0. No v30-based Knots release exists yet. The musig() descriptor code simply isn't in Knots' source tree.

Verified empirically via getdescriptorinfo:

# tr(musig(key1,key2,key3)) — valid BIP-390 keypath descriptor
Core v30.2:  SUCCESS (descriptor with checksum, issolvable=true)
Knots v29.3: ERROR   ("key 'musig(...)' is not valid")
Libre v30.2: SUCCESS (descriptor with checksum, issolvable=true)

# pk(musig(key1,key2)) nested in script leaf — invalid musig usage
Core v30.2:  REJECTED ("Invalid musig() expression")    ← musig-aware
Knots v29.3: REJECTED ("key '...' is not valid")         ← generic
Libre v30.2: REJECTED ("Invalid musig() expression")    ← musig-aware

Impact: Tooling targeting Core v30.2+ or Libre v30.2+ can use musig() descriptors for watchonly wallets, address generation, and PSBT workflows. Same workflows will fail on Knots until a v30-based release ships.

---

What Changed: v29 vs v30

Node	Previous	Current
Core	v29.0.0	v30.2.0
Knots	v29.2.knots20251110	v29.3.knots20260210
Libre	libre-relay-v29.1-1	libre-relay-v30.2-2

1. NEW: BIP-390 musig() descriptors — Core and Libre now parse and validate musig() key expressions. Not present in v29. Knots still on v29 codebase.

2. NEW: musig-aware error classification — Core/Libre v30.2 return specific Invalid musig() expression errors for invalid musig usage, vs generic key-parsing failures. btcaaron's Task 7 can now distinguish "musig aware but rejecting" from "doesn't know what musig is."

3. RESOLVED: reject-reason string convergence — In v29, Core/Libre used mandatory-script-verify-flag-failed while Knots used mempool-script-verify-flag-failed. Core v30.2 adopted the mempool- prefix. All three now match.

4. UNCHANGED: regtest consensus — 16/16 PASS on all three, both runs. No regressions.

---

Untested Policy Boundaries

The suite probes consensus-level Taproot behavior. Known policy divergences in untested areas that would likely surface additional findings:

• OP_RETURN / datacarrier limits — Knots defaults to 42, Core uses 83
• Dust thresholds — Libre may accept lower dust amounts
• Bare multisig relay — Knots may reject what Core/Libre relay
• Full-RBF policy — Libre adopted full-RBF early
• Non-standard script relay — varying policies across implementations

---

Operator Notes

• btcrun + btcaaron installed cleanly with pip install -e .
• Minor btcrun config gap: no port= directive in generated config, so a second regtest silently fails if port 18444 is already bound
• All regtest tests ran in under 30 seconds per node
• Test date: 2026-03-08, Tester: pirq

[aaron-recompile]

Very nice analysis — this is extremely helpful.

The musig descriptor gap in Knots is a great catch.
Thanks for running the full suite and documenting the divergence.