Roots Core v32.1 port review
Exact inputs and authentication
Input observation: 2026-10-31. The unpublished port branch topic/32.1/port starts directly at Bitcoin Core v32.1 commit f0de87f040621468b956f01f526a43a010310501. Old Roots history is source material for a new thematic series; it must not be merged into this base.
The official repository is https://github.com/bitcoin/bitcoin.git. Its advertised refs/tags/v32.1 object is 94f59fa45ea443882655625f4a647608274453e2, peeling to f0de87f040621468b956f01f526a43a010310501. Both match the local tag and the prepared branch's HEAD. Core v31.1 tag object bfa6a4b79cd4c1562fd32857e3147763efae37fb peels to the previous base 9be056a8a72b624dae9623b2f7bded92c2a21c91.
Upstream tag signature verification was explicitly waived by the user. It is not a passing signature check; no upstream keys were imported. Official repository identity, tag object, peeled commit and ancestry remain required. This waiver does not apply to new Roots commit signatures or later release artifact verification.
Publication is independently established by the official release announcement and GitHub release: release ID 426769835, draft=false, prerelease=false, published 2026-10-28T14:23:53Z. The announcement distinguishes v32.1 from v32.0, which was tagged but never fully announced or released.
Selected Roots source and ref refresh
Roots origin remains git@github.com:luganoplanb/bitcoin-roots. The host's SSH configuration permissions prevented the initial read. Read-only HTTPS access to the same repository succeeded without changing origin or host configuration. Fetching only main and roots/31.1 into their existing remote-tracking refs found no movement from the planning snapshot:
| Input | Planning and refreshed commit | Disposition |
|---|---|---|
origin/roots/31.1, selected canonical source | e93a8cca047c12ce9cab97181dd30705ed0e7e83 | Carry all released changes through roots.6. No newer canonical changes observed. |
origin/main | 5d09cb937f0024a6fae0b4ba3da7f0ae7fd0063f | Inventory separately; preserve useful website/tooling in main integration. |
| main's integrated Roots anchor | 34b672ffe2049050068409113783b44b3116b9dc | roots.2; later canonical fixes remain required. |
local roots/31.1 | 83ec87511a52074a1dc95d8c03c5ada4c71c61cc | Stale; excluded as a source selector, left untouched. |
All six remote Roots tag objects and peeled commits match local tags:
| Tag | Tag object | Peeled commit |
|---|---|---|
v31.1-roots.1 | 6cd32ddf7effb622802701b29c79d2cf33d5cdbb | 7cd11e298ec99e9bfc497f0fe3b9a18ab48a61a7 |
v31.1-roots.2 | d6bcaa0a53f984d32a37b1008d026b7ce41c6ac4 | 34b672ffe2049050068409113783b44b3116b9dc |
v31.1-roots.3 | aca0c15393a6e83781f1c883e3364e1899138b60 | 59185c317ef80ef289b46c38ba82081b61a1d101 |
v31.1-roots.4 | c80165c9def05f24ce5874327202b3045e414058 | 2fe385d7f51a2a7e40a778fc3a5c6e580499fbb0 |
v31.1-roots.5 | b08459ad72294b201e3b17365e7c5ebcfc143f4f | 93348f8530aadce4562aef40272b1e0af3e3478b |
v31.1-roots.6 | 0ce3cd96401d3fc593c881350f855957887a7f2c | e93a8cca047c12ce9cab97181dd30705ed0e7e83 |
Ancestry checks establish the append-only chain Core v31.1 → roots.1 → roots.2 → roots.3 → roots.4 → roots.5 → roots.6; roots.2 is also an ancestor of main. Core v31.1 is not an ancestor of v32.1: these release branches diverged. This is an expected upstream ancestry distinction, not missing Roots history; the new series uses the exact official v32.1 base. Source baseline worktree bitcoin-roots-release6 remains at the selected roots.6 commit. The port worktree was clean at input verification. Existing stale/prunable worktree metadata was observed but not modified. The authorized signing-rule edit in the separate main checkout's AGENTS.md is preserved.
New Roots commit signing
The shared repository-local configuration resolves in the port worktree to user.signingkey=5EADD53F2CD1F0B7AEEE920D25FC5C29CD528E32, commit.gpgsign=true, gpg.format=openpgp. Every new Roots commit, including merges, cherry-picks, patch imports and temporary replay commits, must use this key (Denis Roio, jaromil@dyne.org) and have its signature checked. Preserve original authorship and credits. No unsigned fallback is permitted. Configuration verification does not prove that a future signing operation will succeed.
Input verification scope
Focused checks compare official/local tag objects and peeled commits, all six published Roots tag identities, source ref movements, ancestry, branch/worktree identity and signing configuration. They establish source provenance only. No product behavior, native build, hosted CI or release artifact pass is claimed at this input boundary. Complete source/contract accounting and the maintenance baseline follow before this review unit is accepted.
Inventory interpretation and acceptance owners
The inventory covers 38 commits in 9be056a8a72b624dae9623b2f7bded92c2a21c91..e93a8cca047c12ce9cab97181dd30705ed0e7e83, 25 stable IDs in that tip's doc/roots-features.md, its additional address-export prose contract, and 148 net main-only paths in 34b672ffe2049050068409113783b44b3116b9dc..5d09cb937f0024a6fae0b4ba3da7f0ae7fd0063f. Ranges here identify trees or source provenance, not an executable replay plan. No implementation or behavioral acceptance is claimed by a planned disposition.
Owner labels refer to the v32 port work: L2 foundation; L3 local policy; L4 graph/replacement/dynamic dust; L5 mining; L6 networking/cache; L7 wallet backend/export; L8 Qt6; L9 runtime/configuration; L10 maintenance/release/docs; L11 whole-candidate audit/compatibility; L12 canonical/release handoff; L13 separate main integration. Split ownership means each owner must preserve the relevant part; it does not authorize another writer during this review unit. All owners inherit consensus-valid-block acceptance, native interface retention, full applicable suites, signed commits and the absence of release authority.
Fold/adapt means retain the final observable source behavior while using current v32 interfaces and incorporating its follow-up fixes. Retain native identifies inspected upstream behavior to keep; it is not a claim that new-base behavioral equivalence has already passed. Historical evidence, old platform gaps and the v31 reduced-verification waiver cannot satisfy v32 acceptance. No stable contract is dropped. Historical revision labels and generated help are superseded by candidate-specific identity and newly generated help in L10; historical notes and published assets stay intact.
Stable contract dispositions
Source IDs S01–S38 resolve to the complete commit table below. Tests named here are acceptance obligations, normally carried/adapted from the selected roots.6 source. A source filename is not a claim that the pristine v32 tree contains it. The exact source catalog and its linked grammar/subdust/effective-policy contracts remain the authority for defaults, units, errors and dependency edges.
| Stable ID | Planned disposition and source | Owner and observable acceptance |
|---|---|---|
policy-parsing | Fold/adapt S02, S11; whole-token/scoped grammar, ranges, overflow, quarter-vbyte rounding, redacted diagnostics. | L3.1; policy_args_tests.cpp, feature_policy_args.py, policy_args fuzz and independent roots_policy_oracle_tests.cpp; L4.2 instrumented arithmetic. |
effective-policy | Fold/adapt S07, S02–S04; report initialized applied settings, enforcement gates and units, not reparsed guesses. | L9.1 with L3–L5; rpc_effective_policy.py, defaults/overrides/disabled/precedence and actual admission parity; native v32 RPC schema tests. |
policy-standardness | Fold/adapt S02; retain local 1650-byte script/20-byte sigop limits, 2500 legacy sigops, bare-pubkey/multisig/CAT21/token/witness/ephemeral settings and exact reasons. | L3.1–L3.2; transaction_tests.cpp, mempool_sigoplimit.py, mempool_roots_local_policy.py; paired local rejection versus consensus-valid block cases. |
policy-carrier | Fold/adapt S02–S03, S11; final 83-byte/full-count behavior, 1 vbyte/data byte default, fractional rounding, input/witness recognition and extra policy weight. Preserve actual multi-OP_RETURN rejection; resolve misleading old help in L10. | L3/L4; mempool_datacarrier.py, mempool_subdust_fee_penalty.py, feature_taproot.py, mempool_tests.cpp; independent arithmetic and valid-block side. |
policy-dust | Fold/adapt S02–S03, S11; 3000 sat/kvB floor, dynamic off by default, initialized scheduler ownership, retry/shutdown safety, missing-estimate floor fallback and atomic snapshots. Old text describing broken pre-repair startup is historical. | L3 fixed thresholds; L4.2 dynamic lifecycle; mempool_dust.py, mempool_dynamic_dust.py, native publication tests, bounded dynamic_dust_fee sanitizer/fuzz. |
policy-subdust | Fold/adapt S03, S11; enabled deficit penalty affects modified fees only; preserve independent decimal/multiprecision oracles and signed/char boundaries. | L4 with L3/L5; mempool_subdust_fee_penalty.py, native policy oracles, carrier composition and block validity. |
policy-replacement | Fold/adapt S03; full-RBF fee/non-opt-in default, Never/OptIn and v3/TRUC semantics, native graph package invariants. Adapt deprecated RPC fields deliberately. | L4.1; mempool_rbf_policy.py, mempool_truc.py, package/RBF/cluster native and functional suites, fee-delta and ancestor cases. |
policy-coinage | Fold/adapt S02, S04; relay maturity/coin-block age default disabled, next-height/reorg semantics and overflow-safe UTXO arithmetic. | L3 admission/L5 mining; mempool_roots_local_policy.py, miner_tests.cpp, exact boundary and valid-block cases. |
policy-reuse | Fold/adapt S03; spkreuse=allow default and configured conflict tracking across mempool changes/replacement. | L4.1; mempool_rbf_policy.py, affected package/graph native and functional suites; no inference from a renamed upstream structure. |
mining-template | Fold/adapt S04; 1200000-weight/default priority-area, fee/size/resource controls, dependency ordering, residual native chunks and accounting. | L5.1–L5.2; miner_tests.cpp, mining_roots_template.py, mining_prioritisetransaction.py, full GBT/available IPC suites and valid-block tests. |
peer-controls | Fold/adapt S05, S17–S19, S28–S30, S33–S34; final count default 100, 10 decimal MB charged dynamic-storage cap, permissions and current transport privacy. | L6.1–L6.2; permission/addrrelay/getaddr/eviction/Bloom/duplicate-fetch tests, net_peer_connection_tests.cpp, compact extra unit/functional tests and producer benchmark smokes. |
peer-health | Fold/adapt S05, S19; current-connection facts/reset/selection refresh, unavailable-state clearing, no obsolete misbehavior_score. | L6 node API/L8.2 GUI; rpc_net.py, rpc_getblockfrompeer.py, apptests.cpp, sorting/persistent-index/per-peer state tests and rendered peer views. |
wallet-foundation | Fold/adapt S06, S17; descriptor/SQLite boundaries, write-access/directory/symlink guards, signer inherited-FD closure even below lowered soft limits, local unlock and anti-CAT21 locktimes. | L7.1; wallet_tests.cpp, wallet_multiwallet.py, wallet_signer.py, native DB/process tests and old-wallet fixtures where available. |
wallet-rbf | Fold/adapt S06; wallet-derived per-send preference and actual prepared transaction/PSBT signaling. v32 PSBT v2 default remains native. | L7.1 backend/L8.1 GUI; wallettests.cpp, wallet send/PSBT suites, both RBF states and versions 0/2. |
wallet-sweep | Fold/adapt S06; transient WIF keys never imported, local descriptor destination must satisfy signing, preview default, final exact-txid authorization, atomic cancellation/submission. | L7.1/L8.1; wallet_sweepprivkeys.py, wallettests.cpp, test_wallet_privacy_contract.py, incomplete/sufficient multisig/Taproot, locked/watch-only, stale-preview and failure cleanup. |
wallet-coin-control | Fold/adapt S06; wallet facts, effective value, grouped/reused/locked outputs and notification invalidation; locked rows visible but unselectable. | L7.1 interface/L8.1 GUI; wallet_send.py, wallettests.cpp, privacy contract fixtures, same-balance notifications and rendered layouts. |
runtime-config | Fold/adapt S07; forced→CLI→read/write config→JSON→network→default precedence and native scalar/negation rules; complete-read, EOF, rewind/comment, atomic write/reset semantics. | L9.1; getarg_tests.cpp, feature_config_args.py, full read/write config tests, partial-read/write/close failures and redaction. |
operator-stats | Retain native monotonic uptime; fold dormant support from S07/S17 without registration or sampling. Broader statistics remain deliberately unavailable. | L9.1–L9.2; rpc_uptime.py, native clock/source inspection; assert getmempoolstats unavailable. Do not enable dormant schema/model tests as a product feature. |
runtime-resources | Fold/adapt S07/S09; allocation/extension preserves bytes, safe FD saturation/closure and GiB conversion; retain native RAM/cache defaults including new indexes. | L9.2; full filesystem/streams/system/resource tests, signed/unsigned extremes and bounded sanitizer checks; platform branches inspected separately. |
build-portability | Fold/adapt S01, S13–S17; embedded defaults/system LevelDB+secp checks, filesystem/compiler bridges, QR/signer/IPC/taskbar/icon/MiniUPnPc conditions. Drop Qt5 and obsolete libevent dependency/header adaptation because v32 uses Qt6 and native HTTP. | L2.1–L2.2; clean installed-dependency headless/wallet/Qt6 profiles, test_leveldb.py, Homebrew fixtures where still applicable, native crypto probes and affected full suites. Actual unavailable platforms remain gaps. |
identity-desktop | Fold/adapt S08/S09 and later identity revisions; v32 Roots archive tag validation, icons, font-aware QR-PNG export and conditional Windows taskbar. | L2 wiring/L8 GUI/L10 identity and help; test_identity.py, apptests.cpp, wallettests.cpp, fresh-process persistence and stated viewport/font matrix. Protocol IDs/embedded attributions stay unchanged. |
maintenance-ci | Fold/adapt S08/S11/S17/S25 plus relevant main-only compatibility tooling; preserve cadence, full diagnostics, sanitizer/fuzz artifacts and exact-SHA normal CI. | L10.1; full python3 -m unittest discover -s ci/test, classifiers/failure transport fixtures and relevant lint; L11 whole candidate, L13 ordinary integration CI. |
maintenance-release | Fold/adapt S08/S17/S21/S35; Git-native immutable source/tag/new-draft/export/parity, notice bytes, five-platform packaging and signature trust. | L10.2/L12; full release fixtures, signed disposable replay excluding .github/**, blob/mode/symlink parity. Permanent tags, dispatch, manifest/platform signing and publication require later authorization. |
maintenance-docs | Fold/adapt source catalog/privacy/guidance/release notes, S10/S12/S20–S21/S23–S24/S27/S30–S32/S35–S38. Supersede release labels/generated help, preserve history. | L10.3; candidate-help regeneration twice, local links, metadata/docs fixtures and git diff --check; L11 final claims audit/L12 handoff/L13 website. |
maintenance-native | Fold/adapt S08/S09/S11/S17 and source fixture repairs; keep needed behavior-preserving modernization, safe empty writes and test-only assertion hooks. | L2–L9 owners of affected code; native/functional/fuzz regressions and registration; L11 complete audit, not a blanket common-maintenance overlay. |
Additional released address-export contract
The roots.6 catalog describes address export in prose rather than adding a 26th stable bullet ID. It is accounted here explicitly as address-export (supplemental), folded/adapted in L7.2 from S22, S25–S26 and S37, with L8 secret-history/UI integration and L10 documentation. Carry opt-in exportaddresskey, normal local-wallet authorization/unlock, destination type comparison, canonical keys array (full identities/scalars/compression/roles), P2PKH/P2WPKH aliases, recognized P2SH/P2WSH multisig/Miniscript wrappers, Taproot internal/leaf/effective-output mode and local MuSig2 participants. Partial ownership is accurately reported and never broadens export powers. Preserve disabled/private-key-free/external-signer errors, key unavailability, logs/history redaction and HD ancestor-xpub disclosure warnings; Quietproof's native-P2WPKH limitation remains documented. Required evidence includes the entire wallet_exportaddresskey.py, scriptpubkeyman_tests.cpp, wallet_rpc_tests.cpp, key_tests.cpp, rpc_tests.cpp and Qt RPC secret-history tests. A prerelease hardening change disables exportaddresskey, listdescriptors, gethdkeys and derivehdkey entirely by default, including public modes. Only an explicit --allow-dangerous-rpc startup command-line flag enables them; config files and runtime settings cannot. This is an intentional change to the former invocation-only export opt-in, without changing wallet unlock requirements. Current Core addhdkey, derivehdkey, exportwatchonlywallet and new key-provider/schema APIs are overlap to audit, not authorization to export more.
Dependency and non-feature boundaries
L2 precedes compiled port work. L3 admission precedes L4 graph/dust; L3–L4 precede L5 mining and L7 wallet policy integration. L6 uses L2–L3 and L4–L5 outputs where relay touches policy. L8 consumes L6/L7 node interfaces; L9 consumes L3–L7 authoritative settings. L10 reconciles all previous themes, L11 qualifies the complete tree, L12 pins its accepted final candidate, and L13 integrates that candidate separately. effective-policy depends on parsing and every reported policy; carrier extra-weight/subdust/replacement/mining depend on consistent graph weights/modified fees; dust snapshots depend on node lifetime/configuration; wallet GUI relies on backend authorization and notification semantics. None can be retired independently by a symbol rename.
Preserve the catalog's deliberate non-features: no RDTS/BIP110; no Network Watch/feeds/visualizers; no Tor pairing credentials, secret or endpoint-sharing UI; no CoinJoin/payjoin/automatic consolidation/privacy score/anonymity guarantee/chain-analysis dashboard/history redesign; no broad historical Knots consensus/script/wallet/documentation overlays; no retired replay/frozen-state/ generated-evidence/promotion control plane. Dormant statistics and the unused Tor-subprocess declaration remain dormant. Qt5 and old libevent dependency patches are obsolete support, not lost runtime features. The historical catalog excludes a VitePress overlay from the canonical product; the already existing main website is separately retained under L13, without bringing its font/visual dependencies into canonical node or Qt code. There is no unrelated wishlist expansion or fixed commit quota.
Current v32 overlap review
These observations inspect the exact pristine v32.1 files at f0de87f040621468b956f01f526a43a010310501, compared with the selected v31 source. Line locators refer to that base and will move during implementation. They justify adaptation boundaries; later owners still execute current tests.
| Area and base evidence | Consequence and acceptance owner |
|---|---|
doc/dependencies.md:13–42, src/qt/CMakeLists.txt:14,148: CMake 3.22, GCC 12.1/Clang 17, Boost 1.74 and Qt >=6.2; targets use Qt6::. No libevent dependency row. | L2 retains supported probes/platform fixes and Qt6; old Qt5 branches/libevent system-header patches are superseded, without disabling a runtime feature. Existing host dependency availability remains unverified here. |
src/txgraph.h:138–154, src/node/miner.cpp:164,253–268: native chunk/ancestor and policy-weight interfaces remain active. | L4/L5 must compose extra carrier weight, modified fees, replacement, priority-prefix ancestor removal and residual chunks with these APIs; a graph API is not policy equivalence. |
src/net_processing.cpp:502–545: InvToSendBucket owns global backlog plus count/size token buckets (12 MB/600 s, 50 MB cap). | L6 adapts Roots relay controls without restoring old per-peer scheduling; L4/L5 supply consistent effective policy. Full relay/transport tests required. |
src/net.cpp:1978–1985,4294, src/net.h:721: v1 reconnection carries the original node's proxy override through the reconnect queue. | L6 preserves upstream private-broadcast origin privacy on v2 failure, with the current transport regression profile. Old socket hooks cannot discard this field. |
src/net_processing.h:44: native DEFAULT_BLOCK_RECONSTRUCTION_EXTRA_TXN{100}. | Retain native final count default; S34/roots.5 explicitly supersedes the earlier 32768 proposal. L6 still ports byte accounting, fixtures, locks and benchmark behavior; same constant alone does not establish full cache equivalence. |
src/httpserver.cpp:45–79,708: native socket/select handling and HTTP thread pool replace the old libevent transport dependency; libevent mentions elsewhere explain analogous parsing behavior. | L9 adapts errors/limits/configuration to native HTTP; L2 drops obsolete dependency support, L10 updates current build/help guidance. Do not infer dependency use from explanatory comments. |
src/wallet/rpc/spend.cpp:1013,1053,1738: RPC defaults/constructor use PSBT version 2 and accept explicit version selection. | L7/L8 retain this default and verify actual transaction sequences and PSBT versions 0/2 for per-send RBF and sweep. Old PSBT fixtures must be adapted. |
src/wallet/scriptpubkeyman.h:128–130,212–217,267, src/wallet/rpc/wallet.cpp:838,914,957,1142–1151: native solving/signing/key-provider APIs and add/derive HD key/watch-only export RPCs. | L7.2 uses current authorization/provider boundaries; these RPCs do not replace address-specific authorized key export or justify additional disclosure. Check new RPC/OpenRPC schemas and all export error/ownership cases. |
src/rpc/mempool.cpp:346,441,520,627: explicit adjusted/BIP141 size fields and deprecated bip125 gates. | L4/L9 keep native fields truthful alongside Roots effective policy and replacement settings; verify deprecated/default RPC responses, not old unconditional fields. |
src/node/mining_args.cpp:30–99, src/node/miner.cpp:108–117: current BlockCreateOptions, flattening and reserved-weight/sigops/consensus-limit checks. | L5 adapts Roots limits to native option validation and current daemon/IPC consumers, preserving upstream rejection semantics and consensus validity; IPC unavailable is a gap. |
src/node/caches.cpp:23–91: 1 GiB default at >=4 GiB detected RAM on 64-bit, saturating MiB conversion, native txospender-index cache share. | L9 retains native RAM/default/index accounting; port only demonstrated missing resource safeguards and recheck index-specific boundaries. Historical cache text is not a new default instruction. |
src/common/system.cpp:122–125, src/rpc/server.cpp:184–198: startup and uptime use SteadyClock and GetUptime. | Retain native implementation of the narrow uptime sub-contract; L9 runs rpc_uptime.py. Dormant broader statistics are not adopted or enabled. |
src/streams.cpp:95–112: AutoFile::write validates the handle, then the non-obfuscated branch calls fwrite without an empty-span guard. | L9 preserves S09's safe-empty-write contract after null-handle validation and native position/obfuscation bookkeeping; absence of a guard is a source lead, with current regression/sanitizer evidence required. |
No Roots restriction is added to block consensus. Current Core consensus, serialization, transport privacy and native wallet operations remain the base; only necessary policy/resource/interface adaptations belong to their owners.
Complete canonical commit accounting
Every source commit is listed once in chronological order, including documentation, release preparation, tests and follow-up repairs. All 38 recorded authors are Jaromil; preserve that authorship and the older upstream/devjoinedthechat/Plan-B credits within the carried work. Owning contracts above specify acceptance suites; L11 audits this entire table and range-diff again at the final candidate.
| ID | Source commit | Source subject | Disposition / acceptance owner |
|---|---|---|---|
| S01 | 6a60180521daedbc404caf23a8c4a77bcb739de7 | build(roots): establish v31 portability foundation | L2: Fold/adapt foundation, system-library checks and actual platform support; preserve native crypto/embedded defaults. |
| S02 | e5445b7c96193c7ec141705aac5ba96651e0ee30 | feat(policy): preserve Roots parsing and local transaction checks | L3: Fold strict parser/local admission, consensus-valid block fixtures and configured defaults; do not copy obsolete validation signatures. |
| S03 | 846c9de12a5e71259dc1ed829413d68f449ebfd5 | feat(policy): preserve cluster accounting and repair dynamic dust | L4: Fold graph/modified-fee/replacement/reuse and repaired dynamic-dust scheduling, snapshots and safe scaling. |
| S04 | 0fbff64239358c69a266705b189770b0f5570ff3 | feat(mining): preserve Roots priority and template policies | L5: Fold priority prefix, ancestor/residual chunk accounting, template limits and native consensus validity. |
| S05 | 2cd5097c909d1ab2f8bf27eb0df97b24b6c0807e | feat(net): preserve Roots peer controls and health view on v31 | L6/L8: Fold peer controls and current-connection APIs; GUI peer selection/reset belongs to L8. |
| S06 | ab07fc5490af30ce7c6da1c137c8280d1f4815f0 | feat(wallet): preserve Roots signing and spend controls on v31 | L7/L8: Fold wallet DB/signer/sweep/coin-control/backend authorization and GUI RBF/preview/worker boundaries. |
| S07 | 44c15f2cde7d36f2733814d6ab68ce148f1a7079 | feat(rpc): preserve Roots operator interfaces on v31 | L9: Fold initialized effective-policy/configuration, resource guards and dormant statistics; retain native uptime. |
| S08 | 2f8bedaf1c56b3647746773492cb0d6c81a33e86 | chore(roots): integrate v31 identity and maintenance tooling | L8/L10: Fold identity/resources/QR/taskbar and Git-native maintenance/release tooling, attribution and fixture adaptation. |
| S09 | 8e7ffad5d932e07c0c80f3756f7c3a442786877e | fix(roots): restore audited portability and runtime behavior | L8/L9: Fold display-unit startup/persistence ordering and GUI tests; safe empty AutoFile writes/stream sequencing and benchmark fixture correction. |
| S10 | 0c0d2f1bf2e70f85e93423b61d1acd9b0b327578 | docs(roots): record v31 contract preservation evidence | L10/L11: Retain historical preservation evidence as history; regenerate current v32 claims after actual gates, never reuse old pass labels. |
| S11 | 9aaf62146fae1a89d65121468633d5c1701699e9 | fix(test): bound policy oracles and retain CI diagnostics | L3/L4/L10: Fold independent bounded policy oracles including signed-limit/char repair, safe tx-pool fuzz fixtures and full CI diagnostics/artifact transport. |
| S12 | 03bac6b5da7409d16abf1edeaf116b1ffc763e28 | docs(release): prepare v31.1-roots.1 release notes | L10/L12: Retain roots.1 historical release notes; supersede active candidate identity/help and prepare new release handoff without publication. |
| S13 | 64bb4d9137f0b4691ce0c3204dae24c5271931cf | fix(build): preserve Homebrew Boost system header classification | L2: Fold Homebrew Boost system-header classification where current imported targets require it; run maintained fixture. |
| S14 | d959ae99ffe7bef0593c2aca6e797fe702b06c41 | fix(build): retain implicit Homebrew system header flags | L2: Fold implicit Homebrew system-header preservation; check current compiler/imported-target behavior. |
| S15 | c07c02369c52d67365314dbf6c09f52bf34af8cb | fix(build): classify Homebrew IPC dependency headers as system | L2: Fold Homebrew CapnProto/libmultiprocess system-header classification where current IPC dependencies require it. |
| S16 | 83ec87511a52074a1dc95d8c03c5ada4c71c61cc | fix(build): classify Homebrew libevent headers as system | L2: Drop obsolete libevent-only header support because v32 has native HTTP/no libevent dependency; retain historical provenance. |
| S17 | 7cd11e298ec99e9bfc497f0fe3b9a18ab48a61a7 | fix(release): preserve notice bytes and integrate nightly fixes | L6/L7/L9/L10: Fold exact NOTICE bytes (.gitattributes), subprocess descriptor-type fix, dormant RPC nits, wallet fixture ownership and explicit setban test cleanup; retain follow-up regressions. |
| S18 | e5137d7fc9bc3c31848aeac021adbd3a9fbb764b | fix(net): restore explicit peer permission boundaries | L6: Fold explicit addr/all permission implications, additive grants, manual outbound bounds and addrrelay/getaddr/eviction regressions. |
| S19 | f41d71d7d23f8f17d6be125435627da437459269 | refactor(rpc): remove obsolete peer misbehavior score | L6/L8: Fold removal of obsolete misbehavior_score with honest peer state/duplicate block-fetch tests; never reintroduce misleading RPC field. |
| S20 | 662cc3ebfc1ce9de2dfbdf705f19a682ef731ecc | chore(release): prepare roots.2 P2P review candidate | L10: Retain roots.2 historical P2P release notes; supersede active identity/generated docs with v32 candidate. |
| S21 | 34b672ffe2049050068409113783b44b3116b9dc | docs(release): define Folding across Core release cycles | L10/L12/L13: Retain Folding/append-only/new-Core-base maintenance policy; adapt current branch guidance, no fixed series quota. |
| S22 | fad94e12f7b830e820c9f722a8545bffc809f2d5 | feat(wallet): add address-specific private key export | L7.2: Fold opt-in initial address-specific export with local wallet unlock/authorization; latest S37 generalization supersedes narrow implementation. |
| S23 | 3a74a4cd7859981e21ef074ad3cbdde9412c3158 | chore(release): prepare v31.1-roots.3 identity and notes | L10: Retain roots.3 historical notes; replace active candidate version metadata only. |
| S24 | 2b4c8982dd65e87947a085735bdf4455b63ac136 | docs(release): regenerate manuals for v31.1-roots.3 | L10: Supersede roots.3 generated manuals using fresh v32 candidate help, preserving historical source lineage. |
| S25 | b8c98a129f141ac5ba077c33b473120f0cd2ece4 | test(wallet): include address key export in fast regressions | L7/L10: Fold exportaddresskey registration in fast regressions/classification; coverage must include later generalized export. |
| S26 | 59185c317ef80ef289b46c38ba82081b61a1d101 | fix(wallet): compare exported key destinations by type | L7.2: Fold destination type-aware exported-key comparison; preserve regression rather than string-only address matching. |
| S27 | fe7cfe976edabc8c8d374b2bbe2bbaf2bef734c1 | docs: inventory mempool policy and cache review contracts | L3/L4/L6/L10: Historical v31 review retained byte-identically; doc/mempool-policy-review.md separates its measurements/defaults from current applied v32 contracts. See the L10 audit correction below. |
| S28 | bd5422a481aae345cf287bce1c8ab3dc296c20b7 | test: characterize compact reconstruction cache contracts | L6.2: Fold compact extra cache unit/functional characterization, test-only accessors and registered fixtures. |
| S29 | e9ddb283303167331fb6f5e76fedbf95ffa2b272 | bench: measure compact reconstruction cache tradeoffs | L6.2/L10: Native microbenchmarks complement the restored isolated exact-Core-versus-Roots multi-cell experiment, not replace it. Explicit v31.1/v32.1 adapters preserve raw samples, instrumentation and controls; no required performance gain. |
| S30 | 5fa358fbb7ce5e95be8f959888ae25dd70e3b1de | docs: clarify compact reconstruction cache accounting | L6.2/L10: Fold corrected charged-dynamic-storage/cache accounting explanations; historical 32768 default superseded by S34. |
| S31 | f6f9987732b9452b3b7883145de19ed9e8a840a7 | chore(release): prepare v31.1-roots.4 identity and notes | L10: Retain roots.4 historical notes; supersede active revision identity. |
| S32 | 7588cf44afc6bfbc06de67c0fb61f386ff16e13d | docs(release): regenerate manuals for v31.1-roots.4 | L10: Supersede roots.4 generated help through current native generators; no manual edits. |
| S33 | 2fe385d7f51a2a7e40a778fc3a5c6e580499fbb0 | test: declare lock preconditions for cache transaction helper | L6.2: Fold explicit lock preconditions for cache transaction fixture helper and its callers. |
| S34 | 79ebaa31225897f47403a95e510654f9531cdc45 | fix(net): restore Core default for compact block extra transactions | L6.2: Retain final Core count default 100, explicit-override/startup regressions and corrected public help; do not resurrect 32768. |
| S35 | 6faec825074b25c6ece812ef93d57b0cc34fc5cd | chore(legal): attribute Roots copyrights to Plan-B Foundation | L2/L8/L10: Fold Plan-B Foundation legal attribution across carried source/resources and generated metadata; preserve upstream/Jaromil/devjoinedthechat credits. |
| S36 | 93348f8530aadce4562aef40272b1e0af3e3478b | chore(release): prepare v31.1-roots.5 draft | L10/L12: Retain roots.5 historical release notes and latest defaults; supersede active candidate metadata. |
| S37 | add7e6aa8ffbdc39be3070b126662174834d1410 | feat(wallet): generalize address-key export on Roots 31.1 | L7.2/L8: Fold full generalized address export, descriptor/key helpers, provider/partial ownership rules, Taproot/MuSig2 support and all backend/RPC/Qt secret-history regressions. |
| S38 | e93a8cca047c12ce9cab97181dd30705ed0e7e83 | chore(release): prepare v31.1-roots.6 candidate | L10/L12: Retain roots.6 historical notes and provenance; supersede active version/generated help and reconcile full final wallet privacy/export documentation. |
Earlier fixes remain attached to their owning behavior, including L9's native stream edge cases and L8's startup-order persistence repair. The source's old L7–L16 numbering is historical; the v32 owner labels above control this port. S22/S26/S37 are cumulative, not alternative export implementations; S28–S34 are cumulative cache characterization plus final default correction.
Main-only differences and separate integration
The exact roots.2-to-main tree difference contains 148 paths: 111 website paths, 16 mode-only changes and 21 other paths. There are no main-only production C++ or consensus changes in this net difference. The main graph retains much older Roots/Knots history; a reachability log of roots.2..main would include that history, so this inventory uses the exact net path/blob/mode/symlink difference and path provenance. Old branch reachability is not additional product scope.
Each path below has a disposition class and the full source commit reported by git log -1 <pinned-main> -- <path>. This locator identifies the last affecting source change; the full pinned tree and its difference remain authoritative, including earlier fixes. Classes supply owning milestones, behavior and tests:
| Class | Disposition, acceptance owner and tests |
|---|---|
| W | Retain/adapt in L13 website integration only. Keep VitePress config/theme/design/content, documentation mirror symlinks, release catalogue/download links, major-grouped Folding patch browser, cache/renderers, static-HTML checks and publication workflow; preserve historical URLs/assets, qualify new 32.x navigation without fabricating release assets. Run npm test (docs prepare + all Node tests), available site build/static checks and ordinary CI. Browser checks and dependency availability are separate evidence; no installation or deployment is authorized. |
| M | Retain executable modes in L13, reconciling source registration with owning canonical L2/L3/L4/L5/L6/L7/L9/L10. These 16 paths have byte-identical blobs to roots.2 and only change 100644→100755. Do not invent a main-only behavior fix or overwrite later canonical file content. Verify blob/mode parity and relevant full maintenance/functional profiles. |
| D | Retain/adapt clangd compilation-database guidance, config and generated-metadata ignore rules in L13; reconcile current CMake paths and native flags with L2. Run docs/local-link/lint checks and inspect metadata exclusion; generated caches stay outside Git. |
| L | Retain/adapt symlink-aware executable/file handling and website-mirror lint exclusions in L13, without exempting product/build/tests/workflows from ordinary CI. Run relevant lint and website/maintenance fixtures. |
| C | Preserve compatibility harness, CI wrapper/workflow/source-tool regression calls and maintainer guidance. L11.2 imports/adapts and actually qualifies v32 fixtures on both exact binaries; L10 owns applicable canonical CI and L13 retains main tooling. Existing generation guards intentionally qualify only 30.3 (ci/compatibility.py:75,125,142), not 31.1 or 32.1. Keep binary/source/serializer/block provenance, process termination/cleanup, disabled automatic maintenance, bidirectional valid/invalid cases, normalized UTXO checks, lifecycle evidence and artifacts. Run both compatibility source-tool suites and full maintenance; actual matching-binary qualification is a separate required L11 gate. |
| G | Retain/adapt contributor/readme/maintainer guidance in L10/L13, superseding obsolete active release/branch labels, retaining Folding, separate product/tooling parity, ordinary integration CI and immutable old releases. Merge the authorized current AGENTS signing rule into new guidance in L2; never overwrite it from pinned old source. Run documentation/privacy/release fixtures and local links. |
The main-only website privacy guide (5d09cb937f0024a6fae0b4ba3da7f0ae7fd0063f) and README revision (ebdc72c0cd) are guidance, not permission to weaken node privacy or imply a privacy guarantee. Existing main's compatibility follow-ups include provenance/shutdown before fixture restarts, failed-cleanup rejection, automatic-maintenance suppression, lifecycle evidence and UTXO artifacts; retaining the harness files without these checks would be an omission. The catalogue's historical canonical non-feature boundary and this explicit main-only website retention are intentionally distinct.
Exact main-only path accounting
Mode notation is Git tree mode: 100644 regular data, 100755 executable, 120000 symlink; 000000 denotes absence at roots.2. Symlink targets must be inspected and retained/rebased as links during L13, not dereferenced into product copies.
| Path | Modes roots.2→main | Source commit | Class |
|---|---|---|---|
.clangd | 000000→100644 | 789fe585b66eea7ea812f8129d3605ecb467380d | D |
.github/workflows/ci.yml | 100644→100644 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | C |
.github/workflows/deploy-docs.yml | 000000→100644 | 15ea51185b9d891dc0797df72891cc52ddd552ff | W |
.github/workflows/reusable-compatibility.yml | 000000→100644 | d3a7c61cc38195e44c25e2e445bd1dbf0b976656 | C |
.gitignore | 100644→100644 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | D |
AGENTS.md | 100644→100644 | 7e9fe109d2b3419918cb024c1f98282a86425947 | G |
CONTRIBUTING.md | 100644→100644 | db3606f0c0692d4f7286866af2af8c1ebf4ff997 | G |
README.md | 100644→100644 | ebdc72c0cd25603ddc89026faef9fc7bf5d6a538 | G |
ci/compatibility.py | 000000→100755 | 78c585b98aa1e67405b11b26c3f7a0251e28e651 | C |
ci/test/roots-regressions.py | 100644→100755 | 7e9fe109d2b3419918cb024c1f98282a86425947 | M |
ci/test/test_compatibility_ci.py | 000000→100644 | 78c585b98aa1e67405b11b26c3f7a0251e28e651 | C |
ci/test/test_homebrew_boost_headers.py | 100644→100755 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | M |
ci/test/test_homebrew_capnp_headers.py | 100644→100755 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | M |
ci/test/test_homebrew_libevent_headers.py | 100644→100755 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | M |
ci/test/test_roots_compatibility.py | 000000→100755 | 78c585b98aa1e67405b11b26c3f7a0251e28e651 | C |
ci/test/test_roots_regressions.py | 100644→100755 | 7e9fe109d2b3419918cb024c1f98282a86425947 | M |
ci/test/test_workflow_selection.py | 100644→100755 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | M |
contrib/roots/README.md | 100644→100644 | db3606f0c0692d4f7286866af2af8c1ebf4ff997 | G |
contrib/roots/compatibility/README.md | 000000→100644 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | C |
contrib/roots/compatibility/build.py | 000000→100644 | 486de130e3a0a2dd00a5c369862d21af74bbb45f | C |
contrib/roots/compatibility/compare.py | 000000→100755 | f4ad5b388077c657d7cdf827fbd4e4356910111f | C |
contrib/roots/compatibility/fixtures.py | 000000→100644 | 78c585b98aa1e67405b11b26c3f7a0251e28e651 | C |
doc/developer-notes.md | 100644→100644 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | D |
doc/website/.gitignore | 000000→100644 | 6050fc6833412a6e7085001bc395ae2d487c3ced | W |
doc/website/.impeccable/config.json | 000000→100644 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/.impeccable/design.json | 000000→100644 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/.impeccable/surfaces/index-md.md | 000000→100644 | 8e65013f6738516e618eb8766d6e0919b893c64a | W |
doc/website/.vitepress/config.ts | 000000→100644 | 5d09cb937f0024a6fae0b4ba3da7f0ae7fd0063f | W |
doc/website/.vitepress/tests/patch-overview.test.mjs | 000000→100644 | db3606f0c0692d4f7286866af2af8c1ebf4ff997 | W |
doc/website/.vitepress/tests/patch-renderer.test.mjs | 000000→100644 | 097549b6ebd124951f0f73e96c7581a122b2ce93 | W |
doc/website/.vitepress/tests/patch-series.test.mjs | 000000→100644 | 6050fc6833412a6e7085001bc395ae2d487c3ced | W |
doc/website/.vitepress/tests/patch-theme.test.mjs | 000000→100644 | 097549b6ebd124951f0f73e96c7581a122b2ce93 | W |
doc/website/.vitepress/tests/prepare-releases.test.mjs | 000000→100644 | ce2be1b6651f5fe9a38256558ddbc2c5c36ac404 | W |
doc/website/.vitepress/tests/release-assets.test.mjs | 000000→100644 | ce2be1b6651f5fe9a38256558ddbc2c5c36ac404 | W |
doc/website/.vitepress/tests/release-catalogue.test.mjs | 000000→100644 | ce2be1b6651f5fe9a38256558ddbc2c5c36ac404 | W |
doc/website/.vitepress/tests/site.test.mjs | 000000→100644 | 5d09cb937f0024a6fae0b4ba3da7f0ae7fd0063f | W |
doc/website/.vitepress/tests/website-publication.test.mjs | 000000→100644 | 15ea51185b9d891dc0797df72891cc52ddd552ff | W |
doc/website/.vitepress/theme/components/ComparisonTable.vue | 000000→100644 | 7a3bbfe62014184f07c0184c3966bd24eb55547f | W |
doc/website/.vitepress/theme/components/DocumentationIndex.vue | 000000→100644 | cdd88e6c80f23db014163b8e6dccdfc7788cc196 | W |
doc/website/.vitepress/theme/components/Home.vue | 000000→100644 | ca2fa6b28f8c0bc4fb7bedb90d2ac15f7949d732 | W |
doc/website/.vitepress/theme/custom.css | 000000→100644 | ca2fa6b28f8c0bc4fb7bedb90d2ac15f7949d732 | W |
doc/website/.vitepress/theme/index.ts | 000000→100644 | cdd88e6c80f23db014163b8e6dccdfc7788cc196 | W |
doc/website/.vitepress/theme/patches.css | 000000→100644 | 097549b6ebd124951f0f73e96c7581a122b2ce93 | W |
doc/website/DESIGN.md | 000000→100644 | cdd88e6c80f23db014163b8e6dccdfc7788cc196 | W |
doc/website/PRODUCT.md | 000000→100644 | cdd88e6c80f23db014163b8e6dccdfc7788cc196 | W |
doc/website/RELEASE-INPUTS.md | 000000→100644 | db3606f0c0692d4f7286866af2af8c1ebf4ff997 | W |
doc/website/content/CONTRIBUTING.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/INSTALL.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/README.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/SECURITY.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/ci | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/compare.md | 000000→100644 | 8dcc41519a7df406e30834001d72f7c9e16bf539 | W |
doc/website/content/contrib | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/depends | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/CMakeLists.txt | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/Doxyfile.in | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/JSON-RPC-interface.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/README.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/README_doxygen.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/README_windows.txt | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/REST-interface.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/assets-attribution.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/assumeutxo.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/benchmarking.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/bips.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/bitcoin-conf.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/bitcoin_logo_doxygen.png | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/build-freebsd.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/build-netbsd.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/build-openbsd.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/build-osx.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/build-unix.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/build-windows-msvc.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/build-windows.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/cjdns.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/dependencies.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/descriptors.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/design | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/developer-notes.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/dnsseed-policy.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/external-signer.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/files.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/fuzzing.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/guix.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/i2p.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/init.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/man | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/managing-wallets.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/multiprocess.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/multisig-tutorial.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/offline-signing-tutorial.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/p2p-bad-ports.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/policy | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/productivity.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/psbt.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/reduce-memory.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/reduce-traffic.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/release-notes | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/release-notes.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/release-process.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/tor.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/tracing.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/translation_process.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/translation_strings_policy.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/doc/zmq.md | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/documentation.md | 000000→100644 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/getting-started.md | 000000→100644 | 3853ba692b4912eaacc007f824936af6892edfdf | W |
doc/website/content/index.md | 000000→100644 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/principles.md | 000000→100644 | cdd88e6c80f23db014163b8e6dccdfc7788cc196 | W |
doc/website/content/privacy.md | 000000→100644 | 5d09cb937f0024a6fae0b4ba3da7f0ae7fd0063f | W |
doc/website/content/public/bitcoin-roots.lottie.json | 000000→100644 | 8e65013f6738516e618eb8766d6e0919b893c64a | W |
doc/website/content/public/favicon.svg | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/share | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/src | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/test | 000000→120000 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/content/wallet-privacy.md | 000000→100644 | 5d09cb937f0024a6fae0b4ba3da7f0ae7fd0063f | W |
doc/website/package-lock.json | 000000→100644 | 6050fc6833412a6e7085001bc395ae2d487c3ced | W |
doc/website/package.json | 000000→100644 | 15ea51185b9d891dc0797df72891cc52ddd552ff | W |
doc/website/scripts/check-patch-browser.mjs | 000000→100644 | 097549b6ebd124951f0f73e96c7581a122b2ce93 | W |
doc/website/scripts/check-patch-html.mjs | 000000→100644 | 097549b6ebd124951f0f73e96c7581a122b2ce93 | W |
doc/website/scripts/check-static-html.mjs | 000000→100644 | 5d09cb937f0024a6fae0b4ba3da7f0ae7fd0063f | W |
doc/website/scripts/generate-doc-catalog.mjs | 000000→100644 | d41d2e27df060e7e9ac04e7a4e027dce06e904e8 | W |
doc/website/scripts/patch-browser.js | 000000→100644 | db3606f0c0692d4f7286866af2af8c1ebf4ff997 | W |
doc/website/scripts/patch-cache.mjs | 000000→100644 | 097549b6ebd124951f0f73e96c7581a122b2ce93 | W |
doc/website/scripts/patch-overview.mjs | 000000→100644 | db3606f0c0692d4f7286866af2af8c1ebf4ff997 | W |
doc/website/scripts/patch-page.mjs | 000000→100644 | db3606f0c0692d4f7286866af2af8c1ebf4ff997 | W |
doc/website/scripts/patch-renderer.mjs | 000000→100644 | 6050fc6833412a6e7085001bc395ae2d487c3ced | W |
doc/website/scripts/patch-series.mjs | 000000→100644 | 6050fc6833412a6e7085001bc395ae2d487c3ced | W |
doc/website/scripts/patch-theme.js | 000000→100644 | 097549b6ebd124951f0f73e96c7581a122b2ce93 | W |
doc/website/scripts/prepare-releases.mjs | 000000→100644 | 15ea51185b9d891dc0797df72891cc52ddd552ff | W |
doc/website/scripts/release-assets.mjs | 000000→100644 | ce2be1b6651f5fe9a38256558ddbc2c5c36ac404 | W |
doc/website/scripts/release-catalogue.mjs | 000000→100644 | ce2be1b6651f5fe9a38256558ddbc2c5c36ac404 | W |
doc/website/scripts/render-patches.mjs | 000000→100644 | 097549b6ebd124951f0f73e96c7581a122b2ce93 | W |
doc/website/scripts/website-publication.mjs | 000000→100644 | 15ea51185b9d891dc0797df72891cc52ddd552ff | W |
test/cmake/test_identity.py | 100644→100755 | 7e9fe109d2b3419918cb024c1f98282a86425947 | M |
test/cmake/test_leveldb.py | 100644→100755 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | M |
test/functional/mempool_dynamic_dust.py | 100644→100755 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | M |
test/functional/mempool_rbf_policy.py | 100644→100755 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | M |
test/functional/mempool_subdust_fee_penalty.py | 100644→100755 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | M |
test/functional/mining_roots_template.py | 100644→100755 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | M |
test/functional/p2p_compactblocks_extratxs.py | 100644→100755 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | M |
test/functional/p2p_roots_controls.py | 100644→100755 | 7e9fe109d2b3419918cb024c1f98282a86425947 | M |
test/functional/rpc_effective_policy.py | 100644→100755 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | M |
test/functional/wallet_sweepprivkeys.py | 100644→100755 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | M |
test/lint/lint-files.py | 100755→100755 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | L |
test/lint/lint_ignore_dirs.py | 100644→100644 | 0cd0cc423a97133c5e9f6891b921fddb734042bc | L |
test/lint/test_runner/src/lint_docs.rs | 100644→100644 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | L |
test/lint/test_runner/src/util.rs | 100644→100644 | 611a1e6bcc9e10e43d36815fa4f46910ee83f592 | L |
Source-accounting verification
L3 local policy adaptation
The local policy implementation is adapted from S02 e5445b7c96193c7ec141705aac5ba96651e0ee30, with S11 9aaf62146fae1a89d65121468633d5c1701699e9 independent signed-limit/byte oracle corrections. S03 846c9de12a5e71259dc1ed829413d68f449ebfd5 supplies only the common-library placement of the script-size setting, the carrier extra-weight primitive and the pre-staging arithmetic bound at this stage. Jaromil's implementation/contract lineage and upstream Bitcoin Core attribution are retained; new carried test files retain the Plan-B Foundation copyright.
Core v32's accurately counted legacy-sigop algorithm remains authoritative and is parameterized by the configured local bound. Its native ValidateInputsStandardness result API and diagnostic regressions remain available. Admission additionally uses the Roots configurable standardness checks and source rejection reasons. The essential validation.cpp changes are limited to local admission: standardness, carrier accounting, relay coin age, strict sigop density and an arithmetic guard before narrowing adjusted weight into the native graph. Block validation, consensus script flags, serialization, coinbase maturity and the upstream consensus regression assertions are unchanged. No production src/consensus/ or src/script/ changes are needed.
The scoped policy parsing contract preserves whole-token grammar, effective selection, raw-negation checks, ranges, quarter-vbyte rounding and contextual diagnostics. Native Core parsing of inherited controls remains outside that scope. Data-carrier limits default to 83 aggregate bytes with full counting, and multiple OP_RETURN outputs retain their actual local rejection. Explicit fixture allowances isolate upstream consensus/resource assertions from these local restrictions.
Carrier extra-weight arithmetic has an isolated primitive and independent multiprecision oracle; it has no admission/graph consumer yet. Sub-dust modified fees, graph reservations, replacement/TRUC/reuse modes and dynamic-dust lifecycle remain L4; template limits/priority selection remain L5; read/write text config and effective-policy RPC integration remain L9. Supplied staged controls warn that enforcement is pending. Parser acceptance and dormant helpers do not establish these later behaviors. Current L3 execution evidence follows its focused and complete policy gates; historical source results below remain historical.
L3 local restriction and valid-block coverage
mempool_roots_local_policy.py::test_paired_boundaries checks exact admission and submission rejection reasons, absence from the mempool/template, an explicit allowing profile, then acceptance of the same transaction in a block under the rejecting profile. The cases use consensus-valid scripts and rescan MiniWallet after direct block construction. Zero-valued integer bounds remain bounds; they are not advertised as disabled settings.
| Local restriction | Paired coverage and profiles |
|---|---|
| CAT21 marker / bare pubkey / unknown witness outputs | mempool_roots_local_policy.py::run_test: default marker/pubkey rejection, explicit allowing switches, unknown-witness default allowance and configured rejection; same-tx valid blocks. |
| Bare multisig | test_paired_boundaries, bare multisig default/override: bare-multisig, default versus permitbaremultisig=1. |
| Carrier-only transactions | carrier-only default/override: bare-datacarrier, default versus permitbaredatacarrier=1; fixture stays above native 65-byte minimum. |
| Carrier enablement / aggregate byte limit | carrier disabled and carrier83/82-byte boundary: datacarrier, disabled/82 versus default 83. mempool_datacarrier.py also retains default 83/84 and zero/explicit-large limits. |
| Full input/witness carrier counting / nonstandard carriers | mempool_datacarrier.py::test_taproot_carrier: independent switches, exact full-count budget boundary and valid block; OPNet witness carrier additionally checks default txn-datacarrier-nonstandard, explicit allowance and default-profile block acceptance. |
| Multiple OP_RETURN outputs | run_test: exact multi-op-return and same-tx valid block; mempool_accept.py retains native acceptance assertions under explicit nonstandard allowance. |
| Rune / OLGA token filters | run_test and OLGA token filter: tokens-runes / tokens-olga, default token-filter off versus enabled. OLGA isolates the independent nonstandard-carrier/full-count gates. |
| Output / spent-output script bounds | output script bound / spent script bound: scriptpubkey-size / bad-txns-input-script-size, configured 33/23 versus default 1650. |
| ScriptSig bound / fixed 1650 ceiling | scriptSig/redeem serialized bound and fixed scriptSig ceiling: exact scriptsig-size; higher configured bound or explicit nonstandard allowance, plus valid blocks. The separate redeem-size check is dominated by the earlier enclosing scriptSig bound under ordinary admission. |
| Total witness size | default total witness bound: bad-witness-witness-size, default 1650 versus 2000 using valid Taproot NOPs. |
| Witness argument versus tapscript push size | P2WSH/Taproot witness argument80/81 cases reject only 81-byte arguments with the intended stack-item reasons and accept valid blocks; mempool_datacarrier.py accepts an executed 520-byte script push with explicit carrier allowances. feature_taproot.py retains its independent consensus push-limit assertions, including invalid oversized pushes. |
| P2WSH stack count / script size / Taproot annex | Dedicated paired cases retain bad-witness-stackitem-count, bad-witness-script-size and bad-witness-taproot-annex; explicit nonstandard allowance permits admission. Large legacy scripts use bounded data pushes rather than exceeding the consensus 201-opcode ceiling. |
| Legacy sigops / strict density / extreme adjusted weight | Configured legacy-limit and density cases retain bad-txns-input-sigops-toomany-overall / bad-txns-too-many-sigops; default and strict-density 0 allowance are separate. mempool_sigoplimit.py retains 2500/2505, package and extreme too-large-cluster cases with valid-block counterparts. |
| Dust / ephemeral send, nonzero dust and anchor permissions | run_test plus anchor permission, default non-anchor dust permission, multiple dust cap, bare anchor permission: intended dust-nonanchor, dust-nonzero, anchor, dust and bare-anchor reasons; explicit send/dust/anchor permissions or zero dust rate/minimum fee isolate each gate. |
| Relay fee / coin-block age / maturity | default relay fee floor, coin-block gate independent of maturity, maturity gate independent of coin blocks: min relay fee not met, bad-txns-input-immature-coinblocks, bad-txns-input-immature-depth; default/zero/extreme profiles and same-tx valid blocks. run_test also retains exact age boundaries and tip reorg/reconsider behavior. |
Malformed redeem scripts, absent witness/control data and other consensus-invalid fixtures keep their native rejection assertions; they cannot supply a valid-block acceptance counterpart. The full changed-fixture selection includes all affected P2P, package, mining and wallet fixtures. Graph fee accounting, dynamic dust and replacement modes remain L4; independent template selection remains L5. This map does not imply enforcement of staged controls.
L3 verification on 2026-10-31 used the current Linux headless wallet/tests/IPC profile. The full native run passed 841 ordinary cases plus one external-assets warning; CTest passed 380 with one assets skip and no failures. All 38 registered instances from the 37 changed functional files have current passing evidence: 36 passed the complete run, and the fee-estimation/Taproot-wallet fixtures passed after explicit local allowances isolated their native padding/fee assumptions. Their assertions were retained. The extended local-policy test exercised 25 additional rejection/allowance/valid-block pairs. Documentation/include/whitespace and changed-file syntax checks passed. Missing external assets and Python lief/capnp/zmq, current Qt/platform profiles and instrumented fuzz remain explicit gaps for the applicable later owners; registration and historical baselines do not establish those passes. These are execution observations, not a parent milestone acceptance or release qualification.
The source baseline maintenance command python3 -m unittest discover -s ci/test ran from the clean roots.6 baseline at e93a8cca047c12ce9cab97181dd30705ed0e7e83: 153 passed, zero failed, zero skipped, exit 0. This includes the maintained privacy/docs/release/tool fixtures, not native v32 execution. No old sanitizer, Qt or platform result is inherited as a v32 pass.
At the unchanged v32 production base, python3 test/lint/check-doc.py exited 0: 209 used arguments, 230 documented, zero undocumented. Its 21 documented-but-not- matched arguments are informational output, not failed or skipped tests. The inventory audit compares all 38 source commit identities/order, all 25 stable IDs, supplemental export coverage and all 148 main-only path identities, last-affecting commits, blob/mode distinctions and symlinks against Git. All 39 test basenames named in the contract matrix resolve in the selected roots.6 source. Local-link review finds no relative Markdown links in this review document; source paths are intentionally revision-qualified code text. Whitespace, conflict markers, unstaged/untracked scope and empty index are checked separately because ordinary git diff --check excludes this new file.
This completes input/source accounting only. Each implementation owner must replace planned dispositions with actual current-candidate evidence. L11 must reconcile every row, range-diff, intermediate buildability, final credits and whole-candidate compatibility; L12/L13 then pin final canonical/integration identities. No release, tag, publication, main promotion or product behavior change occurred during this inventory.
L4.1 native graph, replacement and fee accounting
Carrier cost, immutable admission dust deficits, configured replacement/TRUC and full-script reuse claims now feed native v32 graph admission. The carried behavior comes from S03 (846c9de12a5e71259dc1ed829413d68f449ebfd5), with released S11 independent signed/decimal oracle handling retained. Jaromil, devjoinedthechat and upstream Core/Knots attribution is preserved. Native ByRatio comparisons, cluster/diagram algorithms, block-removal return values and deprecated RPC gates are retained; no ancestor compatibility index is added.
Entries and staged additions carry the same final modified fee and adjusted weight into replacement diagrams, fee changes, ancestor/descendant summaries, trimming and removal/reorg dependency restoration. Actual fees, BIP141 metrics and block rewards remain independent. The global absolute-fee/weight budgets retain the released conservative main-plus-staged bound, including pending removals. Direct reservation tests cover abort, failed fee changes, committed replacement and return to zero, including intentionally rejected replacements whose final state alone would fit. See doc/feature-contracts/subdust.md.
Independent multiprecision regression reproduced a source discrepancy at signed extremes: repeated relative updates of an already saturated modified fee differed from re-admission using the aggregate stored operator delta. Resident updates now recompute the documented admission formula, with atomic budget failure. The wallet chain-limit query also checks its synthetic fee/weight reservation before staging; an infeasible INT64_MIN delta is rejected. Neither correction widens native storage or changes actual transaction fees or block consensus.
Native RPC size/chunk/fee descriptions include carrier cost and dust deficits. vsize_bip141 and transaction weight retain actual metrics; vsize_adjusted and graph fields report policy accounting. The deprecated fullrbf field remains opt-in and reports the applied input-conflict Always mode, while deprecated bip125-replaceable retains signaling meaning. Runtime configuration and the full effective-policy schema remain L9-owned.
L4.1 focused verification executed all 52 selected amount, mempool, RBF, package, txgraph, cluster-linearization, policy-parser and oracle cases, plus the complete 29-script affected graph/package profile on rebuilt binaries. Pure sigop sizing explicitly disables carrier cost as in the released source; two dust/anchor allowing profiles disable the separately tested deficit. Exact local rejection, submission, template exclusion and same-transaction valid block assertions remain. New subdust cases check independent fee/weight arithmetic, package ancestry, replacement, atomic RPC failure and actual block rewards.
The released code gates prevout conflicts by Never/OptIn but inserts independent script-reuse conflicts later, subject to replacement economics. The older catalog statement that Never disables all replacement was broader than this runtime path. This port preserves the observed implementation and makes its scope explicit in current help/docs. mempool_graph_policy.py characterizes both Never and OptIn with independent inputs and non-signaling v2 originals: one satoshi below the exact incremental relay requirement rejects while equality replaces, with exact mempool and actual/modified fee outcomes. The original default reuse, direct-prevout and TRUC mode assertions remain.
Dynamic-dust publication, recomputation, locking and persisted operator-delta serialization are covered by the following L4.2 implementation. At the L4.1 boundary, newly carried arithmetic fuzz sources were registration, not an instrumented pass. The subdust script retains the optional --test-mining-priority assertions as an explicit L5 gate.
L4.2 dynamic dust and lifecycle
The complete released S03/S11 contracts now use native v32 estimator and scheduler ownership. One task is initialized after successful chainstate loading and estimator construction, before mempool reload; retries cannot capture an abandoned pool. Native scheduler stop/join precedes owner destruction. Block targets preserve the raw LONG-horizon 80% estimate, and mempool targets preserve actual fees at native policy-size positions. Safe thousandth scaling and the configured floor/missing-estimate fallback retain independent arithmetic oracles. Full snapshots, admission and node/wallet dust interfaces synchronize on the mempool lock; resident deficits are immutable.
Persistence now writes the stored operator delta rather than the modified-fee delta, preventing a dust deficit from being serialized as an operator change and subtracted twice on restart. Reorg admission captures one fresh deficit; native confirmation clearing of operator deltas remains intact. Startup setting changes, default/off, enabled/blocksonly, scheduler, missing/warmed estimates, reindex retry, restart and reorg are observable lifecycle fixtures. Independent P2SH, P2WSH, Taproot and mixed-input carrier fixtures calculate exact adjusted weight/fee boundaries while retaining BIP141 metrics and valid-block acceptance. Mining priority remains L5; integrated runtime configuration and effective-policy RPC schema remain L9. Current full-suite and instrumented execution results are recorded separately from the historical L4.1 boundary.
L5.1 template selection
Released S04 (0fbff64239358c69a266705b189770b0f5570ff3) supplies the Roots priority prefix, residual graph, resource controls and ancestor-accounting regressions. Jaromil, Plan-B and upstream attribution is retained. The material v32 adaptation keeps validated optional BlockCreateOptions, native coinbase fields and rejection diagnostics. Reserved weight below 2000, above consensus limits or above the maximum block weight remains an error. Serialized size retains the released 1000-byte reservation/clamping, priority defaults to 100000 bytes, block weight to 1200000 and minimum mining fee to 1000 sat/kvB.
Priority ordering uses confirmed coin age, then actual fee per policy vsize, then descending txid, and only admits transactions after their parents. A scoped native graph builder selects the remaining subset after the prefix. Removed ancestors contribute neither fees nor resources to residual chunks; policy fees and adjusted weight retain L4's signed budgets and native chunk algorithms. Actual BIP141 weight and sigops govern block limits, preserving the v32 sigop-adjusted-weight regression; actual fees alone determine coinbase reward. Local template exclusions never change block validation. The empty-prefix path uses the resident graph without copying it; residual storage is O(mempool) with native bounded cluster optimization and no live graph mutation.
Focused tests retain one/two-ancestor subtraction, waiting descendants, ties, modified-fee/rounding boundaries, actual rewards and valid blocks with excluded transactions. Resource cases exercise equality/one-above for serialized size, actual weight and sigops, low/high weight and sigop reservations, and native invalid-option diagnostics. IPC/schema qualification remains L5.2-owned.
L5.2 mining interfaces and qualification
GBT and mining clients retain the current v32 interface/schema. GBT calls the same mining implementation with resolved node defaults and cooldown disabled; its transaction fees/sigops and coinbase reward remain actual amounts. IPC's native BlockCreateOptions wire fields are useMempool, blockReservedWeight and coinbaseOutputMaxAdditionalSigops. Size, priority, fee floor and maximum weight inherit server settings; the schema does not offer per-request overrides for these fields. Reserved weight is always supplied by IPC clients (8000 when omitted), overriding the startup reservation as in upstream v32. Native option errors still apply. No wire field, ordinal or compatibility behavior is changed, so additional schema generation is unnecessary; the C++ profile uses native CMake-generated bindings against the current headers.
ipc_tests/mining_socketpair connects generated C++ mining proxies to the node's actual mining implementation over a socket pair. It covers server policy inheritance and resolved-policy retention across waitNext, size/weight equality boundaries, mempool omission, reserve/sigop errors, actual fee/reward vectors and coinbase reconstruction from v32 witness/required-output/locktime fields. Valid submissions through both methods, detailed invalid/non-final/duplicate-invalid results, duplicate results and explicit deprecated-method errors cross the wire. The native algorithms and block validation remain unchanged. Python's existing fee-only and 4-million-weight override fixtures explicitly select those policies instead of relying on the former upstream defaults; their oracles are retained.
The installed native Cap'n Proto profile and IPC CLI can run without Python bindings. Python capnp is unavailable in the installed Python 3.13 interpreter; interface_ipc.py and interface_ipc_mining.py must therefore report explicit skips, not passes. The C++ socket test does not establish the Python clients' full startup, cooldown, cancellation, concurrent-wait and malformed-payload profiles. L11 owns this remaining qualification gap; no dependency installation or disabled-IPC profile is used to claim it has passed.
Core v32 also consumes BlockAssembler for synthetic mempool fee estimates. The full native gate exposed priority-prefix transactions missing from its fee-chunk percentiles after Roots defaults became active. This consumer now explicitly selects fee-only assembly at consensus block capacity with v32's 1 sat/kvB synthetic floor. Percentiles and the health activity window use network capacity rather than the local mining default. Thus local template priority, fee and weight settings do not redefine network fee competition. Native percentile/cache/health oracles retain their expected fee fractions; their fixtures explicitly use network capacity. The affected functional estimator profile also explicitly retains native mining capacity, fee floor and fee-only ordering while exercising its existing low-feerate assertions.
L6.1 peer permissions, transport and current-connection state
The final released peer behavior is adapted from S05 (2cd5097c909d1ab2f8bf27eb0df97b24b6c0807e), the roots.2 permission repair (e5137d7fc9bc3c31848aeac021adbd3a9fbb764b) and obsolete-score removal (f41d71d7d23f8f17d6be125435627da437459269). This is a material v32 adaptation preserving Jaromil, Plan-B, devjoinedthechat and upstream lineage; it is not a byte-equivalent replay of those commits.
Bare whitelist/whitebind grants retain native bounded address handling; addr/all explicitly bypass address-processing and GETADDR caching, while message-size limits, inbound-only GETADDR and one response per connection remain. Socket and range grants add together. Outgoing address grants and default loopback Bloom apply only to manual connections; automatic full-relay, block-relay, address-fetch, feeler and private-broadcast connections receive neither. Explicit Bloom disable removes the loopback grant. Tor inbound binds, including the first ordinary bind used by the automatic onion service, suppress address grants while retaining explicit socket grants. The wildcard-target regression reproduced unintended all grants because accepted sockets report a concrete local address rather than the wildcard listener. Matching the target family/port for local proxy connections now preserves suppression; other ports and non-local peers retain normal grants.
forceinbound implies noban and download; ordinary noban does not imply force. Its fallback cannot evict no-ban or outbound/manual peers. The native v32 relay-only eviction filter and protected-peer argument remain intact for relay scheduling. forced_inbound records an actual displaced slot independently of the permission. Native v32 additionally enforces a transaction-relay cap after VERSION and Bloom transitions; that later event does not change the socket-admission flag. The force-slot fixture explicitly allows all inbound slots to relay transactions, while the native connection-limit/eviction tests retain their separate default-cap oracles. Clearnet v2-only suppresses explicit v1 and v1 retry; default transport behavior, private-broadcast proxy retention on retry, no-direct-connect protection, and SOCKS credential redaction retain their current v32 paths. Native saturating local-address scores already satisfy the released overflow repair and are unchanged.
Peer RPC exposes last block announcement, actual forced-inbound state and optional CPU load per mille of this connection's duration. CPU measurement failure/backwards readings and non-positive connection durations do not publish a load. New connections receive fresh scalars; internal discouragement remains native, and misbehavior_score remains absent. Duplicate manual fetch from the same peer preserves in-flight tracking; requesting a replacement peer transfers the request. Current v32 header/pruning safeguards remain. Fee filters retain their enabled default, explicit disable and released 100/1000 sat/kvB quantization seeds according to incremental relay fee. Cache accounting and the restored default of 100 transactions are separately owned by L6.2.
L6.1 focused Linux verification passed the complete Roots controls script and 30 related functional instances, including both transports for peer RPC, disconnect/ban and timeouts, native relay limits/rate scheduling, private broadcast/proxy fallback, Bloom/address permissions and duplicate/replacement block fetches. All 53 selected native networking/permission/eviction/DoS/peer manager/private-broadcast cases passed. Initial fixtures needed the v32 RPC connection helper and directional Bloom-service expectations; their complete affected scripts then passed on unchanged binaries. The wildcard regression's failed pre-fix capture and passing complete controls/unit cases establish that repair. This focused evidence does not replace the L6 full native/CTest and compact-cache gates, or later Qt/platform and external qualification.
L6.2 compact reconstruction cache and current resource behavior
The latest released cache behavior combines S05 with roots.4 characterization bd5422a481aae345cf287bce1c8ab3dc296c20b7, its lock-precondition correction 2fe385d7f51a2a7e40a778fc3a5c6e580499fbb0, and roots.5 default restoration 79ebaa31225897f47403a95e510654f9531cdc45. The default remains 100, matching the exact Core v32.1 header/help and observed 101-offer eviction. Explicit disabled, one-slot and 400-slot wraparound profiles also agree with that Core binary; the disabled profile offers five transactions. The old 32768 default is historical. Current byte/threshold, ownership, parser and benchmark contracts are in compact-reconstruction-cache.md.
The common ring insertion path charges native dynamic storage for each occupied reference, independently bounds count and metadata, evicts before adding, and applies the per-entry cutoff to replacements as well as rejects. Released unit tests retain exact charge and one-byte-below boundaries, sparse wraparound, multiple eviction, duplicate references, shared-owner lifetime and real reconstruction availability. The helper's lock exclusions include v32's inventory mutex before entering the native replacement producer.
The full P2P cache test retains real replacement/orphan producers and paired policy-rejected but consensus-valid blocks through hit, disabled and evicted paths. Core v32 ignores unsolicited compact blocks from peers not selected for high-bandwidth announcements. The adapted fixture submits the known header and explicitly requests each block before supplying its compact representation; native request/transport rules remain unchanged. This also makes competing candidate-block probes exercise reconstruction rather than an ignored message.
Current Linux gates passed 861 ordinary native cases plus one external-assets warning (27127425 assertions), CTest 381 passes plus one assets skip, and all 39 owned networking/transport/privacy/compact functional instances. Separate exact-Core count parity and all four producer/cache benchmark smokes passed. The benchmark profile retains wallet/tests/IPC; unavailable ZeroMQ remains the existing explicit qualification gap. Qt/platform/Python IPC and external-assets qualification retain their later owners; none is inferred from these results.
Four native benchmarks call current insertion/replacement methods, and three native reconstruction consumer cases also execute. Warm insertion timings exclude fixture construction and snapshots; the cold producer includes manager construction, download/broadcast bookkeeping, locks and destruction. JSON/CSV results use separate per-offer and per-operation denominators. Native dynamic charge is reported separately from the explicitly modeled count-times-slot-size payload. For the local default structural fixture, 100 occupied slots charged 41600 dynamic bytes, with 4800 modeled slot-payload bytes; this is not RSS or unique retained allocation. Historical exported/instrumented v31 experiment timings are not reused, and no improvement or throughput threshold is claimed. The final source-inventory pass also retains complete IBD-relay and pruned-node service profiles. Their fixture oracles distinguish released incremental-fee quantization from the minimum relay fee, and loopback-only Bloom services from global service advertisement.
L7.1 wallet foundations and transient sweep
The backend adapts released S06 (ab07fc5490af30ce7c6da1c137c8280d1f4815f0), with final roots.6 process/filesystem refinements and Plan-B attribution from S17 and S35. This is a material v32 adaptation of Jaromil's wallet work, retaining upstream and devjoinedthechat lineage rather than claiming a byte-equivalent replay. Descriptor/SQLite boundaries remain native. Core v32 already supplies the directory-write probe and SQLite rejection, including the native in-memory database distinction; those implementations are retained. Wallet discovery excludes node database trees and avoids recursive symlink/reparse traversal. Both multiwallet transports reproduced a lexical-path exclusion failure when v32 canonicalized the wallet directory but the data directory retained a symlink alias. Exclusions now compare canonical filesystem paths; the original relative wallet names and directory-error handling remain intact. Unix signer children close actual inherited descriptors, including those above a lowered soft limit, while retaining standard streams and the close-on-exec error pipe. Native Windows process handling is retained; handle isolation and non-Linux fallback execution remain separate platform qualifications.
Transient WIF sweep uses shared native UTXO scanning through the chain interface and requires a complete locally available signing path in the selected descriptor wallet. Cached public expansion and encrypted-key availability are sufficient for this eligibility check; it retrieves no wallet secret and does not unlock the wallet. Incomplete multisig/Taproot/MuSig paths remain ineligible. Supplied secrets stay in the transient signing provider, and the interface decodes secure strings with secure Base58 working storage. The default returns a signed preview. Atomic cancellation and submission are mutually exclusive, and the GUI-facing interface checks the reviewed txid before submission. GUI presentation and secret-history integration belong to L8.
The fee snapshot uses v32's native MinimumFeeRateResult and synchronized fallback authorization. Review found that the released sweep consumed a zero fee when estimation was unavailable and fallback disabled. This port rejects that case before submission, matching native wallet sends. It retains one fee snapshot during convergence, native signing options and native transaction version defaults. Backend RBF tests inspect actual prepared sequences and serialize/decode PSBT versions 0 and 2 for both wallet preferences and explicit per-send choices. Normal wallet key APIs remain native; generalized address export is separately owned by L7.2.
The complete sweep fixture retains CLI fee-precedence checks and the later released read/write-configuration branch behind --test-runtime-config. L9 must run that branch after its configuration implementation; its absence at this backend boundary is a named integration dependency, not a passing configuration result. Current focused execution and full-parent qualification are recorded separately from these source dispositions.
Focused Linux verification passed all 131 selected native cases across every wallet suite and the affected key/descriptor/script/RPC/process/filesystem suites (764450 assertions). The complete 78-instance wallet/tool/PSBT/UTXO-scan functional selection passed 75 and explicitly skipped three unavailable previous-release profiles. Both sweep RPC/CLI modes and both multiwallet modes passed. The initial three functional failures are retained: the two directory alias failures above, and a native MuSig script-path fixture funded at v32's 1 sat/vB synthetic estimate while Roots required its adjusted policy size. That nonce/participant fixture now specifies 2 sat/vB, retaining Roots policy and every signing/nonce assertion. All selected native and functional cases were renewed after these corrections. Existing Qt6 wallet-model, send and coin-control consumer objects also compile with the new backend interfaces; this is not a Qt presentation or runtime qualification. Full L7 reconciliation and the export-specific gates follow L7.2.
L7.2 generalized address-key export
The backend carries Jaromil's opt-in primitive S22 (fad94e12f7b830e820c9f722a8545bffc809f2d5), the roots.3 destination-type correction S26 (59185c317ef80ef289b46c38ba82081b61a1d101), and the complete roots.6 generalized contract S37 (add7e6aa8ffbdc39be3070b126662174834d1410). The selected comparison endpoint remains released roots.6 e93a8cca047c12ce9cab97181dd30705ed0e7e83. This is a thematic adaptation to Core v32 descriptor loading, RPCMethod, range accessors and separate MuSig2 functions, retaining the upstream, Jaromil and Plan-B credits. It does not use the separate user-owned export branch as an import source.
Opt-in means explicitly invoking exportaddresskey; the released RPC has no separate startup enable option. It requires a local descriptor wallet with private keys and the normal wallet unlock authorization. The wallet lock covers authorization, extraction and serialization. Public metadata alone does not grant export authority. Canonical keys records preserve complete public identities, secret scalars, compression and script/Taproot/MuSig2 roles; the historical single-key aliases remain available where applicable. Optional taproot_output returns an effective output secret only when locally available. Multisig and MuSig2 results describe local participants, without promising a complete spending path. Address and wrapper checks retain the final typed destination comparison and reject inconsistent metadata. Nonce snapshots, descriptor caches and wallet contents are unchanged by export.
Native v32 addhdkey, derivehdkey, exportwatchonlywallet, ExportDescriptors and ExportWatchOnlyWallet implementations are retained unchanged. Existing HD master secrets already held locally can be explicitly exported, including the documented ancestor-disclosure risk. Export does not search unknown derivation paths or add authority to private-key-free or external-signer wallets. The privacy manual states those limits and the explicit-invocation and unlock requirements. Qt console history integration remains L8's presentation responsibility; daemon log redaction is checked here.
The carried complete export fixture uses independent BIP32, BIP327 and BIP341 oracles and both RPC/CLI transports. Additional checks exercise native HD key derivation, locked authorization and public-only export/restore boundaries. OpenRPC tests assert the canonical record schema, optional legacy aliases and argument default in both transports. Source equality checks complement these behavioral tests; they do not substitute for them. Focused child evidence and later full-parent qualification remain distinct.
Focused Linux verification passed all 159 selected native cases (767537 assertions), including all 25 added provider/export cases. Both complete export transports and all 25 selected current wallet/RPC/schema profiles passed, with no selected skips. The native v32 key API, public-only wallet, descriptor, encryption, signer, MuSig2, Taproot, Miniscript, PSBT, raw RPC, help, OpenRPC and sweep profiles are included. Initial failures are retained: the carried nonce test required v32's public-nonce argument to its session identity, and the new functional fixture required refreshing CMake's configuration-time symlink glob before either transport could start. The repaired native build and complete export transports passed with current source/configuration/binary identities. Raw runtime wallets and private-key fixture values are excluded from evidence; full sanitized logs and the complete reconstructible dirty-diff digest are retained. Full L7 native/CTest and complete wallet/tool reconciliation await the parent continuation.
L7 parent qualification
The preceding child results describe their individual checkpoints. Parent qualification rebuilt all six policy consumers after the include cleanup and froze those binaries before execution. The complete default native run passed 901 cases normally and reported one assets warning, with 26961793 assertions. The assets case returned before processing fixtures because DIR_UNIT_TEST_DATA is unset; it is not a qualified assets pass. Complete CTest discovered 382 tests with none disabled: 381 passed and the same assets prerequisite produced one skip. All 28 released export additions have passing witnesses in the current full native report.
One combined functional selection covered 88 unique wallet, tool, PSBT, UTXO-scan, signer and affected key/RPC/schema instances: 85 passed and three previous-release profiles skipped. Both complete export transports, both sweep transports and OpenRPC RPC/CLI passed. A fresh prerequisite check found the effective previous-release directory absent; no option disabled that capability to obtain a pass. Ancient migration requires Core 14.3, migration requires 28.2, and backward compatibility needs its declared previous-release matrix. Those profiles and external assets remain explicit L11 qualification gaps.
Complete sanitized native, CTest and functional captures retain commands, UTC observations, exits, individual results and source/configuration/binary identities. Historical failures and child checkpoints remain preserved. The only source-tree change after the final product runs is this review paragraph; no executable, fixture or configuration input changed. Native v32 HD/public-only wallet APIs remain unchanged, and export remains an explicit invocation under normal local-wallet authorization. No separate export-enable switch exists. Qt console/history integration remains L8, the retained sweep runtime-config branch must run under L9, and optional/platform/sanitizer qualification remains with its declared later owners. These results do not qualify those gaps or establish a release or deployment.
L8.1 Qt6 wallet controls and asynchronous lifetimes
The Qt wallet changes adapt the released roots.6 source at e93a8cca047c12ce9cab97181dd30705ed0e7e83, especially Jaromil's ab07fc5490af30ce7c6da1c137c8280d1f4815f0 wallet theme and the console filter additions in fad94e12f7b830e820c9f722a8545bffc809f2d5 and add7e6aa8ffbdc39be3070b126662174834d1410. Bitcoin Core and Plan-B Foundation source attribution is retained. This is an adaptation to native v32 Qt6 and wallet interfaces, rather than a replay of Qt5 build internals.
Core v32 removed the send dialog's RBF checkbox. The port restores an explicit per-send choice initialized from the accepted wallet walletrbf preference, and preserves explicit RBF signaling for fee bumps. Send confirmation uses the prepared transaction's actual sequences and reports effective inputs, size, fee rate and change. Loaded PSBT0 and PSBT2 proposals reconstruct the unsigned transaction for signaling feedback; native sends retain v32's default PSBT2. Coin control uses wallet-provided effective value, input size and status facts, keeps filtering independent of selection, and invalidates selections on wallet notifications even when balances do not change. No privacy score is added.
The transient sweep dialog has masked input, guarded cache disposal, a non-broadcast preview and default-Cancel confirmation. It submits the exact preview transaction ID through the accepted wallet interface. Notifications also invalidate a preview during the confirmation's nested event loop. Worker cancellation, queued-key cleanup and joining on model destruction preserve the authorization boundary; a surviving dialog uses a guarded model pointer. After submission starts, cancellation cannot reverse broadcast. Qt and host buffers prevent any claim of complete memory erasure. RPC console/history filtering covers both sweepprivkeys and exportaddresskey, including nested export calls and wallet/null/Taproot output modes.
Focused Linux qualification uses the Qt6 RelWithDebInfo wallet/IPC profile and the minimal platform plugin. The complete Qt test executable covers both RBF states, PSBT0/2 display, stale send/sweep confirmations, same-balance input invalidation, Cancel as the sweep default, real preview/exact-txid submission, secret error cleanup and model/worker destruction. Four focused native wallet cases pass with 143 assertions; five functional instances pass for send and both RPC/CLI sweep/export transports. The first GUI fixture failed because it had no fee estimates and the native default disables fallback fees. It now opts into a fallback only for the disposable sweep test; production wallet defaults are unchanged. Compiler/API adaptation and filter-selection failures are retained with the failed test attempts.
These are behavior checks, not rendered qualification. The viewport/font matrix, fresh-process settings and peer refresh checks belong to L8.2. Windows/macOS rendering and optional platform profiles remain untested here; the sweep runtime-configuration branch remains an L9 gate. Complete parent native/CTest and owned functional reconciliation follow both Qt children.
L8.2 persistence, peer views and desktop rendering
The peer model and unavailable-state repair adapt Jaromil's released 2cd5097c909d1ab2f8bf27eb0df97b24b6c0807e; Roots resources, splash, font-aware QR export and Windows taskbar progress adapt 2f8bedaf1c56b3647746773492cb0d6c81a33e86. Existing upstream and Plan-B attribution is retained. Native Qt6 resource registration and enum streaming remain in use; Qt5 compatibility branches and generated translations are not imported.
Persistent peer indexes now resolve current rows after refresh replaces model storage. Empty refreshes do not emit invalid indexes. Selected-peer state is cleared before attempting to fill it, including session IDs when switching to v1. The live test uses two loopback regtest peers, changes sorting while keeping selection, forces the native unavailable-state try-lock path, then disconnects the selected peer. Table columns expand to fit the current font and facts without shrinking a wider saved column. Long identifiers and wide tables keep native scrolling; the resizable details pane is exercised at three quarters of the viewport width.
Starting Block uses the per-connection VERSION height retained by L6, matching the released GUI: nonnegative advertised heights (including zero) are shown, and negative or temporarily unavailable state is Unknown. The Qt regression sends real VERSION messages from disposable loopback sockets with sentinel 845123, zero and negative heights through PeerManager and the node interface. It checks selection, persistent indexes, sorting, current-row refresh, native try-lock unavailability, recovery, explicit reset and disconnect. The rendered matrix also captures the sentinel height. This restores the GUI fact without restoring the removed startingheight or misbehavior_score RPC fields.
The aggregate coin label describes safe local-key outputs, excluding locked, immature, unsafe, watch-only and external-signer outputs from that total. Watch-only inputs remain selectable for PSBT preparation. This does not promise that a partially owned construction can finish signing, or rate its privacy. QR tests cover the quiet zone, PNG/clipboard parity, the actual receive Save action and refresh when the configured money font changes. Taskbar source and its Windows-only CMake option use the Qt6 native-event API; this Linux build does not execute that Windows path.
The display-unit regression uses separate writer and reader processes for all four units, reading language before constructing the production OptionsModel. It verifies both the stored enum type and the restored model value. Qt6 already registers the enum's stream operators; the released Qt5-specific registration repair is unnecessary on this baseline.
Rendered qualification uses actual Qt 6.8.2 xcb widgets on owned Xvfb screens at 1280x800 and 1920x1080, with application fonts of 9 and 13.5 points (100% and 150%). Device pixel ratio is 1 and logical DPI is 96; Xvfb's reported physical DPI is recorded separately and is not hardware HiDPI qualification. The four-cell matrix covers coin control and its scrolled facts, sweep preview, confirmation and cancellation, send/RBF and PSBT confirmations, and populated peer table/details/unavailable state. Tests use disposable explicit regtest wallets and peers; no entered spending key, wallet database or cookie belongs in the screenshot evidence. Each matrix execution runs the full Qt suite, including all eight fresh-process settings operations. Initial import/build, watch-only-count and Xvfb-wrapper/readiness failures are retained. Missing xauth is handled by an owned local Xvfb server with TCP disabled, without an installation or host configuration change.
The L6/L8 interface correction restores the per-connection VERSION starting height, and the GUI now displays nonnegative advertised values, including zero. Real node-state regression and renewed rendered evidence cover sentinel 845123, selection/refresh, and Unknown for negative or unavailable state. The global product name/version integration remains owned by L10; Roots icon, splash and desktop resource preservation does not qualify that pending label. Windows/macOS rendering remains untested. Parent native/CTest and full owned functional gates follow the final-child checkpoint.
L9.1 configuration and effective policy
This adapts Jaromil's released S07 (44c15f2cde7d36f2733814d6ab68ce148f1a7079) at the pinned roots.6 endpoint. Forced settings, CLI, read/write text config, JSON settings, network config, default config and built-in defaults retain their order. Native scalar and list negation rules remain in use. The default text file is network-datadir bitcoin_rw.conf; explicit relative/absolute paths and -noconfrw are covered. Its released flat key interpretation is retained. The selected ArgsManager resolves its network before SelectParams without changing global chain parameters. Core v32's nonrecursive args mutex requires lock-aware persistence helpers rather than nested public-method locking.
Complete input and value validation precede publication. The carried reader distinguishes valid EOF from failed reads, including a final invalid line's rewind/comment copy. Read/write text updates close the staged output and replace the file before publishing settings; failed reads, writes, closes and reset removals preserve the selected input. Known sensitive values are redacted in CLI, text and JSON settings diagnostics, including section-prefixed JSON keys. The JSON logging path previously bypassed sensitive-option flags and now uses the same registered flags. Numeric JSON Boolean access retains released conversion behavior.
The additive getmempoolinfo.policy object reports initialized authoritative options and the synchronized live dust rate under the existing mempool lock. Its units, gates, resolved modes and help are retained in native v32 RPCMethod and OpenRPC schemas. The native deprecated fullrbf alias remains hidden unless requested; removed peer RPC fields are not restored. Native libevent-free HTTP and monotonic uptime implementations are unchanged. Connection limits are exercised at defaults, CLI overrides, read/write first-value selection and CLI overrides of that file; zero and negative limits retain their startup errors. Dormant statistics remain unavailable.
Focused Linux qualification passes all 83 selected native cases (223804 assertions) and all 26 owned configuration, policy, HTTP/RPC and sweep functional instances, with no selected skips. Both sweep runtime-config transports and both effective-policy/OpenRPC transports execute. Native regressions include initial/partial/post-rewind read faults, empty/no-newline inputs, failed reset removal, and child-isolated partial-write/close failures using file-size limits. Initial failures are retained: test setup confused an empty path with typed negation; a connection-limit fixture expected a clamp instead of the native startup error; and the large-block HTTP fixture used Core-sized carriers under Roots defaults. That fixture now explicitly selects larger script/carrier and block limits only on its disposable node, preserving the full backpressure assertions and production defaults. These are the focused L9.1 child results; resource and sanitizer checks appear below, followed by parent qualification.
L9.2 resources and dormant support
The released S07 runtime repairs and S09 (8e7ffad5d932e07c0c80f3756f7c3a442786877e) empty-write correction are carried forward. Advisory allocation preserves live bytes across extension and interior ranges. The fallback reads existing chunks, fills only their missing tails, and performs the required seeks between reads and writes. Windows extends only after a successful size query and seek; the destructive Apple truncation path uses the preserving common fallback instead. Core v32's wider offsets also require checked addition and representability checks before any allocation or seek. Empty AutoFile writes return only after validating the handle, including for closed/null files and obfuscated streams.
Native v32 FD saturation, checked consteval GiB literals, RAM detection and cache calculation are retained unchanged. Subprocess closure above a lowered descriptor soft limit remains as accepted in L7. Cache tests cover signed inputs and native architecture bounds without allocating those amounts. Native RAM warnings reserve 2 GiB before taking 75 percent of the remainder; the released v31 calculation is not restored. Statistics source and tests use native btcsignals and RPCMethod, but options, RPC registration and production sampling remain unwired. Both RPC transports reject getmempoolstats, and dormant options fail startup parsing.
Focused Linux checks pass all 102 selected native cases (22607 assertions) and three functional instances, with no selected skips. Five GiB compile contracts include expected compile-time overflow/narrowing rejections. Signed and unsigned FD syscall models pass 24 cases under ASan/UBSan and integer instrumentation. Bounded standard ASan/UBSan probes pass 16 allocation scenarios across the actual utility library and an exact-source forced fallback, two AutoFile profiles, and the dormant model. These are not a full native sanitizer suite. The broader integer-instrumented utility run stops before main in libstdc++ string-map logging initialization; its diagnostic is retained without a suppression or product-pass claim. Initial chrono test API errors and a missing probe link dependency are also retained with their repairs. Windows/macOS, other POSIX ABIs and 32-bit execution remain unqualified and owned by L11. These are the focused L9.2 child results; full parent qualification follows the separate child checkpoint.
L9 parent qualification
Both existing out-of-source profiles build all configured targets with -j 2: the policy profile includes the daemon, IPC, wallet, CLI, tools and benchmarks; the Qt6 profile includes its GUI and tests, with its existing optional features unchanged. All builds finish before testing, and executable/configuration hashes are frozen and checked across the gates. Full native runs each exercise 916 cases successfully: 27040946 assertions in policy and 27029782 in Qt6. Each registers 917 cases; the external script-assets case returns with a warning because its data is unavailable and is not counted as exercised. Full configured CTest passes 382 of 383 tests in policy and 383 of 384 in Qt6; each skips only script-assets. The Qt target runs all six suites with 22 reported passes, no failures/skips, and eight fresh-process display-unit settings operations.
All 33 owned functional instances pass with no skips. This includes the complete 26-instance configuration/settings/read-write/effective-policy/HTTP/RPC/sweep selection, retaining both sweep runtime-config transports and both effective-policy and OpenRPC transports. Seven additional instances cover initialization, shutdown, abort handling, file locks, directory symlinks, isolated help and startup notice. Bounded sanitizer and native-preservation evidence from L9.2 is reused only after matching its current source, generated probes, library, binary and configuration identities. No broader sanitizer or platform qualification is inferred.
Owned tracked and untracked source passes include/guard, locale-use, Python syntax and whitespace checks. Optional whole-tree locale lint reports existing uses in four accepted test files, all unchanged from the starting commit and outside L9; that diagnostic remains for L11's full candidate reconciliation. Ruff/mypy qualification is unavailable with the installed lint dependencies. Successful explicit native fixture databases are removed only after their runs exit zero, retaining complete diagnostic text and evidence; failed runtimes are preserved. The missing external assets, Windows/macOS/other POSIX/32-bit execution and complete candidate sweep remain L11-owned. Parent review and signed commit acceptance are separate from these current local gates.
L10.3 candidate identity and generated documentation
The default product identity is Bitcoin Roots 32.1-roots.1, with Plan-B Foundation copyright and separate historical Bitcoin Core attribution. The explicit CLIENT_VERSION_TAG packaging hook remains tied to the exact Core version. Numeric client/protocol identifiers and the Satoshi wire identity are unchanged. This is an unpublished candidate; the default version label does not establish release approval or publication.
The feature catalog, wallet privacy guide, build/maintenance guidance and curated candidate notes describe current v32 interfaces and retained Roots behavior. The released v31.1-roots.6 notes are preserved byte-for-byte in the historical release-notes directory. Current guidance covers native PSBT v2, removed fixed-fee interfaces, graph/package invariants, Qt6/HTTP/IPC, cache defaults, dormant statistics and the complete key-export disclosure boundary. Historical contributor credits remain intact.
All nine candidate tools were freshly built out of source with -j2 in the existing Linux Qt6 profile. Its wallet, IPC and QR support are enabled; ZMQ, USDT and MiniUPnPc are disabled. Native manpage and example-configuration generators ran twice against these binaries through private regtest/help wrappers with isolated settings. All seven generated manuals and the example configuration have identical hashes across both runs. No generated manual was edited by hand. Public help omits debug-only and unavailable optional-build sections; the generation profile does not qualify those omitted features.
The complete local maintenance suite passes 154 methods with no failures or skips. Focused CTest passes the three build-identity methods and all six Qt suites (22 reported passes, no failures/skips). A fresh isolated regtest node checks current network identity, export/sweep/effective-policy help and the intentional absence of getmempoolstats, then exits cleanly. Documentation option lint, ShellCheck/Bash syntax for 18 owned scripts, workflow/Python/JSON structure and whitespace checks pass. All disposable Git fixture commits are signed and audited with the required key. Initial v32 translation API compile errors and a missing documentation link are retained with their repairs.
These are focused L10.3 Linux results, with synthetic archive/signing/draft and mocked transport fixtures. They do not establish hosted CI, five-platform packages, full candidate acceptance or publication. Missing optional lint tools and broader native/functional/platform gates remain explicit. The L10.2 signed snapshot/replay applies only to its frozen source; later identity/docs changes require a renewed final accepted-SHA portable replay. Parent review, full applicable qualification and commit authorization remain separate.
L10 audit correction: historical review and offline cache tooling
The preservation audit found that the released S29 cache experiment helpers, tests and guide had been omitted while native producer microbenchmarks were carried. Those are distinct tools. The restored contrib/roots/cache_bench* helpers inject insertion and InitData thread CPU/wall timers and reference snapshots only into disposable exact Git exports. Their Linux workload captures process RSS/HWM/CPU, actual reconstruction hits/requests/roundtrips/wire bytes and per-cell raw samples. It retains paired seeds, warmup/sample separation, bounded ownership/timeouts, failures and provenance. There was no perf/BPF collector in the released tool and none is added here.
The current adapter recognizes only exact v31.1/v32.1 version tuples, pinned Core bases and reviewed insertion/InitData bodies. Current logging APIs and the 100-entry default require intentional adaptations; nominal full overlap cannot imply full hits when 128 offers exceed that capacity. Production cache defaults (100 entries/10 decimal MB), relay/permissions, consensus and privacy behavior are unchanged. Native microbenchmarks and source/unit/compile checks do not establish an instrumented experiment run or production performance result.
The S27 policy review and its later released updates are preserved byte-for-byte under doc/historical/mempool-policy-review-v31.1-roots.6.md. The active doc/mempool-policy-review.md records that historical/current distinction and links the current parsing/effective-policy/subdust and preservation contracts. The old 32,768-slot measurements, warmed-vector limitations and unanswered questions remain historical. They cannot qualify v32 or justify a default change; current owning evidence must be assessed independently.
L11.1 folded-series audit and qualification fixtures
The renewed audit starts from accepted candidate 98166e6a54bccd7e44cbf89d4ac58e338a109fac, tree 1681cf5294af773f7361b52b2d021b5dbf5a3268, directly above exact Core f0de87f040621468b956f01f526a43a010310501. The twelve new commits form a linear series, including the ordinary starting-height and audit-correction commits. There is no old Roots merge, fixup/squash residue or new committed build, database, log or evidence artifact. Each commit has the required signing and primary fingerprint. The complete range-diff is a folding aid rather than a one-to-one patch-equivalence claim: all 38 released source commits and their 350 distinct changed paths remain individually accounted for above. The 148 main-only paths retain their separate qualification/integration owners; main's older integrated product does not replace later released repairs.
Current contract and source dispositions
The 25 stable IDs and supplemental address export resolve to current code and observable regressions, not merely to historical case totals:
| Contract | Current implementation and regression boundary |
|---|---|
policy-parsing | src/node/mempool_args.cpp retains strict carrier/dynamic-dust grammar; policy_args_tests.cpp and feature_policy_args.py retain boundary/error oracles. |
effective-policy | getmempoolinfo.policy reports initialized settings through src/rpc/mempool.cpp; rpc_effective_policy.py checks values, schema, precedence and admission parity. It is not a new geteffectivepolicy RPC. |
policy-standardness | src/policy/policy.cpp and admission-only MemPoolAccept checks retain configured script/input/witness limits and reasons; local-policy and sigop fixtures retain valid-block separation. |
policy-carrier | DatacarrierBytes/CalculateExtraTxWeight retain aggregate recognition and quarter-vbyte arithmetic; carrier, Taproot and independent oracle fixtures retain their distinct policy/consensus expectations. |
policy-dust | CTxMemPool::GetDustRelayFee/UpdateDynamicDustFee and node-owned scheduling retain atomic snapshots and floor fallback; dust/lifecycle fixtures remain registered. |
policy-subdust | Admission and mempool entries retain the deficit's modified-fee effect; native multiprecision and functional decimal oracles remain independent of production calculations. Actual fees/rewards remain distinct. |
policy-replacement | Admission uses configured RBF/TRUC gates over native graph accounting; replacement/package/TRUC regressions retain fee and ancestor boundaries. |
policy-coinage | Admission retains next-height maturity and capped coin-block arithmetic; mining priority derives current UTXO age without changing block validity. |
policy-reuse | CTxMemPool::GetScriptConflicts and staged admission retain creation/spend conflict tracking and package/RBF boundaries; spkreuse=allow stays default. |
mining-template | BlockAssembler::addPriorityTxs and residual addChunks retain ordering and accounting; template and generated C++ IPC regressions cover the supported interface. |
peer-controls | Explicit permission implications, current transport privacy and compact-cache accounting remain in networking code; native and functional fixtures preserve count 100, 10 decimal MB and either-zero disabling. |
peer-health | VERSION-derived per-connection m_starting_height reaches node statistics and the real Qt Starting Block view. Unavailable state clears honestly; deprecated startingheight/misbehavior_score RPC fields remain removed. |
wallet-foundation | Native SQLite write-access validation remains present; database discovery and Unix signer descriptor closure retain their corrections. Wallet lock/unlock and descriptor boundaries remain native. |
wallet-rbf | The wallet-derived GUI choice reaches the prepared transaction; decoded transaction/PSBT tests distinguish signaling from replacement policy and preserve native PSBT v2. |
wallet-sweep | CanSweepTo/CreateSweepTransaction, backend authorization and the guarded GUI preview retain local satisfaction, transient keys, cancellation and exact-txid submission boundaries. |
wallet-coin-control | Wallet interfaces and coin-control rows retain factual effective value, locked/reused/grouped outputs and notification invalidation; no privacy score is introduced. |
runtime-config | Read/write config merging and ModifyRWConfigFile/EraseRWConfigFile retain precedence, complete-read and atomic-publication safeguards; settings/config fixtures retain failure cases. |
operator-stats | Native monotonic uptime remains active. Statistics sources remain dormant: no option/RPC registration or production sampling is added. Model source is not a shipped statistics service. |
runtime-resources | File allocation, FD saturation, native checked byte-unit conversion and safe empty AutoFile writes retain their regressions. Native RAM/index cache defaults remain unchanged. |
build-portability | Current system-library probes and Qt6 target wiring remain registered; obsolete Qt5/libevent support is not reinstated. Unavailable platform execution remains unqualified. |
identity-desktop | Roots identity/icons/QR behavior retain native numeric version and Satoshi wire identity. Platform notices now separate Foundation 2026 from Core 2009-year credit. |
maintenance-ci | Current classifiers, export regression selection, failure diagnostics and bounded assurance cadence remain maintained; local fixture results do not establish hosted CI. |
maintenance-release | Git-native immutable source/tag checks, signed portable imports, hosting exclusions and archive/notice/draft fixtures remain intact; publication authority is separate. |
maintenance-docs | Current feature/privacy/build/help guidance is separate from byte-preserved historical notes and policy measurements. Generated manuals remain tied to their recorded native help inputs. |
maintenance-native | Scoped stream/process/test/fuzz adaptations remain attached to their owning behavior; broad historical overlays are not imported by name alone. |
| Supplemental address export | ExportAddressKeys, checked descriptor derivation/MuSig2 provenance and secure key extraction retain the full roots.6 contract, unlock/opt-in boundaries, partial ownership and RPC/Qt secret-history regressions. |
The six previously undisposed S27/S29 tooling/review paths are restored by the accepted correction. The historical policy review is byte-identical to the released source; its active replacement explicitly separates v31 measurements from current contracts. The three S35 platform metadata inputs now retain both copyright holders and their correct year ranges, with numeric bundle/resource versions unchanged. These are source/configured-preprocessing results; they do not claim Windows/macOS runtime execution.
Only four original net-source paths remain absent. FindLibevent.cmake and its Homebrew fixture are obsolete because v32's native HTTP transport has no libevent dependency. src/test/mock_process.cpp is replaced by the mock dispatcher and inherited-descriptor/lowered-limit regressions in src/test/system_tests.cpp. The old doc/roots-v31-port-review.md remains a pinned historical review; this v32 review replaces active acceptance claims. Neither an absent filename nor a renamed upstream API establishes retirement of an operator behavior.
The remaining native-file adoptions have observable replacements. Qt5-only branches in the address filter, translation generation, options/model startup and Qt test linkage are superseded by the supported Qt6 paths and retained Qt6 persistence tests. SQLite's directory write check and SOCKS credential-log redaction are already present in exact Core v32. Native _GiB uses checked conversion, with current resource callers/tests retaining their bounds. Legal help moved to src/common/license_info.cpp; the ordinary client version code and numeric identifiers remain native. Upstream contributor/build/dependency/ release guidance remains upstream guidance, complemented by current Roots build and maintainer documents rather than copied historical instructions.
Released fixture repairs and marker isolation
The audit found that the before-mining synchronization repair in the released wallet output-type fixture had been lost. Bob submits locally, so the current fixture again calls sync_mempools() before Alice mines each returned payment. Framework synchronization after mining does not replace this ordering barrier. The coin-selection and final mixed-input assertions are unchanged.
Two other released fixture adaptations also remain necessary on v32. The sequence-lock fixture deliberately lowers a transaction's modified fee to zero; its two nodes now explicitly disable the priority area so that the transaction remains available for subsequent lock assertions. The UTXO-to-SQLite fixture creates arbitrary tiny outputs; it now disables the subdust penalty only for that conversion experiment, while retaining native coinstatsindex setup. Fresh failures before these repairs reproduced premature priority inclusion and the dust-deficit modified-fee rejection respectively. No production mining, wallet, admission default or consensus assertion was changed.
Released fixture-only rejectparasites=0 flags are not blindly restored. Current GetAntiFeeSnipingLockTime in src/wallet/spend.cpp excludes the reserved CAT21 marker before ordinary wallet transactions are produced. Output-type grouping, backup and wallet-group fixtures exercise that current path with normal local policy. This source-based disposition is separate from the two necessary non-marker isolation settings above.
Intermediate buildability and accepted-source replay
Historical precommit source captures are reconciled to each accepted thematic commit, not to an equal tip tree. Original dirty patches reconstruct the exact accepted complete trees for L2/L3/L4/L5/L6/L9. L8's complete 3,059-path source capture matches its accepted tree. L7's changed source identities match its accepted commit; its earlier documentation snapshot is distinguished from post-build executable identities. The ordinary L6 correction was built in the headless profile while protected Qt/documentation work was dirty; those GUI inputs were outside that profile, and its owned node/fixture inputs match the correction commit. Later L4 fuzz, L5 IPC and L6 benchmark build receipts cover their final compiled fixtures. L1 is documentation-only.
The original L10 complete source capture matches its accepted 3,128-path tree. The later 14-path audit correction preserves runtime compilation inputs; actual native translation-unit preprocessing is identical across its resource-only header change. Its configured platform metadata, helper suites and bounded isolated experiment have distinct receipts. This supports the affected Linux profiles, without inventing a fresh platform build or treating the experiment as an uninstrumented production performance result.
Recorded builds used their original prior HEAD plus dirty source and generated Git metadata. Their later clean commit SHA was not their version stamp. Pre-build binary hashes and superseded binaries are not asserted to be current post-build identities. Complete source/command/configuration/binary provenance and original successful build logs support the intermediate compilation claims; the final tree's green result alone is not used for them.
The accepted 98166e6a54bccd7e44cbf89d4ac58e338a109fac export has an actual exact-Core replay of twelve signed commits ending at 58c7da17032832204ca4fd27d938e83f5246993b. All 3,136 source entries, product blobs/modes/symlinks, author/date/message attribution and CRLF bytes agree, while .github/** remains exact Core. The earlier e979 replay remains historical; these source/replay/accepted identities are separate. Qualification fixture and review-document changes after this accepted source require final accepted-SHA reconciliation at the later parent/handoff gate.
The aggregate audit leaves consensus directories, primitive serialization, script interpreter, cryptographic dependencies and chain parameters unchanged. Necessary validation.cpp touches stay within local admission/package policy; descriptor/key extensions support wallet export and transient sweep rather than changing consensus validation or signing algorithms. No translation subtree is mass-rebranded. Contributor/source credits remain explicit. Full candidate, external compatibility, hosted/platform and instrumented acceptance gates remain separate work; this audit is not their passing result.
L11.2 external compatibility qualification
The main-only external harness is retained as a separately qualified source package for later main integration. Its baseline is main commit 5d09cb937f0024a6fae0b4ba3da7f0ae7fd0063f; seven compatibility-tool paths are adapted without importing those tools into this product branch. The maintained 30.3 profile remains the default, with its regression support intact. The explicit 32.1 profile requires exact Core f0de87f040621468b956f01f526a43a010310501 and Roots 98166e6a54bccd7e44cbf89d4ac58e338a109fac, committed version metadata, official Core origin/tag identity, ancestry, CLI/RPC version and fresh controlled build provenance. Caller-provided binaries remain unqualified. The upstream Core-tag signature check was explicitly waived; exact source matching remains required and no upstream signer-authentication result is asserted.
All 70 external-tool regressions passed. Fresh exports built both binaries with caches disabled before the six-case full comparison executed. Roots produced the ordinary transaction's block through its native producer and Core accepted the returned bytes. Core produced a block containing the Roots-policy-rejected carrier transaction and Roots accepted its returned bytes. The report retains both directions, raw block/transaction identities, producer/receiver binaries, pre-submit tips and equal post-exchange UTXO state. Eight admission checks, valid/invalid block outcomes, sigop/subdust boundaries, disconnect/reconnect and two successful restarts per node remain distinct. Four intermediate and two final shutdowns exited successfully. This qualifies the pinned native component pair; later source SHAs require their own exact qualification decision.
L11.3 initial candidate matrix and corrected-source renewal
The initial Linux matrix below used accepted 98166e6 native source plus the declared qualification fixture/document changes. Its logs, snapshots and tested binaries remain historical evidence for that exact state. GNU 14.2 RelWithDebInfo enables wallet, Qt 6.8.2, QR, IPC, external signer, benchmark, experimental kernel and chainstate targets. Compilation and linking settle before each affected binary is tested. Frozen source, compiler/configuration/generated-header and binary identities accompany every profile; earlier source snapshots, failures and tested binaries remain retrievable. The bounded source/lint correction described below has a separate signed private snapshot and renewed source/configuration/binary evidence. The table records the initial matrix; its numeric totals are not transferred to the corrected binaries. Current acceptance requires the renewed matrix and the external gates below.
| Current execution | Observed scope and result |
|---|---|
| Full native unit executable | 917 cases passed, 26,949,822 assertions, no skipped cases. |
| Complete configured CTest | All 386 registered tests passed with no skips, including embedded-library, IPC, kernel and GUI registrations. |
| Complete extended functional runner | All 519 selected instances completed: 503 passed, 16 prerequisite skips, no failures. This includes 197 benchmark sanity instances and 322 functional/script instances; benchmark sanity is not performance measurement. |
| Full Qt6 executable | Six classes passed all 10 substantive registered functions plus 12 setup/cleanup functions; 22 PASS rows, no failures/skips. |
| Maintainer fixtures | The complete 179-test current maintenance suite passed; source/lint proposal validation is recorded separately from canonical application. |
| Current bounded instrumented fuzz | All nine maintained fast targets passed with 238,077 measured inputs, including dynamic_dust_fee and policy_args. Each uses the five-second campaign, five-second input, seed-one and 15-second outer bounds. |
| Current resource instrumentation | Clang 19 ASan/UBSan passed 16 allocation scenarios, two stream profiles and the dormant-model exercise. Five GiB compile contracts include four expected rejections; signed/unsigned FD models passed 24 cases under address/undefined/integer/float-divide instrumentation. These models do not establish another platform's ABI. |
The official unit-test asset file was fetched independently of any full fuzz corpus, pinned to qa-assets commit b33d85102d169b54d966ea315ad81a636680aefa and checked against its Git blob and SHA256. Its 2,244 vectors executed in script_assets_test with 141,917 passing assertions. The historical missing-assets warning/CTest skip above is closed for this current profile. Both bind-address fixtures also executed using already assigned routable IPv4/IPv6 addresses; no host address/routing change was made.
Four native test/fuzz sources now use locale-independent formatting or literal boundary inputs. Integer one-past-limit fixtures remain independent of the production parser, with their original accept/reject assertions and inputs. Three Core mining experiments explicitly request their required mining profile: fee-free orphan checks disable the Roots priority area, disconnect-pool FIFO checks use full-size fee-ordered blocks, and the upload-budget experiment uses full-size blocks for its fixed request count. The initial failures and repaired runs are retained. Production mining/admission defaults and consensus rules remain unchanged.
The first native XML-file-sink capture reproduced a Boost cleanup failure in the forked lock-directory test. The identical case and complete native suite passed with normal CI-style output; external CTest JUnit provides structured results without that file-sink lifetime interaction. The initial fuzz starts executed zero inputs because the local environment omitted current CI's existing dependency suppressions. The passing campaigns use the unchanged Core v32 suppression files and current CI sanitizer environment. No new suppression or ignored test was added. Leak detection was enabled in the bounded fuzz/resource runs; these are not full-corpus or unbounded leak-free claims.
Sixteen functional prerequisite skips remain explicit: seven previous-release profiles, two Python Cap'n Proto interfaces, ZeroMQ, five USDT profiles and the plyvel-backed UTXO corruption test. Existing local/tool paths were checked. Required-feature configure probes fail for missing ZeroMQ, USDT, system LevelDB and system secp256k1 dependencies; native embedded libraries remain enabled. Rust lint infrastructure, ruff/mypy/lief, actionlint, PowerShell and offline Markdown-link tooling are unavailable in this environment. Current hosted Clang 23, instrumented MSan libc++, cross toolchains/emulators and native Windows/macOS runners remain unqualified. Their exact prerequisites and case identities belong to the canonical review/CI handoff owner.
The root approved the exact bounded source/lint correction under final qualification ownership: fifteen executable script modes, the benchmark helper's path-correct include guard, six release scripts and one CI helper using the native C.UTF-8 locale, and removal of the unused policy/policy.h include from policy/settings.cpp. Both externally visible policy defaults remain 20. Incidental unused static initialization disappears, so binary identity is renewed; the initial matrix is not relabeled as current evidence. All test assertions remain intact and no production default changes.
The corrected full source and intended executable modes are frozen in a required-key-signed private qualification snapshot. The canonical index stays untouched; metadata-dependent lints run against the byte/mode-equivalent snapshot. Portable export/replay, deterministic release fixtures and current native/instrumented gates are renewed from that identity. The original seven external compatibility-tool paths and their initial native qualification stay unchanged. A separate copied package deliberately pins the exact corrected snapshot, retains every provenance/generation guard and fresh controlled-build requirement, and runs its full regression and six-case bidirectional exchange qualification. The eventual accepted product SHA remains a distinct identity requiring final reconciliation; no private snapshot implies release authority.
Read-only queries found no existing 32.1 review branch/PR or normal pull-request workflow run at the pinned candidate SHA. The canonical review handoff must obtain normal exact-SHA hosted CI under its own authority. No public ref, workflow dispatch, release signing or publication is performed by this qualification. Final parent/whole-branch review remains required.