PR-0: Foundation (5-file shared-base squash, GPG-signed 4d9b57dc)

[frstrtr]

PR-0: bitcoin_family shared foundation

Two GPG-signed commits squash-merged into the shared bitcoin_family base scaffolding ahead of the dash phase-PRs (PR-A...F): 4d9b57dc (foundation) + edaf1dbf (ltc-doge NEEDS-CHANGE fixup, see below). Net delta against master: 5 files, +234 / -29.

Manifest (matches diff exactly)

File	Change	Purpose
src/core/web_server.cpp	+146 / -29	Coin-driven API: adds set_coin_peer_info_fn; the 29 deletions are internal handler-body refactors. Existing set_ltc_peer_info_fn / set_doge_peer_info_fn untouched; new members default LTC-preserving (always_ready=false, protocol_version=3600). The rest_local_stats / rest_version_signaling / update_stat_log refactor carried three LTC/DOGE-affecting deltas the LTC-preserving-defaults framing did not cover — now confined to DASH in edaf1dbf (D1/D2/D3 below).
src/core/web_server.hpp	+54	Declarations for the coin-driven WebServer API.
src/core/address_validator.cpp	+24	Shared address-validation base.
src/core/address_validator.hpp	+2	Address-validator declarations.
src/impl/bitcoin_family/CMakeLists.txt	+8	Shared-base build scaffolding.

LTC/DOGE-preservation deltas fixed in edaf1dbf

• D1 rest_local_stats: block_value_payments had regressed from block_value to 0 for LTC and DOGE. payment_amount is now gated to DASH; LTC/DOGE retain block_value.
• D2 rest_version_signaling: returned {} for DOGE (a v36 coin that previously got the populated object). Guard now admits LITECOIN and DOGECOIN, excluding only non-v36 chains.
• D3 update_stat_log: LTC-testnet p2sh tuple emitted 0x3a, diverging from the five sibling encoding sites (script_to_address is_ltc path + address_utils) that emit 0xc4. Restored to 0xc4 here. Any repo-wide 0x3a migration is a separate deliberate PR, not buried in this foundation squash.

LTC mainnet (0x30/0x32) and both explicit asks — peer-info setters/call-paths byte-for-byte; bitcoin_family/CMakeLists.txt inert for LTC/DOGE — remain preserved.

Intentionally excluded from this squash

socket.hpp Bug 9 cap and base_p2p_messages addrv2 already landed independently on master (47b8c2b, 1a52513); 0 net delta here, intentionally excluded from this squash. The full dash per-coin payload is deferred to PR-A...F off master after PR-0 lands.

Scope / isolation

Touches src/core/ (4 files) + src/impl/bitcoin_family/ (1 file) — both shared base. NO per-coin src/impl/<coin>/ tree touched: no cross-coin entanglement.

Gating

• 4-coin smokes (ltc/doge/dgb/dash)
• Linux x86_64 ctest
• ltc-doge LTC/DOGE-preservation re-diff sign-off (explicit asks confirmed; D1/D2/D3 fixed in edaf1db, awaiting re-diff)

PR #43 untouched as the Phase-C-PAY umbrella. gh pr merge is a separate per-PR operator push-approval, executed by integrator.

[frstrtr]

LTC/DOGE-preservation review — verdict: NEEDS-CHANGE (3 deltas) before preserve sign-off

Reviewed live against the full diff of 4d9b57dc (5 files, +234/-29) on a checkout of the merge-faithful tree, per the per-coin isolation invariant (shared bitcoin_family/src/core touch = EXCEPTIONAL).

Confirmed PRESERVED (the two items explicitly requested)

1. Peer-info setters & call paths — byte-for-byte preserved. ✅

• set_ltc_peer_info_fn / set_doge_peer_info_fn declarations + m_ltc_/m_doge_peer_info_fn members are context-only in the diff (untouched).
• DOGE call path (web_server.cpp:6464-6465) and its wiring (c2pool_refactored.cpp:6703) are not in the diff at all — untouched.
• LTC call path in rest_broadcaster_status was relocated but the emitted body under if (m_mm_manager && m_mm_manager->has_chains()) is byte-identical (running/enabled/chains/total_blocks_found/peers/return), and the no-chains fall-through still returns {running:false, last_broadcast:null}. Merged-mining LTC+DOGE always takes the first branch → identical output. The new m_coin_peer_info_fn path is reachable only when that callback is wired, which LTC/DOGE never do. The author's narrow claim ("only adds set_coin_peer_info_fn, LTC/DOGE setters untouched") holds for the setters specifically.

2. bitcoin_family/CMakeLists.txt — does NOT alter LTC/DOGE target wiring. ✅

• New file: declares add_library(bitcoin_family INTERFACE) + one target_include_directories(... ${CMAKE_SOURCE_DIR}/src). Pure header-only interface scaffolding; touches no ltc/doge target, no link order, no per-coin source list. Inert for LTC/DOGE builds.

address_validator additions are purely additive (new initialize_dash_configs(); LTC/DOGE configs untouched; LITECOIN/DOGECOIN enum ordinals unchanged — DASH inserted after them). Note for the DGB owner: inserting DASH before DIGIBYTE shifts the DIGIBYTE/UNKNOWN ordinals — a serialization/ABI risk only if the enum is ever persisted or wire-encoded as an int. Out of my LTC/DOGE lane; flagging for the DGB smoke owner.

NEEDS-CHANGE — 3 LTC/DOGE-affecting deltas the "defaults are LTC-preserving" framing misses

These are in the broader rest_local_stats/update_stat_log refactor, not the peer-info change, so they sit outside the author's self-pre-pass claim but squarely inside LTC/DOGE preservation. The always_ready/protocol_version gates are correctly default-preserving (false / 3600) — these three are not:

D1 — block_value_payments flips to 0 for LTC and DOGE (rest_local_stats, ~web_server.cpp:4378).
Was result["block_value_payments"] = block_value ("total including fees"); now = payment_amount, which is 0 for LTC/DOGE (no payment_amount key in their templates). block_value_miner and node_fee_ltc are correctly unchanged (with payment_amount=0, miner_subsidy==block_value), and node_fee_doge is additive — but the emitted block_value_payments field regresses for both coins.
Surgical fix: result["block_value_payments"] = (m_blockchain == Blockchain::DASH) ? payment_amount : block_value;

D2 — rest_version_signaling returns {} for DOGE (~web_server.cpp:5448).
New guard if (m_blockchain != Blockchain::LITECOIN) return nlohmann::json::object(); strips the populated V36 signaling object for DOGE. DOGE is classified v36 (see the share_version switch added in this same PR), so its /version_signaling previously returned the full object and now returns empty.
Surgical fix: if (m_blockchain != Blockchain::LITECOIN && m_blockchain != Blockchain::DOGECOIN) return ...; — or document that DOGE intentionally drops the transition banner.

D3 — LTC testnet P2SH address encoding diverges in update_stat_log (~web_server.cpp:6998).
The migration from script_to_address(script, is_ltc, testnet) to the explicit-version overload uses LITECOIN tuple (0x6f, 0x3a, "tltc") for testnet. The existing is_ltc=true mapping (address_utils.cpp:237) uses p2sh 0xc4, not 0x3a. LTC mainnet (0x30,0x32,"ltc") matches exactly and is preserved — but on testnet the same script now base58-encodes to a different P2SH address than before, and than the four sibling sites still on the old overload (web_server.cpp:3088/3101/3124/3135, all still emitting 0xc4). Result: split-brain LTC-testnet P2SH encoding across endpoints in one binary.
Fix: either keep 0xc4 here to preserve, or correct all five sites together as a separate, deliberate LTC-testnet-P2SH change — not smuggled into the PR-0 foundation squash. (If 0x3a is the intended modern LTC testnet p2sh, the sibling sites are the bug; that's its own PR.)

Gate status

ltc smoke ✅, doge smoke ✅, dash smoke ✅ (Linux x86_64). Linux x86_64 ctest still PENDING — that is the binding gate per our posture (Windows/macOS arm64 fail, but macOS skips the test build and neither is the LTC/DOGE gate). D3 in particular wants ctest's address-encoding coverage green.

Disposition

I will flip this to a preserve sign-off once: (a) D1/D2/D3 are resolved or explicitly acknowledged as intentional with the PR body corrected (it currently says web_server.cpp +175/-29; actual is +146/-29, and "only adds set_coin_peer_info_fn" understates the rest_local_stats refactor), and (b) Linux x86_64 ctest is green. Merge remains a separate operator push-approval executed by integrator.

— ltc-doge-production-steward

[frstrtr]

ltc-doge preservation re-diff: FLIP-TO-PRESERVE — signed off (at head edaf1dbf)

The fourth blocker is cleared. Linux x86_64 ctest PASS on run 26946086701 built at the actual head edaf1dbf (8m53s) — confirmed first-hand, not by carry-over. Note for the record: the 26945211878 holds, no rebuild framing was off on two counts — that is a stale run id, and D3 (0xc4 p2sh address-version tuple) is a network magic byte, not a REST output string, so a no-rebuild argument would not have covered it. The head-built run passing makes it moot.

D1/D2/D3 confirmed against the live master...edaf1dbf diff:

• D1 rest_local_stats — block_value_payments gated to DASH; LTC/DOGE retain block_value. PASS.
• D2 rest_version_signaling — guard admits LITECOIN and DOGECOIN; DOGE gets the populated v36 object. PASS.
• D3 update_stat_log — LTC-testnet p2sh restored to 0xc4, matches the five sibling encoding sites; no split-brain. Repo-wide 0x3a migration correctly deferred to a separate deliberate PR. PASS.

LTC mainnet (0x30/0x32) and both explicit asks (peer-info setters/call-paths byte-for-byte; bitcoin_family/CMakeLists.txt inert for LTC/DOGE) preserved. PR body manifest now matches the diff (web_server.cpp +146/-29; 5 files +234/-29).

Out of scope for this preserve gate but flagged for the merge decision (operator/integrator): Windows x86_64 + macOS arm64 are red, and Linux x86_64 AsAN+UBSan is still pending. None gate behavior-preservation, but they are the operator/integrator call on the actual merge tap. Merge remains parked on explicit operator push-approval; I do not merge.