fix 1.23.8: ladder harness-style wallet serve for coin settle

2026-08-29 02:40:48 +02:00 · 2026-08-29 02:40:48 +02:00 · ba899dd8b5
commit ba899dd8b5
parent 2451c66173
8 changed files with 233 additions and 22 deletions
--- a/CLI-AUTOMATION-NOTES.md
+++ b/CLI-AUTOMATION-NOTES.md
@ -35,6 +35,18 @@ Legend:

 ---

+
+## 1a. Harness-style wallet daemon (mon ladder + e2e)
+
+| | |
+|--|--|
+| **Seen** | Ladder `OK_BANK` with `avail=0` while mon e2e / harness get full coins; force-select 400 without `selected_exchange` |
+| **Area** | **cli** / **core** / **ops** |
+| **Problem** | Harness uses a long-lived **WalletService** (unix socket client). One-shot `wallet-cli` starts and stops the shepherd each call — after bank `transfer_done`, coins often stay `pendingIncoming` until something keeps the wallet awake. `run-until-done` is banned (hangs). |
+| **Workaround** | Mon: `advanced serve --unix-path $WSOCK` then `--wallet-connection=$WSOCK` (e2e default; ladder `LADDER_WALLET_SERVE=1` via `ladder/lib_wallet_serve.sh`). Force-select body must be `{"reserve_pub":"…","selected_exchange":"<exchange payto from /keys>"}`. Mint may include `"exchange_url":"$EX"`. Poll balance + bank status; never `run-until-done`. |
+| **Wanted** | Documented wallet-cli “serve for automation” profile matching harness WalletClient. |
+
+
 ## 2. Withdraw completion is not observable from CLI alone

 | | |
--- a/TESTS.md
+++ b/TESTS.md
@ -25,7 +25,7 @@ Every check line has a **global** run number and a **grouped** id:
 | **sanity** | `check_sanity.sh` | `bank` `exchange` `merchant` |
 | **server** | `check_server.sh` | (flat `server-NN` or host groups) |
 | **e2e** | `check_e2e.sh` | `prereq` `load` `bank` `wallet` `bankwd` `settle` `pay` `shop` `paivana` `dig` `report` |
-| **ladder** | `check_amount_ladder.sh` + `ladder/lib_pay.sh` | `plan` `load` `withdraw` `pay` `report` — classic / max-search; phases `ladder` · `max-ladder` · `goa-ladder` (alias) |
+| **ladder** | `check_amount_ladder.sh` + `ladder/lib_pay.sh` + `ladder/lib_wallet_serve.sh` | `plan` `load` `withdraw` `pay` `report` — classic / max-search; harness-style `advanced serve` (`LADDER_WALLET_SERVE=1`); phases `ladder` · `max-ladder` · `goa-ladder` (alias) |

 **Why groups:** flat `www-001`…`www-080` was hard to talk about.  
 `www.bank-04` / `e2e.bankwd-02` pin the failure to a logical block.
--- a/2
+++ b/2
@ -1 +1 @@
-1.23.7
+1.23.8
--- a/VERSIONS.md
+++ b/VERSIONS.md
@ -17,6 +17,7 @@ Git tags: `vMAJOR.FEATURE.FIX` (e.g. `v1.8.0`). File `VERSION` omits the `v` pre

 | Tag | Date (UTC) | Notes |
 |-----|------------|--------|
+| **v1.23.8** | 2026-08-29 | **Bugfix / ladder:** harness-style wallet daemon — `ladder/lib_wallet_serve.sh` (`advanced serve` + `--wallet-connection`, default `LADDER_WALLET_SERVE=1`). Mint sends `exchange_url`; settle/pay-wait skip one-shot `run-pending` when serve is up. Aligns mon ladder with taler-harness WalletService + e2e serve path (no run-until-done). CLI-AUTOMATION-NOTES §1a. |
 | **v1.23.7** | 2026-07-22 | **FP stage host-agent:** ship `taler-monitoring-fp-stage.{service,timer}`; `install-host-agent.sh --fp-stage` (stagepaysan) enables FP mon only, disables GOA timers on that host, enforces suite tip knobs. Same sticky/publish pipeline as GOA. |
 | **v1.23.6** | 2026-07-22 | **Strict suite tip for mon pages:** every host-agent run must `fetch`+`reset` to `origin/main` and **refuse** if HEAD ≠ remote tip (`SUITE_REQUIRE_REMOTE_TIP=1`). Overlay is refreshed **from** the tip after update (never re-applied over tip by default — was publishing stale site-gen). `git clean` after reset. Mon HTML still embeds that COMMIT in sticky/meta. |
 | **v1.23.5** | 2026-07-22 | **Bugfix / host-agent:** multi-agent shared `/var/www/monitoring-sites` — stop full-tree `rsync --delete` (wiped sibling pages e.g. `/monitoring-max-ladder/` after main/surface/aptdeploy runs). Publish **only this job’s** `HTML_OK_DIR`/`HTML_ERR_DIR` per host; `apply-monitoring-live.sh` merges without tree-delete. |
--- a/check_amount_ladder.sh
+++ b/check_amount_ladder.sh
@ -10,8 +10,8 @@
 # Flow (bank landings):
 #   1) GET /intro/auto-account.json  → personal *account-* (balance 0)
 #   2) Mint pool withdrawals as explorer + confirm when selected
-#   3) wallet-cli accept-uri only (no run-until-done)
-#   4) bank confirm when selected; settle = balance + transfer_done
+#   3) wallet-cli accept-uri via harness-style advanced serve (no run-until-done)
+#   4) bank confirm when selected; settle = balance + transfer_done (serve keeps shepherd)
 #   5) pay via free-amount template (GOA: goa-free) or private orders
 #
 # Modes (LADDER_MODE):
@ -89,6 +89,9 @@ elapsed_ms() {
 : "${LADDER_MAX_SEED:=}"
 # After withdraw, wait for spendable available>0 before pay (secs total; shows countdown)
 : "${LADDER_PAY_WAIT_AVAILABLE_S:=180}"
+# Harness-style long-lived wallet daemon (advanced serve). 1=on (default).
+# Without serve, coins often stay pendingIncoming after bank transfer_done.
+: "${LADDER_WALLET_SERVE:=1}"

 # Historic libeufin-ish absolute ceiling used as default LADDER_MAX_AMOUNT on GOA
 LADDER_ABS_CEILING="4503599627370496"
@ -225,13 +228,18 @@ ladder_over() {
 }

 wcli() {
-  # taler-helper-sqlite3 is Python ≥3.11; prefer Homebrew/local python on macOS
-  # so Apple /usr/bin/python3 3.9 does not FATAL the sqlite backend (no reservePub).
+  # Prefer harness-style serve socket (long-lived shepherd); else one-shot --wallet-db.
+  # PATH: taler-helper-sqlite3 needs Python ≥3.11 (Homebrew/local on macOS).
  local _path="${PATH:-}"
  case ":${_path}:" in
    *:/opt/homebrew/bin:*) ;;
-    *) _path="/opt/homebrew/bin:/usr/local/bin:${_path}" ;;
+    *) _path="/opt/homebrew/bin:/usr/local/bin:${HOME}/.local/bin:${_path}" ;;
  esac
+  PATH="$_path"
+  if command -v wallet_serve_wcli >/dev/null 2>&1; then
+    wallet_serve_wcli "$@"
+    return $?
+  fi
  if [ -n "${CLI_JS:-}" ] && [ -f "$CLI_JS" ]; then
    PATH="$_path" node "$CLI_JS" --wallet-db="$WDB" --no-throttle "$@"
  else
@ -481,6 +489,8 @@ source "$ROOT/metrics.sh"
 # Payment module (classic + max-search payable)
 # shellcheck source=ladder/lib_pay.sh
 source "$ROOT/ladder/lib_pay.sh"
+# shellcheck source=ladder/lib_wallet_serve.sh
+source "$ROOT/ladder/lib_wallet_serve.sh"
 METRICS_DIR="$REPORT_DIR"
 ALT_UNITS_FILE="${REPORT_DIR}/alt_unit_names.json"
 export METRICS_DIR CUR WDB CLI_JS ALT_UNITS_FILE
@ -578,18 +588,31 @@ wallet_prepare() {
  local label="${1:-wallet}"
  WDB="$SCRATCH/wallet-${label}.sqlite3"
  export WDB
-  if [ ! -f "$WDB" ]; then
-    wcli exchanges add "$EX" >"$SCRATCH/ex-add-$label.out" 2>&1 || true
-    wcli exchanges update "$EX" >"$SCRATCH/ex-upd-$label.out" 2>&1 || true
-    wcli exchanges accept-tos "$EX" >"$SCRATCH/ex-tos-$label.out" 2>&1 || true
+  # Start harness-style daemon once per DB (shepherd stays up for wirewatch settle).
+  if [ ! -f "$WDB" ] || [ -z "${WSERVE_PID:-}" ] || ! kill -0 "${WSERVE_PID:-0}" 2>/dev/null; then
+    wallet_serve_install_trap
+    if wallet_serve_start; then
+      if [ -n "${WSERVE_PID:-}" ] && [ -S "${WSOCK:-}" ]; then
+        ok "wallet serve" "pid=$WSERVE_PID sock=$WSOCK (harness-style; no run-until-done)"
+      fi
+    else
+      if [ "${LADDER_WALLET_SERVE:-1}" = "1" ]; then
+        warn wallet "wallet serve unavailable"           "problem: advanced serve did not come up — falling back to one-shot CLI (coins may stay pendingIncoming). See $SCRATCH/wallet-serve.log"
+      fi
+    fi
  fi
+  # ToS / exchange — via serve socket when up
+  wcli exchanges add "$EX" >"$SCRATCH/ex-add-$label.out" 2>&1 || true
+  wcli exchanges update "$EX" >"$SCRATCH/ex-upd-$label.out" 2>&1 || true
+  wcli exchanges accept-tos "$EX" >"$SCRATCH/ex-tos-$label.out" 2>&1 || true
 }

 t0=$(now_ms)
 rm -f "$SCRATCH/wallet-main.sqlite3"
+wallet_serve_stop 2>/dev/null || true
 wallet_prepare "main"
 ms_tos=$(elapsed_ms "$t0")
-ok "wallet exchange + ToS (${ms_tos}ms) — cumulative DB for withdraw+pay (no run-until-done)"
+ok "wallet exchange + ToS (${ms_tos}ms) — cumulative DB for withdraw+pay (serve=${LADDER_WALLET_SERVE:-1})"

 : "${LADDER_MAX_RUNGS:=99}"
 if [ "${LADDER_MODE}" = "max" ]; then
@ -664,10 +687,11 @@ ladder_withdraw_rung() {

  # mint from explorer pool
  t0=$(now_ms)
+  # e2e/harness: pass exchange_url so bank binds the op to our exchange
  code=$(curl -sS -m 30 -o "$SCRATCH/wd-$tag.json" -w '%{http_code}' \
    -H "Authorization: Bearer ${TOK}" \
    -H 'Content-Type: application/json' \
-    -d "{\"amount\":\"${AMT}\"}" \
+    -d "{\"amount\":\"${AMT}\",\"exchange_url\":\"${EX}\"}" \
    "${BANK}/accounts/${EXP_USER}/withdrawals")
  ms_mint=$(elapsed_ms "$t0")
  WID=$(python3 -c 'import json;d=json.load(open("'"$SCRATCH"'/wd-'"$tag"'.json"));print(d.get("withdrawal_id") or "")' 2>/dev/null || true)
@ -960,11 +984,14 @@ sys.exit(0 if Decimal(sys.argv[1]) > Decimal(sys.argv[2]) else 1)
        | python3 -c 'import json,sys; d=json.load(sys.stdin); print(d.get("transfer_done"), d.get("status"))' 2>/dev/null || echo "?")
      if echo "$xfer" | grep -qi True; then
        bank_done=1
-        # nudge wallet to pick up wire (bounded; not run-until-done)
+        # With harness-style serve, shepherd is already running — poll balance only.
+        # One-shot CLI: bounded run-pending (never run-until-done).
        wcli transactions >"$SCRATCH/tx-settle-$tag.out" 2>&1 || true
-        if command -v timeout >/dev/null 2>&1 || command -v gtimeout >/dev/null 2>&1; then
-          _to=$(command -v gtimeout 2>/dev/null || command -v timeout)
-          "$_to" 8 wcli advanced run-pending >"$SCRATCH/run-pending-settle.out" 2>&1 || true
+        if [ -z "${WSERVE_PID:-}" ] || [ ! -S "${WSOCK:-}" ]; then
+          if command -v timeout >/dev/null 2>&1 || command -v gtimeout >/dev/null 2>&1; then
+            _to=$(command -v gtimeout 2>/dev/null || command -v timeout)
+            "$_to" 8 wcli advanced run-pending >"$SCRATCH/run-pending-settle.out" 2>&1 || true
+          fi
        fi
        # after bank done, still prefer coins; allow remaining rounds for available
        note="bank transfer_done waiting coins avail=${after} $xfer"
--- a/ladder/README.md
+++ b/ladder/README.md
@ -30,3 +30,16 @@ There is **no** `check_goa_ladder.sh` — removed; use the phases above.
 | `LADDER_FREE_TEMPLATE` | `goa-free` | template id (hacktivism) |
 | `LADDER_FREE_TEMPLATE_INSTANCE` | merchant instance | e.g. `goa-demo-cp4zqk` |
 | `LADDER_PAY_WAIT_AVAILABLE_S` | `90` | after withdraw, wait for spendable **available**>0; logs **withdraw board in start order** (OK/OK_BANK) with live `pendingIncoming`→`done` + timer (same wallet DB) |
+
+
+## Wallet daemon (harness-style)
+
+taler-harness keeps a **WalletService** up for the whole withdraw/confirm/wirewatch path.
+Mon ladder defaults to the same idea:
+
+| Env | Default | Meaning |
+|-----|---------|---------|
+| `LADDER_WALLET_SERVE` | `1` | `advanced serve` + `--wallet-connection` (no `run-until-done`) |
+| `LADDER_WALLET_SERVE=0` | — | one-shot CLI only (coins may stay `pendingIncoming` / `OK_BANK`) |
+
+Implementation: `ladder/lib_wallet_serve.sh` (shared with the e2e pattern).
--- a/ladder/lib_pay.sh
+++ b/ladder/lib_pay.sh
@ -371,12 +371,14 @@ ladder_wait_available() {
      info pay "timer 0s left — stop waiting (available still ${CUR}:0)"
      break
    fi
-    # nudge wallet (same DB); do not run-until-done (can hang)
+    # serve: shepherd already running (harness-style). one-shot: bounded run-pending.
    wcli transactions >"$SCRATCH/tx-wait-pay.out" 2>&1 || true
-    if command -v timeout >/dev/null 2>&1 || command -v gtimeout >/dev/null 2>&1; then
-      local _to
-      _to=$(command -v gtimeout 2>/dev/null || command -v timeout)
-      "$_to" 8 wcli advanced run-pending >"$SCRATCH/run-pending.out" 2>&1 || true
+    if [ -z "${WSERVE_PID:-}" ] || [ ! -S "${WSOCK:-}" ]; then
+      if command -v timeout >/dev/null 2>&1 || command -v gtimeout >/dev/null 2>&1; then
+        local _to
+        _to=$(command -v gtimeout 2>/dev/null || command -v timeout)
+        "$_to" 8 wcli advanced run-pending >"$SCRATCH/run-pending.out" 2>&1 || true
+      fi
    fi
    # progress board every ~5s (tick every 2s → every 3rd)
    if [ $((tick % 3)) -eq 0 ]; then
--- a/ladder/lib_wallet_serve.sh
+++ b/ladder/lib_wallet_serve.sh
@ -0,0 +1,156 @@
+# ladder/lib_wallet_serve.sh — long-lived wallet daemon (harness-style)
+#
+# taler-harness uses WalletService + WalletClient over a unix socket so the
+# wallet shepherd stays up while bank confirm + wirewatch complete. Mon e2e
+# already mirrors that with `advanced serve`. Ladder historically used one-shot
+# CLI only → coins often stayed pendingIncoming (OK_BANK) until pay-wait.
+#
+# Source after lib.sh. Callers set: SCRATCH, WDB, CLI_JS or WALLET_CLI, NODE_BIN.
+# Optional: LADDER_WALLET_SERVE=1 (default), E2E_WALLET_SERVE (e2e alias).
+#
+# Public:
+#   wallet_serve_resolve_bins   — NODE_BIN + CLI path
+#   wallet_serve_start          — advanced serve --unix-path
+#   wallet_serve_stop           — kill daemon
+#   wallet_serve_wcli args…     — client via socket if up, else --wallet-db
+#   wallet_serve_install_trap   — EXIT cleanup (idempotent)
+
+: "${LADDER_WALLET_SERVE:=1}"
+# e2e reuses same default name
+: "${E2E_WALLET_SERVE:=${LADDER_WALLET_SERVE}}"
+
+WSOCK="${WSOCK:-}"
+WSERVE_PID="${WSERVE_PID:-}"
+_WALLET_SERVE_TRAP_INSTALLED="${_WALLET_SERVE_TRAP_INSTALLED:-0}"
+
+wallet_serve_resolve_bins() {
+  # Prefer explicit WALLET_CLI / CLI_JS (.mjs); else find_wallet_cli from lib.sh
+  if [ -z "${CLI_JS:-}" ] || [ ! -f "${CLI_JS:-}" ]; then
+    if [ -n "${WALLET_CLI:-}" ] && [ -f "${WALLET_CLI}" ]; then
+      CLI_JS="$WALLET_CLI"
+    elif command -v find_wallet_cli >/dev/null 2>&1; then
+      CLI_JS=$(find_wallet_cli 2>/dev/null || true)
+    fi
+  fi
+  if [ -n "${CLI_JS:-}" ] && [ -f "$CLI_JS" ]; then
+    WALLET_CLI="$CLI_JS"
+  fi
+  if [ -z "${NODE_BIN:-}" ] || [ ! -x "${NODE_BIN:-}" ]; then
+    NODE_BIN=$(command -v node 2>/dev/null || true)
+  fi
+  # PATH for taler-helper-sqlite3 (harness/e2e preflight)
+  local _path="${PATH:-}"
+  case ":${_path}:" in
+    *:/opt/homebrew/bin:*) ;;
+    *) _path="/opt/homebrew/bin:/usr/local/bin:${HOME}/.local/bin:${_path}" ;;
+  esac
+  PATH="$_path"
+  export PATH NODE_BIN CLI_JS WALLET_CLI
+}
+
+wallet_serve_stop() {
+  if [ -n "${WSERVE_PID:-}" ] && kill -0 "$WSERVE_PID" 2>/dev/null; then
+    kill "$WSERVE_PID" 2>/dev/null || true
+    wait "$WSERVE_PID" 2>/dev/null || true
+  fi
+  WSERVE_PID=""
+  if [ -n "${WSOCK:-}" ]; then
+    rm -f "$WSOCK" 2>/dev/null || true
+  fi
+}
+
+wallet_serve_start() {
+  # Returns 0 if serve is up (or intentionally disabled); 1 if wanted but failed.
+  wallet_serve_resolve_bins
+  local want="${LADDER_WALLET_SERVE:-1}"
+  # e2e callers may only set E2E_WALLET_SERVE
+  if [ "${E2E_WALLET_SERVE:-}" = "0" ]; then
+    want=0
+  fi
+  if [ "$want" != "1" ]; then
+    return 0
+  fi
+  if [ -z "${NODE_BIN:-}" ] || [ ! -x "${NODE_BIN}" ]; then
+    return 1
+  fi
+  if [ -z "${WALLET_CLI:-}" ] || [ ! -f "${WALLET_CLI}" ]; then
+    return 1
+  fi
+  if [ -z "${WDB:-}" ]; then
+    return 1
+  fi
+  if [ -z "${SCRATCH:-}" ]; then
+    SCRATCH=$(mktemp -d)
+  fi
+  WSOCK="${WSOCK:-$SCRATCH/wallet.sock}"
+  wallet_serve_stop
+  rm -f "$WSOCK"
+  # harness: daemon owns DB; clients use --wallet-connection only (no concurrent --wallet-db)
+  "$NODE_BIN" "$WALLET_CLI" --wallet-db="$WDB" --no-throttle --skip-defaults \
+    advanced serve --unix-path "$WSOCK" \
+    >"${SCRATCH}/wallet-serve.log" 2>&1 &
+  WSERVE_PID=$!
+  local i
+  for i in $(seq 1 50); do
+    if [ -S "$WSOCK" ] && kill -0 "$WSERVE_PID" 2>/dev/null; then
+      return 0
+    fi
+    # died early
+    if ! kill -0 "$WSERVE_PID" 2>/dev/null; then
+      break
+    fi
+    sleep 0.1
+  done
+  wallet_serve_stop
+  return 1
+}
+
+wallet_serve_install_trap() {
+  [ "${_WALLET_SERVE_TRAP_INSTALLED}" = "1" ] && return 0
+  _WALLET_SERVE_TRAP_INSTALLED=1
+  # Append to existing EXIT trap if any — stop serve first
+  # shellcheck disable=SC2064
+  trap 'wallet_serve_stop' EXIT
+}
+
+# Run wallet-cli: prefer live serve socket (harness WalletClient path).
+# Optional first arg: max seconds (integer) → wrapped with timeout when available.
+wallet_serve_wcli() {
+  wallet_serve_resolve_bins
+  local maxc=""
+  if [[ "${1:-}" =~ ^[0-9]+$ ]]; then
+    maxc=$1
+    shift
+  fi
+  local -a cmd
+  if [ -n "${NODE_BIN:-}" ] && [ -x "${NODE_BIN}" ] && [ -n "${WALLET_CLI:-}" ] && [ -f "${WALLET_CLI}" ]; then
+    if [ -n "${WSERVE_PID:-}" ] && [ -S "${WSOCK:-}" ] && kill -0 "$WSERVE_PID" 2>/dev/null; then
+      cmd=( "$NODE_BIN" "$WALLET_CLI" --wallet-connection="$WSOCK" --no-throttle "$@" )
+    else
+      cmd=( "$NODE_BIN" "$WALLET_CLI" --wallet-db="${WDB:?}" --no-throttle --skip-defaults "$@" )
+    fi
+  elif [ -n "${CLI_JS:-}" ] && [ -f "$CLI_JS" ]; then
+    if [ -n "${WSERVE_PID:-}" ] && [ -S "${WSOCK:-}" ]; then
+      cmd=( node "$CLI_JS" --wallet-connection="$WSOCK" --no-throttle "$@" )
+    else
+      cmd=( node "$CLI_JS" --wallet-db="${WDB:?}" --no-throttle "$@" )
+    fi
+  else
+    if [ -n "${WSERVE_PID:-}" ] && [ -S "${WSOCK:-}" ]; then
+      cmd=( taler-wallet-cli --wallet-connection="$WSOCK" --no-throttle "$@" )
+    else
+      cmd=( taler-wallet-cli --wallet-db="${WDB:?}" --no-throttle "$@" )
+    fi
+  fi
+  if [ -n "$maxc" ]; then
+    if command -v timeout >/dev/null 2>&1; then
+      timeout -k 2 "$maxc" "${cmd[@]}"
+    elif command -v gtimeout >/dev/null 2>&1; then
+      gtimeout --kill-after=2 "$maxc" "${cmd[@]}"
+    else
+      "${cmd[@]}"
+    fi
+  else
+    "${cmd[@]}"
+  fi
+}
 @ -1 +1 @@
 .23.7
 .23.8