release 1.13.5: progress bar env-aware estimates and phase re-fit

Stage jobs estimated 139 checks (urls 95 + monpages 20 + phantom
post-check 12 + devtesting) but only ~46 ran, so bars sat at 42/139
then snapped to 46/46. Estimate from CHECK_BANK/LANDING and mon hosts;
after each phase set total = done + remaining; no suite credit for
host-agent monpages post-check outside the process.

VERSION

@@ -1 +1 @@
-1.13.4
+1.13.5

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 |
 |-----|------------|--------|
+| **v1.13.5** | 2026-07-19 | **Bugfix:** progress bar — env-aware phase estimates (CHECK_BANK/LANDING, mon hosts); re-fit total after each phase (done+remaining); drop fake +12 for host-agent monpages post-check (fixes stage 42/139→46/46 snap). |
 | **v1.13.4** | 2026-07-19 | **Bugfix:** monpages race-proof — atomic HTML write; rsync `--delay-updates` + settle; fetch retries; larger head windows; pre-publish soft content (`MONPAGES_PRE_PUBLISH`); bootstrap markers. |
 | **v1.13.3** | 2026-07-19 | **Bugfix:** sticky **generated** time in **Europe/Zurich** (CEST/CET · MESZ/MEZ), not UTC `…Z`; ISO keeps offset for age JS. |
 | **v1.13.2** | 2026-07-19 | **Bugfix:** sticky version link uses **commit** URL (`/src/commit/<sha>`) not `/src/tag/v…` (tags not yet pushed → Forgejo fat 404); tags v1.11–v1.13.1 pushed to origin. |

lib.sh

@@ -574,8 +574,9 @@ set_progress_total() {
   # Optional: expected number of numbered checks in this run.
   # PROGRESS_TOTAL=0 → no percent, only "done=N".
   # Does not reset GLOBAL_N / PROGRESS_DONE if already mid-run (state file).
+  # May shrink when parent re-fits after a short phase (total = done + remaining est),
+  # but never below PROGRESS_DONE.
   local n="${1:-0}"
-  # Never shrink below already completed work (re-estimate mid-run safe)
   if [ "${PROGRESS_DONE:-0}" -gt 0 ] && [ "$n" -gt 0 ] && [ "$n" -lt "$PROGRESS_DONE" ]; then
     n=$PROGRESS_DONE
   fi

taler-monitoring.sh

@@ -396,36 +396,73 @@ if [ "${#PHASES[@]}" -eq 0 ]; then
 fi
 
 # Global expected check count for progress bar done/total (override: PROGRESS_TOTAL=N).
-# One estimate for the whole run — phase summary() must NOT reset this mid-run.
+# Auto estimate is env-aware (CHECK_BANK=0 etc.) and re-fit after each phase so stage
-# Prefer slightly high; lib.sh only grows if short (never shrinks until progress_finish).
+# jobs do not sit at 42/139 then snap to 46/46. Phase summary() must NOT snap mid-run.
+# Host-agent monpages post-check is outside this process — never counted here.
+_progress_estimate_phase() {
+  local p="$1" n=0 hosts=1
+  hosts=$(printf '%s' "${MON_HOSTS:-x}" | wc -w)
+  [ "$hosts" -lt 1 ] && hosts=1
+  case "$p" in
+    urls)
+      # Core: exchange + merchant + webui + perf + terms (~28–40). Bank/landing optional.
+      # Stage (CHECK_BANK=0 CHECK_LANDING=0) lands near ~34 numbered lines — keep close.
+      n=28
+      [ "${CHECK_BANK:-1}" != "0" ] && n=$((n + 24))
+      [ "${CHECK_LANDING:-1}" != "0" ] && n=$((n + 18))  # intro + stats.json + qr
+      n=$((n + 6))  # headroom (SPA / terms / privacy / alt paths)
+      ;;
+    inside)   n=30 ;;
+    versions) n=35 ;;
+    sanity)   n=35 ;;
+    server)   n=20 ;;
+    # e2e: ATM ladder emits many INFO lines (coins before/after each note)
+    e2e)      n=240 ;;
+    ladder|goa-ladder) n=120 ;;
+    auth401)  n=70 ;;
+    aptdeploy) n=40 ;;
+    surface|ecosystem) n=120 ;;
+    monpages|pages)
+      # slash + bare (+ optional _err) per inventory host; GOA full suite ≈20 URLs
+      case "${TALER_DOMAIN:-}" in
+        *hacktivism*) n=24 ;;
+        *lefrancpaysan*|*francpaysan*) n=$((hosts * 2 + 6)); [ "$n" -lt 12 ] && n=12 ;;
+        *) n=$((hosts * 2 + 4)); [ "$n" -lt 8 ] && n=8 ;;
+      esac
+      [ "$n" -gt 48 ] && n=48
+      ;;
+    mattermost) n=25 ;;
+    mail) n=40 ;;
+    devtesting|franken|fake-franken|fake_franken) n=10 ;;
+    *) n=15 ;;
+  esac
+  printf '%s' "$n"
+}
+
+# Recompute PROGRESS_TOTAL = done + estimate(remaining phases). Allows shrink after short phases.
+_progress_refit_remaining() {
+  [ "${PROGRESS_TOTAL_AUTO:-0}" = "1" ] || return 0
+  local rem=0 rp
+  for rp in "$@"; do
+    rem=$((rem + $(_progress_estimate_phase "$rp")))
+  done
+  if type _mon_state_load >/dev/null 2>&1; then
+    _mon_state_load
+  fi
+  set_progress_total "$(( ${PROGRESS_DONE:-0} + rem ))"
+}
+
+PROGRESS_TOTAL_AUTO=0
 if [ "${PROGRESS_TOTAL:-0}" = "0" ] || [ -z "${PROGRESS_TOTAL:-}" ]; then
+  PROGRESS_TOTAL_AUTO=1
   _pt=0
   for p in "${PHASES[@]}"; do
-    case "$p" in
+    _pt=$((_pt + $(_progress_estimate_phase "$p")))
-      urls)     _pt=$((_pt + 95)) ;;   # +qr + webui SPA fingerprints
-      inside)   _pt=$((_pt + 30)) ;;
-      versions) _pt=$((_pt + 35)) ;;   # + disk probes can be many
-      sanity)   _pt=$((_pt + 35)) ;;
-      server)   _pt=$((_pt + 20)) ;;
-      # e2e: ATM ladder emits many INFO lines (coins before/after each note)
-      e2e)      _pt=$((_pt + 240)) ;;
-      ladder|goa-ladder) _pt=$((_pt + 120)) ;;
-      auth401)  _pt=$((_pt + 70)) ;;
-      aptdeploy) _pt=$((_pt + 40)) ;;
-      surface|ecosystem) _pt=$((_pt + 120)) ;;
-      monpages|pages) _pt=$((_pt + 20)) ;;
-      mattermost) _pt=$((_pt + 25)) ;;
-      mail) _pt=$((_pt + 40)) ;;
-      devtesting|franken|fake-franken|fake_franken) _pt=$((_pt + 12)) ;;
-    esac
   done
-  # monpages post-check after htmlify can add another inventory pass
-  case " ${PHASES[*]} " in
-    *" monpages "*|*" pages "*) _pt=$((_pt + 12)) ;;
-  esac
   set_progress_total "$_pt"
   unset _pt
 fi
+export PROGRESS_TOTAL_AUTO
 
 # Whole-run wall clock (default 10 min). Override: RUN_TIMEOUT=0 (unlimited).
 : "${RUN_TIMEOUT:=600}"
@@ -516,9 +553,13 @@ unset _INSIDE_ACCESS_HINT
 chmod +x "$ROOT"/check_*.sh 2>/dev/null || true
 
 ec=0
-for p in "${PHASES[@]}"; do
+_phase_idx=0
+_n_phases=${#PHASES[@]}
+while [ "$_phase_idx" -lt "$_n_phases" ]; do
+  p="${PHASES[$_phase_idx]}"
   if [ "$RUN_TIMED_OUT" = "1" ]; then
     RUN_SKIPPED_PHASES+=("$p")
+    _phase_idx=$((_phase_idx + 1))
     continue
   fi
   left=$(mon_seconds_left)
@@ -526,6 +567,7 @@ for p in "${PHASES[@]}"; do
     RUN_TIMED_OUT=1
     RUN_TIMEOUT_AT_PHASE="${RUN_TIMEOUT_AT_PHASE:-$p}"
     RUN_SKIPPED_PHASES+=("$p")
+    _phase_idx=$((_phase_idx + 1))
     continue
   fi
   case "$p" in
@@ -546,7 +588,21 @@ for p in "${PHASES[@]}"; do
       run_phase devtesting "$ROOT/check_devtesting.sh" || ec=1
       ;;
   esac
+  # After each phase: refit total = actual done + estimate(remaining). Fixes
+  # stage over-estimate (e.g. 42/139 while only ~4 checks left).
+  if [ "${PROGRESS_TOTAL_AUTO:-0}" = "1" ]; then
+    _rem_args=()
+    _j=$((_phase_idx + 1))
+    while [ "$_j" -lt "$_n_phases" ]; do
+      _rem_args+=("${PHASES[$_j]}")
+      _j=$((_j + 1))
+    done
+    _progress_refit_remaining "${_rem_args[@]+"${_rem_args[@]}"}"
+    unset _rem_args _j
+  fi
+  _phase_idx=$((_phase_idx + 1))
 done
+unset _phase_idx _n_phases
 
 # Extraordinary run-budget failure: always report at end; HTML links top → here.
 if [ "$RUN_TIMED_OUT" = "1" ]; then