export-runner: fix new-symbol discovery aborting with a year-0000 backtest

discover_start's stdout is its result channel (read via command
substitution in process_symbol), but the probes inside it stream the
cBot's "[BOT] …" progress on stdout too. That noise leaked into the
captured value, so start_idx parsed to 0 and the resume walk began at
year 0 -> an unparseable "--start=01/01/0" backtest that retried out and
hard-aborted the whole run. Only a brand-new symbol (no files on disk yet)
hits this path, so it stayed invisible until AUDUSD was added.

Redirect every probe call in discover_start to stderr (1>&2) so its
stdout carries only the "YEAR MONTH" result, and document the contract.
Add scripts/test-export-runner.sh: source-level tests that stub probe_run
to emit the same stdout noise and assert a clean result (RED without the
redirect). Make SYMBOL_FILE overridable for scoped runs/tests.

docs/export-runner-quirks.md

@@ -79,3 +79,24 @@ data dir untouched):
   yesterday, and left the trusted middle `03` byte-identical.
 - **Clean-empty safety**: `run_one` on an empty future window returns rc 3 (skip), never the
   rc 2 that would abort the run.
+
+## Post-deploy fix — discovery stdout pollution (year-0000 abort)
+
+The first **new** symbol added after the rewrite (AUDUSD) aborted on the scheduled
+2026-06-25 run with `--start=01/01/0 --end=31/01/0000` ("Value for parameter start can't
+be parsed"). Root cause: a function-output contract violation, not bad data.
+
+`process_symbol` resolves a brand-new symbol's first month via `sd=$(discover_start …)` —
+command substitution, so `discover_start`'s **stdout is its result channel**. But the
+probes inside it (`probe_run` → `run_one`) stream the cBot's `[BOT] …` progress on stdout
+too. Those lines leaked into `sd`; `start_idx=$(midx ${sd% *} ${sd#* })` then parsed
+garbage and collapsed to `0` → the walk started at year 0, month 1 → an unparseable
+`01/01/0` backtest. (The abort was correct behaviour: it stopped a ~24 000-iteration walk
+from year 0 to now, and nothing was written to the data dir — the scrape failed before any
+file op.) Only the no-existing-files branch calls `discover_start`, so only a *new* symbol
+could trigger it — invisible until AUDUSD.
+
+**Fix**: every probe call inside `discover_start` is redirected `1>&2`, so its stdout
+carries only the result line. A documented CONTRACT comment now guards the function.
+Covered by `scripts/test-export-runner.sh` (source-level; stubs `probe_run` to emit the
+same stdout noise and asserts a clean `YEAR MONTH` result — RED without the redirect).

scripts/export-runner.sh

@@ -31,7 +31,7 @@ BOT_FILE="MS_SimpleExport.algo"
 PWD_FILE="/mnt/Robots/password.txt"          # container-internal path (console reads it)
 IMAGE="ghcr.io/spotware/ctrader-console:latest"
 HOST_ROBOTS_PATH="$REPO_ROOT/runtime/Robots" # holds MS_SimpleExport.algo + password.txt
-SYMBOL_FILE="$REPO_ROOT/runtime/ScrapeSymbols.txt"
+SYMBOL_FILE="${SYMBOL_FILE:-$REPO_ROOT/runtime/ScrapeSymbols.txt}"  # overridable for a scoped run/test
 RETRIES="${RETRIES:-3}"                       # attempts before a transient failure aborts
 RUN_TIMEOUT="${RUN_TIMEOUT:-600}"             # seconds per backtest before it counts as a hang
 DISCOVER_MIN_YEAR="${DISCOVER_MIN_YEAR:-2005}" # earliest year the first-data probe scans
@@ -211,12 +211,18 @@ probe_run() {
 # nothing if the symbol has no data at all. Yearly probe uses December (catches data
 # that started any time that year); monthly probe uses the FULL month (catches data
 # starting after the 10th — the old probe only tested days 01–10).
+#
+# CONTRACT: this function's STDOUT carries ONLY the result line; the caller reads it via
+# command substitution. probe_run -> run_one stream the cBot's "[BOT] …" lines on stdout,
+# so every probe call here MUST be redirected to stderr (`1>&2`) — otherwise that progress
+# noise pollutes the captured value and start_idx is parsed from garbage (see the AUDUSD
+# year-0000 abort, docs/export-runner-quirks.md). All diagnostics here go to stderr too.
 discover_start() {
     local S=$1 MODE=$2 y mm MM edate found_year=""
     echo "🌑 Discovery: locating first data month for $S ($MODE)..." >&2
     for ((y=DISCOVER_MIN_YEAR; y<CURRENT_YEAR; y++)); do
         echo "  🧪 year $y" >&2
-        if probe_run "$S" "$MODE" "01/12/$y" "31/12/$y"; then found_year=$y; break; fi
+        if probe_run "$S" "$MODE" "01/12/$y" "31/12/$y" 1>&2; then found_year=$y; break; fi
     done
     if [[ -z "$found_year" ]]; then found_year=$CURRENT_YEAR; fi   # brand-new this year → check months below
     for ((mm=1; mm<=12; mm++)); do
@@ -225,7 +231,7 @@ discover_start() {
         edate="$(month_end_or_clamp "$found_year" "$mm")"
         [[ -z "$edate" ]] && break
         echo "  🧪 month $found_year-$MM" >&2
-        if probe_run "$S" "$MODE" "01/$MM/$found_year" "$edate"; then echo "$found_year $mm"; return 0; fi
+        if probe_run "$S" "$MODE" "01/$MM/$found_year" "$edate" 1>&2; then echo "$found_year $mm"; return 0; fi
     done
     return 1   # no data anywhere
 }

scripts/test-export-runner.sh

@@ -0,0 +1,76 @@
+#!/usr/bin/env bash
+#
+# Source-level unit tests for export-runner.sh. The runner is written to be sourceable
+# (main runs only when executed directly), so we can load its functions, stub the parts
+# that would spawn containers, and assert pure logic. No Docker, no broker, no data dir.
+#
+# Run: scripts/test-export-runner.sh   (exit 0 = all pass)
+set -uo pipefail
+
+HERE="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+RUNNER="$HERE/export-runner.sh"
+
+# The runner sources runtime/export.env and hard-requires HOST_DATA_PATH/CTID/ACCOUNT.
+# Provide throwaway values so sourcing succeeds without the real deployment config.
+export ENV_FILE; ENV_FILE="$(mktemp)"
+cat > "$ENV_FILE" <<'ENV'
+HOST_DATA_PATH=/nonexistent
+CTID=test@example.com
+ACCOUNT=0
+ENV
+trap 'rm -f "$ENV_FILE"' EXIT
+
+# shellcheck disable=SC1090
+source "$RUNNER"
+
+PASS=0 FAIL=0
+check() { # check NAME EXPECTED ACTUAL
+    if [[ "$2" == "$3" ]]; then echo "  ✓ $1"; PASS=$((PASS+1))
+    else echo "  ✗ $1 — expected [$2], got [$3]"; FAIL=$((FAIL+1)); fi
+}
+
+# --- discover_start contamination regression -------------------------------------------
+# run_one streams "[BOT] …" progress on STDOUT; discover_start's STDOUT is its result
+# channel (read via command substitution). A stub that mimics that noise must NOT leak
+# into the result. Pre-fix this returned garbage like "0-000-…" -> start_idx 0 -> the
+# year-0000 "--start=01/01/0" abort. CURRENT_YEAR/MONTH come from the host clock; pick a
+# data-start strictly before today so the assertion is stable regardless of the date.
+emit_noise() { echo "  [BOT] Progress 50% | bla"; echo "  [BOT] | Info | noise"; }
+
+# Scenario 1: data starts in a past December (yearly probe hits), Jan of that year empty,
+# Feb has data -> expect "<year> 2".
+TARGET_YEAR=$(( CURRENT_YEAR - 1 ))
+probe_run() { emit_noise
+    case "$3" in
+        "01/12/$TARGET_YEAR") return 0 ;;   # yearly probe: this December has data
+        "01/02/$TARGET_YEAR") return 0 ;;   # monthly probe: February is the first month
+        *) return 1 ;;
+    esac
+}
+got="$(discover_start FOO M1)"
+check "discover_start: result is clean (no [BOT] leakage)" "$TARGET_YEAR 2" "$got"
+
+# Scenario 2: no data in any past year, first data in an early month of the current year.
+probe_run() { emit_noise
+    [[ "$3" == "01/0$(( CURRENT_MONTH > 1 ? CURRENT_MONTH - 1 : CURRENT_MONTH ))/$CURRENT_YEAR" ]] && return 0
+    return 1
+}
+got="$(discover_start FOO M1)"
+# first month of the current year that the stub reports data for:
+want_m=$(( CURRENT_MONTH > 1 ? CURRENT_MONTH - 1 : CURRENT_MONTH ))
+check "discover_start: current-year fallback, clean result" "$CURRENT_YEAR $want_m" "$got"
+
+# Scenario 3: genuinely no data anywhere -> empty stdout, rc 1.
+probe_run() { emit_noise; return 1; }
+got="$(discover_start FOO M1)"; rc=$?
+check "discover_start: no data -> empty result" "" "$got"
+check "discover_start: no data -> rc 1" "1" "$rc"
+
+# --- month index round-trip (sanity on the arithmetic the resume loop depends on) ------
+i=$(midx 2026 5)
+check "midx/idx_year round-trip"  "2026" "$(idx_year "$i")"
+check "midx/idx_month round-trip" "5"    "$(idx_month "$i")"
+
+echo "-----"
+echo "$PASS passed, $FAIL failed"
+[[ $FAIL -eq 0 ]]