Remove the silent two-library default from trigger-add-submodules.sh (#73)

* remove the silent two-library default from trigger-add-submodules.sh

* fix the coderabbitai review

* fix the first reviewer comments

* fix the second reviewer comments

Author: whisper67265

.github/workflows/assets/lib.sh

@@ -2,8 +2,10 @@
 # Shared shell library for add-submodules and start-translation workflows.
 # Source env.sh before lib.sh so ORG, MODULE_ORG, BOT_NAME, BOT_EMAIL, BOOST_ORG, MASTER_BRANCH,
 # LOCAL_BRANCH_PREFIX, TRANSLATION_BRANCH_PREFIX, WEBLATE_ENDPOINT_PATH, and TRANSLATIONS_REPO
-# are set. Workflows also set GITHUB_TOKEN, LANG_CODES, and (for start-translation)
-# WEBLATE_URL / WEBLATE_TOKEN in the step env before sourcing.
+# are set. Workflows also set GITHUB_TOKEN (from secrets.SYNC_TOKEN — broad PAT),
+# LANG_CODES, and (for start-translation) WEBLATE_URL / WEBLATE_TOKEN in the step
+# env before sourcing. Some workflows set GH_TOKEN to the ephemeral Actions token
+# (e.g. heartbeat.yml for gh CLI); that is not the same privilege level as GITHUB_TOKEN.
 # Call validate_secrets (or validate_secrets weblate) after sourcing env.sh and lib.sh.
 #
 # Per-submodule batch return convention (see docs/ARCHITECTURE.md §6):
@@ -576,7 +578,7 @@ validate_secrets() {
   local require_weblate=0
   [[ "${1:-}" == "weblate" ]] && require_weblate=1
 
-  _require_nonempty GITHUB_TOKEN "SYNC_TOKEN secret is not set."
+  _require_nonempty GITHUB_TOKEN "SYNC_TOKEN secret is not set (mapped to GITHUB_TOKEN in workflow env)."
   if [[ -n "${LANG_CODE:-}" ]]; then
     validate_lang_codes "$LANG_CODE"
   elif [[ -n "${LANG_CODES:-}" ]]; then

.github/workflows/heartbeat.yml

@@ -30,6 +30,7 @@ jobs:
 
       - name: Check the last successful scheduled sync
         env:
+          # Ephemeral Actions token for gh CLI — not the SYNC_TOKEN PAT in GITHUB_TOKEN elsewhere.
           GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
         run: |
           set -euo pipefail

CHANGELOG.md

@@ -36,9 +36,17 @@ are a separate namespace — see [README](README.md#releases) and
   per-step verification and create-tag coverage.
 - [endpoint-contract.md](docs/endpoint-contract.md) Outbound Weblate section:
   request schema is now the source of truth for payload fields.
+- `scripts/trigger-add-submodules.sh` now requires `--submodules`; the
+  `unordered, json` script default is removed so the operator script cannot
+  silently diverge from raw API auto-discovery.
+- Bats coverage for `scripts/trigger-dispatch-common.sh`, `trigger-add-submodules.sh`,
+  and `trigger-start-translation.sh`.
 
 ### Fixed
 
+- `build_dispatch_json` in `trigger-dispatch-common.sh`: pass key/value pairs
+  positionally via jq `--args` and `$ARGS.positional` instead of building an
+  intermediate array.
 - Clone and finalize steps now fail fast: **`clone_repo`**, **`sync_translations_branch`**,
   and **`finalize_translations_*`** propagate non-zero status instead of continuing with
   incomplete state.

README.md

@@ -192,7 +192,10 @@ These wrappers use only **`exit 0`** (success, including **`--help`**) and **`ex
 Copy **`.env.example`** to **`.env`** and set **`GH_TOKEN`** (or **`GITHUB_TOKEN`**)
 with permission to call **`POST /repos/{owner}/{repo}/dispatches`** on the target
 repo. The workflows still use GitHub **secrets** and **variables** on the server as
-documented below.
+documented below. **Workflow jobs** map tokens differently: **`SYNC_TOKEN`** (broad
+PAT) is set as **`GITHUB_TOKEN`**; the ephemeral Actions token used by **`gh`** (e.g.
+in **`heartbeat.yml`**) is set as **`GH_TOKEN`**. Those names are interchangeable
+only in the local trigger scripts, not across client and workflow contexts.
 
 ---
 

docs/GETTING-STARTED.md

@@ -66,7 +66,8 @@ linked README sections for scope, format, and which workflows consume each value
 
 ## 1. Local trigger setup (optional)
 
-To fire dispatches from a clone of this repo instead of the GitHub API or UI:
+To fire dispatches from a clone of this repo instead of the GitHub API
+(`POST …/dispatches` or `gh api`):
 
 ```bash
 cp .env.example .env   # set GH_TOKEN (GITHUB_TOKEN is also accepted)
@@ -75,6 +76,10 @@ cp .env.example .env   # set GH_TOKEN (GITHUB_TOKEN is also accepted)
 - **`GH_TOKEN`** is **client-side only** — permission to call
   `POST /repos/{owner}/{repo}/dispatches`. Workflows still use the GitHub
   **secrets** from step 0 on the server.
+- In **workflow jobs**, naming is inverted relative to the client scripts:
+  the broad **`SYNC_TOKEN`** PAT is exported as **`GITHUB_TOKEN`**, while the
+  ephemeral Actions token (e.g. for `gh` in **`heartbeat.yml`**) is exported as
+  **`GH_TOKEN`**. The two names are interchangeable only on the client side.
 - Requires **curl** and **jq** or Python 3.
 
 See [README § Scripts](../README.md#scripts-local-repository_dispatch) for script
@@ -117,14 +122,12 @@ scripts/trigger-add-submodules.sh \
 ```
 
 - Omit **`--lang-codes`** to use repository variable **`LANG_CODES`**.
-- Omit **`--submodules`** to use the script default **`DEFAULT_SUBMODULES`**
-  (`unordered, json`; see
-  [trigger-add-submodules.sh](../scripts/trigger-add-submodules.sh#L40)) — **not**
-  auto-discovery from **`boostorg/boost`**. The script substitutes that default
-  before building the payload
-  ([`SUBMODULES` assignment](../scripts/trigger-add-submodules.sh#L99)), so `client_payload.submodules`
-  is never omitted. Full discovery runs only when the workflow receives a dispatch
-  **without** a `submodules` field (raw API / GitHub UI).
+- **`--submodules` is required** when using this script (comma-separated library
+  names). To process all Boost libraries, omit `submodules` from a **raw**
+  `repository_dispatch` payload sent via the GitHub API (for example
+  `POST …/dispatches` or `gh api repos/{owner}/{repo}/dispatches`) — the workflow
+  then auto-discovers the full list from **`boostorg/boost`** `.gitmodules` via
+  **`resolve_add_submodules_names`**.
 
 ### Expected outcome
 
@@ -238,7 +241,7 @@ Runs automatically **daily at 00:00 UTC** (`0 0 * * *`) or on manual dispatch.
 | Item     | Value                                                                 |
 | -------- | --------------------------------------------------------------------- |
 | Workflow | [`.github/workflows/sync-translation.yml`](../.github/workflows/sync-translation.yml) |
-| Script   | **None** — use the dispatches API, GitHub UI, or the curl example below |
+| Script   | **None** — use the dispatches API or the curl example below |
 | Trigger  | `repository_dispatch` with `event_type: sync-translation` (no `client_payload`) |
 
 ### Trigger (JSON)

docs/endpoint-contract.md

@@ -101,7 +101,7 @@ submodule pointer updates, and Boost release refs in `client_payload.version`.
 | Workflow         | `.github/workflows/add-submodules.yml`                                                                                |
 | Body shape       | `{"event_type":"add-submodules","client_payload":{...}}`                                                              |
 | `client_payload` | All optional: `version`, `submodules` (list-like string), `lang_codes` (comma-separated). See [README](../README.md). |
-| Script           | `scripts/trigger-add-submodules.sh` builds JSON with `jq` or Python; omits empty optional fields. Both trigger scripts source shared [`scripts/trigger-dispatch-common.sh`](../scripts/trigger-dispatch-common.sh). |
+| Script           | `scripts/trigger-add-submodules.sh` builds JSON with `jq` or Python; omits empty optional fields; **`--submodules` is required** on the script even though the API field is optional. Both trigger scripts source shared [`scripts/trigger-dispatch-common.sh`](../scripts/trigger-dispatch-common.sh). |
 
 ### `event_type: start-translation`
 

scripts/trigger-add-submodules.sh

@@ -12,7 +12,7 @@
 #
 # Usage:
 #   scripts/trigger-add-submodules.sh [--repo OWNER/NAME] [--token PAT] \
-#     [--version REF] [--submodules 'a, b'] [--lang-codes zh_Hans,ja]
+#     --submodules 'a, b' [--version REF] [--lang-codes zh_Hans,ja]
 #
 # If --repo is omitted: GITHUB_REPOSITORY, then git origin, then DEFAULT_REPO below.
 
@@ -34,18 +34,13 @@ source "$_ASSETS_DIR/lib.sh"
 source "$_REPO_ROOT/scripts/trigger-dispatch-common.sh"
 unset _REPO_ROOT _ASSETS_DIR
 
-# ---------------------------------------------------------------------------
-# Typical run — edit these. CLI flags override (except --token uses env/PAT).
-# ---------------------------------------------------------------------------
-DEFAULT_SUBMODULES="unordered, json"
-
 usage() {
   cat <<'EOF'
 Trigger add-submodules.yml via repository_dispatch (POST .../dispatches).
 
 Usage:
   scripts/trigger-add-submodules.sh [--repo OWNER/NAME] [--token PAT] \
-    [--version REF] [--submodules 'a, b'] [--lang-codes zh_Hans,ja]
+    --submodules 'a, b' [--version REF] [--lang-codes zh_Hans,ja]
 
 Requires: curl; jq or Python 3 (python3 / python)
 Auth: .env (GH_TOKEN), GH_TOKEN / GITHUB_TOKEN in env, or --token (needs repo scope on the target).
@@ -54,7 +49,7 @@ Options:
   --repo OWNER/REPO     Target repository (default: GITHUB_REPOSITORY, then origin, then DEFAULT_REPO)
   --token PAT           GitHub token
   --version REF         Boost ref; default DEFAULT_VERSION in script
-  --submodules LIST     default DEFAULT_SUBMODULES in script
+  --submodules LIST     comma-separated library names (required)
   --lang-codes CSV      optional; omit → workflow uses repo vars.LANG_CODES
 EOF
 }
@@ -65,18 +60,22 @@ VERSION=""
 SUBMODULES=""
 LANG_CODES=""
 
+shift_pair_into() {
+  printf -v "$1" '%s' "${2:-}"
+}
+
 while [[ $# -gt 0 ]]; do
   case "$1" in
     --repo)
-      REPO="${2:-}"; shift 2 || exit 1 ;;
+      shift_pair_into REPO "$2"; shift 2 || exit 1 ;;
     --token)
-      TOKEN="${2:-}"; shift 2 || exit 1 ;;
+      shift_pair_into TOKEN "$2"; shift 2 || exit 1 ;;
     --version)
-      VERSION="${2:-}"; shift 2 || exit 1 ;;
+      shift_pair_into VERSION "$2"; shift 2 || exit 1 ;;
     --submodules)
-      SUBMODULES="${2:-}"; shift 2 || exit 1 ;;
+      shift_pair_into SUBMODULES "$2"; shift 2 || exit 1 ;;
     --lang-codes)
-      LANG_CODES="${2:-}"; shift 2 || exit 1 ;;
+      shift_pair_into LANG_CODES "$2"; shift 2 || exit 1 ;;
     -h|--help)
       usage; exit 0 ;;
     *)
@@ -86,6 +85,16 @@ while [[ $# -gt 0 ]]; do
   esac
 done
 
+# Trim leading/trailing whitespace; reject empty or whitespace-only values.
+SUBMODULES="${SUBMODULES#"${SUBMODULES%%[![:space:]]*}"}"
+SUBMODULES="${SUBMODULES%"${SUBMODULES##*[![:space:]]}"}"
+
+if [[ -z "$SUBMODULES" ]]; then
+  echo "error: --submodules is required (comma-separated library names)" >&2
+  usage >&2
+  exit 1
+fi
+
 require_curl || exit 1
 
 TOKEN="$(resolve_trigger_token "$TOKEN")" || exit 1
@@ -96,7 +105,6 @@ REPO="$(resolve_trigger_repo "$REPO")" || {
 }
 
 VERSION="${VERSION:-$DEFAULT_VERSION}"
-SUBMODULES="${SUBMODULES:-$DEFAULT_SUBMODULES}"
 
 validate_event_type "$EVENT_ADD_SUBMODULES"
 

scripts/trigger-dispatch-common.sh

@@ -8,9 +8,11 @@ DEFAULT_REPO="cppalliance/boost-docs-translation"
 DEFAULT_VERSION="boost-1.90.0"
 
 infer_repo_from_git() {
-  local url root o r
-  root="$(git rev-parse --show-toplevel 2>/dev/null)" || return 1
-  url="$(git -C "$root" remote get-url origin 2>/dev/null)" || return 1
+  local url="${1:-}" root o r
+  if [[ -z "$url" ]]; then
+    root="$(git rev-parse --show-toplevel 2>/dev/null)" || return 1
+    url="$(git -C "$root" remote get-url origin 2>/dev/null)" || return 1
+  fi
   if [[ "$url" =~ github\.com[:/]([^/]+)/([^[:space:]]+) ]]; then
     o="${BASH_REMATCH[1]}"
     r="${BASH_REMATCH[2]}"
@@ -44,6 +46,9 @@ resolve_trigger_repo() {
   return 1
 }
 
+# Client-side dispatch auth: prefers GH_TOKEN, then GITHUB_TOKEN. Workflow jobs use a
+# different mapping — SYNC_TOKEN PAT in GITHUB_TOKEN; ephemeral Actions token in GH_TOKEN
+# (e.g. heartbeat.yml). The names are not interchangeable across those two contexts.
 resolve_trigger_token() {
   local explicit="${1:-}"
   local token="${explicit:-${GH_TOKEN:-${GITHUB_TOKEN:-}}}"
@@ -67,26 +72,21 @@ build_dispatch_json() {
   local event_type="$1"
   shift
   if command -v jq >/dev/null 2>&1; then
-    local json_pairs='[]' key val
-    while [[ $# -gt 0 ]]; do
-      key="$1"
-      val="$2"
-      shift 2
-      json_pairs="$(jq -n --argjson arr "$json_pairs" --arg k "$key" --arg v "$val" \
-        '$arr + [{key: $k, value: $v}]')"
-    done
-    jq -n --arg event_type "$event_type" --argjson pairs "$json_pairs" \
+    jq -n --arg event_type "$event_type" --args \
       '{
         event_type: $event_type,
         client_payload: (
-          {}
-          | reduce pairs[] as $p (
-              .;
-              if ($p.value | length) > 0 then . + {($p.key): $p.value} else . end
-            )
+          reduce range(0; $ARGS.positional | length; 2) as $i (
+            {};
+            if ($ARGS.positional[$i + 1] | length) > 0
+            then . + {($ARGS.positional[$i]): $ARGS.positional[$i + 1]}
+            else .
+            end
+          )
         )
-      }'
-    return 0
+      }' \
+      "$@"
+    return $?
   fi
   local py=""
   command -v python3 >/dev/null 2>&1 && py="python3"
@@ -102,7 +102,7 @@ for i in range(0,len(pairs),2):
         d[k]=v
 print(json.dumps({"event_type":et,"client_payload":d}))' \
       "$event_type" "$@"
-    return 0
+    return $?
   fi
   return 1
 }