fix(ci): cancel long rust build subprocesses (#3942)

.github/workflows/pr-ci.yml

@@ -14,6 +14,8 @@ permissions:
 env:
   NPM_CONFIG_STORE_DIR: .pnpm-store
 
+# Keep only the newest run per PR/branch. Long Rust/Tauri build commands must
+# run through scripts/ci-cancel-aware.sh so cancellation reaches descendants.
 concurrency:
   group: ${{ github.workflow }}-${{ github.event.pull_request.number || github.head_ref || github.ref }}
   cancel-in-progress: true
@@ -56,11 +58,9 @@ jobs:
               - 'app/vite.config.*'
               - 'app/tailwind.config.*'
               - 'app/postcss.config.*'
-              - 'scripts/ci-cancel-aware.sh'
             i18n:
               - '.github/workflows/pr-ci.yml'
               - 'app/src/**'
-              - 'scripts/ci-cancel-aware.sh'
             rust-core:
               - '.github/workflows/pr-ci.yml'
               - 'Cargo.toml'
@@ -69,6 +69,7 @@ jobs:
               - 'src/**'
               - 'tests/**'
               - 'scripts/ci-cancel-aware.sh'
+              - 'scripts/test-rust-e2e.sh'
               - 'scripts/test-rust-with-mock.sh'
             rust-tauri:
               - '.github/workflows/pr-ci.yml'
@@ -116,7 +117,7 @@ jobs:
             pnpm-store-${{ runner.os }}-
 
       - name: Install dependencies
-        run: bash scripts/ci-cancel-aware.sh pnpm install --frozen-lockfile
+        run: pnpm install --frozen-lockfile
 
       - name: Save pnpm store cache
         if: always() && steps.pnpm-cache.outputs.cache-hit != 'true'
@@ -126,17 +127,17 @@ jobs:
           key: ${{ steps.pnpm-cache.outputs.cache-primary-key }}
 
       - name: Type check TypeScript files
-        run: bash scripts/ci-cancel-aware.sh pnpm --filter openhuman-app compile
+        run: pnpm --filter openhuman-app compile
         env:
           NODE_ENV: test
 
       - name: Check Prettier formatting
-        run: bash scripts/ci-cancel-aware.sh pnpm --filter openhuman-app format:check
+        run: pnpm --filter openhuman-app format:check
         env:
           NODE_ENV: test
 
       - name: Run ESLint
-        run: bash scripts/ci-cancel-aware.sh pnpm --filter openhuman-app lint
+        run: pnpm --filter openhuman-app lint
         env:
           NODE_ENV: test
 
@@ -164,7 +165,7 @@ jobs:
             pnpm-store-${{ runner.os }}-
 
       - name: Install dependencies
-        run: bash scripts/ci-cancel-aware.sh pnpm install --frozen-lockfile
+        run: pnpm install --frozen-lockfile
 
       - name: Save pnpm store cache
         if: always() && steps.pnpm-cache.outputs.cache-hit != 'true'
@@ -174,7 +175,7 @@ jobs:
           key: ${{ steps.pnpm-cache.outputs.cache-primary-key }}
 
       - name: Verify i18n coverage (missing / extra / drifted keys across locales)
-        run: bash scripts/ci-cancel-aware.sh pnpm i18n:check
+        run: pnpm i18n:check
 
   rust-quality:
     name: Rust Quality (fmt, clippy)
@@ -207,7 +208,7 @@ jobs:
           shared-key: pr-rust-quality
 
       - name: Check Rust formatting
-        run: bash scripts/ci-cancel-aware.sh cargo fmt --all -- --check
+        run: cargo fmt --all -- --check
 
       - name: Run clippy (core crate)
         run: bash scripts/ci-cancel-aware.sh cargo clippy -p openhuman
@@ -258,7 +259,7 @@ jobs:
 
       - name: Install dependencies
         if: steps.playwright-artifact-cache.outputs.cache-hit != 'true'
-        run: bash scripts/ci-cancel-aware.sh pnpm install --frozen-lockfile
+        run: pnpm install --frozen-lockfile
 
       - name: Save pnpm store cache
         if: always() && steps.pnpm-cache.outputs.cache-hit != 'true'
@@ -277,7 +278,7 @@ jobs:
         if: steps.playwright-artifact-cache.outputs.cache-hit != 'true'
         run: |
           set -euo pipefail
-          bash scripts/ci-cancel-aware.sh pnpm --filter openhuman-app test:e2e:web:build
+          pnpm --filter openhuman-app test:e2e:web:build
 
           RUST_HOST_TRIPLE="$(rustc -vV | awk '/^host: / { print $2 }')"
           WEB_CORE_TARGET_DIR="target/e2e-web-${RUST_HOST_TRIPLE}"
@@ -335,7 +336,7 @@ jobs:
             pnpm-store-${{ runner.os }}-
 
       - name: Install dependencies
-        run: bash scripts/ci-cancel-aware.sh pnpm install --frozen-lockfile
+        run: pnpm install --frozen-lockfile
 
       - name: Save pnpm store cache
         if: always() && steps.pnpm-cache.outputs.cache-hit != 'true'
@@ -345,7 +346,7 @@ jobs:
           key: ${{ steps.pnpm-cache.outputs.cache-primary-key }}
 
       - name: Run Vitest with coverage
-        run: bash ../scripts/ci-cancel-aware.sh pnpm test:coverage
+        run: pnpm test:coverage
         working-directory: app
         env:
           NODE_ENV: test
@@ -562,7 +563,7 @@ jobs:
           shared-key: pr-rust-core
 
       - name: Install dependencies
-        run: bash scripts/ci-cancel-aware.sh pnpm install --frozen-lockfile
+        run: pnpm install --frozen-lockfile
 
       - name: Save pnpm store cache
         if: always() && steps.pnpm-cache.outputs.cache-hit != 'true'
@@ -577,7 +578,7 @@ jobs:
           touch app/.env
 
       - name: Run Rust E2E suite
-        run: bash scripts/ci-cancel-aware.sh pnpm test:rust:e2e
+        run: pnpm test:rust:e2e
         env:
           RUST_BACKTRACE: "1"
 
@@ -627,7 +628,7 @@ jobs:
             pnpm-store-${{ runner.os }}-
 
       - name: Install JS dependencies (test harness)
-        run: bash scripts/ci-cancel-aware.sh pnpm install --frozen-lockfile
+        run: pnpm install --frozen-lockfile
 
       - name: Save pnpm store cache
         if: always() && steps.pnpm-cache.outputs.cache-hit != 'true'
@@ -653,14 +654,14 @@ jobs:
           chmod +x target/e2e-web-*/debug/openhuman-core
 
       - name: Install Playwright Chromium headless shell
-        run: bash scripts/ci-cancel-aware.sh pnpm --filter openhuman-app exec playwright install chromium-headless-shell
+        run: pnpm --filter openhuman-app exec playwright install chromium-headless-shell
 
       - name: Run Playwright web E2E suite (shard ${{ matrix.shard }}/4)
         env:
           OPENHUMAN_WORKSPACE: ${{ runner.temp }}/openhuman-playwright-workspace
         run: |
           mkdir -p "$OPENHUMAN_WORKSPACE"
-          bash scripts/ci-cancel-aware.sh bash app/scripts/e2e-web-session.sh \
+          bash app/scripts/e2e-web-session.sh \
             --shard=${{ matrix.shard }}/4
 
       - name: Upload Playwright E2E failure artifacts

app/scripts/e2e-web-build.sh

@@ -24,4 +24,4 @@ fi
 echo "Building web E2E bundle with backend ${VITE_BACKEND_URL}"
 pnpm run build:web
 echo "Building standalone openhuman-core for web E2E into ${E2E_WEB_CORE_TARGET_DIR}..."
-CARGO_TARGET_DIR="$E2E_WEB_CORE_TARGET_DIR" cargo build --manifest-path "$REPO_ROOT/Cargo.toml" --bin openhuman-core
+CARGO_TARGET_DIR="$E2E_WEB_CORE_TARGET_DIR" "$REPO_ROOT/scripts/ci-cancel-aware.sh" cargo build --manifest-path "$REPO_ROOT/Cargo.toml" --bin openhuman-core

scripts/ci-cancel-aware.sh

@@ -10,6 +10,7 @@ fi
 OS_NAME="$(uname -s 2>/dev/null || echo unknown)"
 CHILD_PID=""
 RECEIVED_SIGNAL=""
+CHILD_OWNS_PROCESS_GROUP=0
 
 is_windows_shell() {
   case "$OS_NAME" in
@@ -35,6 +36,10 @@ terminate_tree_term() {
     return
   fi
 
+  if [ "$CHILD_OWNS_PROCESS_GROUP" = "1" ]; then
+    kill -TERM -- "-$pid" 2>/dev/null || true
+  fi
+
   local descendants=""
   descendants="$(collect_descendants_unix "$pid")"
   if [ -n "$descendants" ]; then
@@ -53,6 +58,10 @@ terminate_tree_kill() {
     return
   fi
 
+  if [ "$CHILD_OWNS_PROCESS_GROUP" = "1" ]; then
+    kill -KILL -- "-$pid" 2>/dev/null || true
+  fi
+
   local descendants=""
   descendants="$(collect_descendants_unix "$pid")"
   if [ -n "$descendants" ]; then
@@ -103,19 +112,33 @@ cleanup() {
   return "$status"
 }
 
+start_child() {
+  if ! is_windows_shell && command -v setsid >/dev/null 2>&1; then
+    setsid "$@" &
+    CHILD_OWNS_PROCESS_GROUP=1
+  else
+    "$@" &
+    CHILD_OWNS_PROCESS_GROUP=0
+  fi
+  CHILD_PID=$!
+}
+
 trap 'forward_cancel INT' INT
 trap 'forward_cancel TERM' TERM
 trap 'forward_cancel HUP' HUP
 trap cleanup EXIT
 
 echo "[ci-cancel-aware] exec: $(printf '%q ' "$@")" >&2
-"$@" &
-CHILD_PID=$!
+start_child "$@"
 
 set +e
 wait "$CHILD_PID"
 status=$?
 set -e
 
-CHILD_PID=""
+# A trapped cancellation can interrupt wait before the child exits. Keep
+# CHILD_PID set so the EXIT cleanup can escalate from TERM to KILL.
+if ! kill -0 "$CHILD_PID" 2>/dev/null; then
+  CHILD_PID=""
+fi
 exit "$status"

scripts/test-rust-e2e.sh

@@ -133,9 +133,9 @@ echo "[rust-e2e] Running ${#SUITES[@]} suite(s) serially."
 for suite in "${SUITES[@]}"; do
   if [ "${#EXTRA_ARGS[@]}" -gt 0 ]; then
     echo "[rust-e2e]   cargo test --manifest-path Cargo.toml --test $suite -- ${EXTRA_ARGS[*]}"
-    cargo test --manifest-path Cargo.toml --test "$suite" -- "${EXTRA_ARGS[@]}"
+    "$SCRIPT_DIR/ci-cancel-aware.sh" cargo test --manifest-path Cargo.toml --test "$suite" -- "${EXTRA_ARGS[@]}"
   else
     echo "[rust-e2e]   cargo test --manifest-path Cargo.toml --test $suite"
-    cargo test --manifest-path Cargo.toml --test "$suite"
+    "$SCRIPT_DIR/ci-cancel-aware.sh" cargo test --manifest-path Cargo.toml --test "$suite"
   fi
 done