refactor(e2e): simplify suite runner

Signed-off-by: Drew Newberry <anewberry@nvidia.com>

e2e/run.sh

@@ -90,22 +90,15 @@ catalog_has_entry() {
 }
 
 vm=
-vm_set=0
 gateway_config=
-gateway_config_set=0
 suite_name=
-suite_set=0
 with_configurations=()
 
 while [ "$#" -gt 0 ]; do
 	case "$1" in
 	--vm)
 		require_value "$1" "$#" "${2:-}"
-		if [ "${vm_set}" -eq 1 ]; then
-			die "--vm may be supplied only once"
-		fi
 		vm=$2
-		vm_set=1
 		shift 2
 		;;
 	--with)
@@ -115,20 +108,12 @@ while [ "$#" -gt 0 ]; do
 		;;
 	--gateway-config)
 		require_value "$1" "$#" "${2:-}"
-		if [ "${gateway_config_set}" -eq 1 ]; then
-			die "--gateway-config may be supplied only once"
-		fi
 		gateway_config=$2
-		gateway_config_set=1
 		shift 2
 		;;
 	--suite)
 		require_value "$1" "$#" "${2:-}"
-		if [ "${suite_set}" -eq 1 ]; then
-			die "--suite may be supplied only once"
-		fi
 		suite_name=$2
-		suite_set=1
 		shift 2
 		;;
 	-h | --help)
@@ -141,10 +126,10 @@ while [ "$#" -gt 0 ]; do
 	esac
 done
 
-if [ "${gateway_config_set}" -ne 1 ]; then
+if [ -z "${gateway_config}" ]; then
 	die "--gateway-config is required"
 fi
-if [ "${suite_set}" -ne 1 ]; then
+if [ -z "${suite_name}" ]; then
 	die "--suite is required"
 fi
 if ! command -v python3 >/dev/null 2>&1; then
@@ -162,9 +147,9 @@ if [ ! -f "${suite_path}" ]; then
 	die "unknown suite: ${suite_name}"
 fi
 mode=host
-if [ "${vm_set}" -eq 1 ] || [ "${#with_configurations[@]}" -gt 0 ]; then
+if [ -n "${vm}" ] || [ "${#with_configurations[@]}" -gt 0 ]; then
 	mode=vm
-	if [ "${vm_set}" -eq 0 ]; then
+	if [ -z "${vm}" ]; then
 		vm=ubuntu
 	fi
 fi
@@ -180,6 +165,9 @@ if [ "${mode}" = vm ]; then
 	if ! command -v nix >/dev/null 2>&1; then
 		die "Nix is required for VM mode"
 	fi
+	if ! command -v base64 >/dev/null 2>&1; then
+		die "base64 is required for VM mode"
+	fi
 	if ! vm_catalog="$(cd "${ROOT}" && nix run .#test-guest -- --list)"; then
 		die "failed to read the Nix test-guest catalog"
 	fi
@@ -226,10 +214,9 @@ if [ -n "${CARGO_BUILD_JOBS:-}" ]; then
 fi
 
 cd "${ROOT}"
-target_dir="$(
-	mise x -- cargo metadata --format-version=1 --no-deps |
-		python3 -c 'import json, sys; print(json.load(sys.stdin)["target_directory"])'
-)"
+target_dir="$(e2e_cargo_target_dir "${ROOT}")"
+
+ensure_build_nofile_limit
 
 echo "==> Building native host openshell CLI"
 mise x -- cargo build "${cargo_jobs[@]}" -p openshell-cli --bin openshell
@@ -237,7 +224,6 @@ host_cli_bin="${target_dir}/debug/openshell"
 
 echo "==> Preparing ${linux_musl_target} build target"
 mise x -- rustup target add "${linux_musl_target}" >/dev/null
-ensure_build_nofile_limit
 
 echo "==> Building Linux openshell-sandbox (${linux_musl_target})"
 mise x -- cargo zigbuild "${cargo_jobs[@]}" \
@@ -248,7 +234,6 @@ mise x -- cargo zigbuild "${cargo_jobs[@]}" \
 linux_sandbox_bin="${target_dir}/${linux_musl_target}/release/openshell-sandbox"
 
 host_gateway_bin=
-guest_cli_bin=
 guest_gateway_bin=
 if [ "${mode}" = host ]; then
 	echo "==> Building native host openshell-gateway"
@@ -258,17 +243,6 @@ if [ "${mode}" = host ]; then
 		--features bundled-z3
 	host_gateway_bin="${target_dir}/debug/openshell-gateway"
 else
-	if ! command -v base64 >/dev/null 2>&1; then
-		die "base64 is required for VM mode"
-	fi
-	echo "==> Building Linux openshell CLI (${linux_musl_target})"
-	CXXSTDLIB=c++ mise x -- cargo zigbuild "${cargo_jobs[@]}" \
-		--release \
-		--target "${linux_musl_target}" \
-		-p openshell-cli \
-		--bin openshell
-	guest_cli_bin="${target_dir}/${linux_musl_target}/release/openshell"
-
 	echo "==> Preparing ${linux_gateway_rust_target} build target"
 	mise x -- rustup target add "${linux_gateway_rust_target}" >/dev/null
 	echo "==> Building Linux openshell-gateway (${linux_gateway_zig_target})"
@@ -289,32 +263,25 @@ else
 	guest_gateway_bin="${target_dir}/${linux_gateway_rust_target}/release/openshell-gateway"
 fi
 
-for binary in "${host_cli_bin}" "${linux_sandbox_bin}"; do
+expected_binaries=("${host_cli_bin}" "${linux_sandbox_bin}")
+if [ "${mode}" = host ]; then
+	expected_binaries+=("${host_gateway_bin}")
+else
+	expected_binaries+=("${guest_gateway_bin}")
+fi
+for binary in "${expected_binaries[@]}"; do
 	if [ ! -x "${binary}" ]; then
 		echo "ERROR: expected built binary at ${binary}" >&2
 		exit 1
 	fi
 done
-if [ "${mode}" = host ] && [ ! -x "${host_gateway_bin}" ]; then
-	echo "ERROR: expected built gateway at ${host_gateway_bin}" >&2
-	exit 1
-fi
-if [ "${mode}" = vm ]; then
-	for binary in "${guest_cli_bin}" "${guest_gateway_bin}"; do
-		if [ ! -x "${binary}" ]; then
-			echo "ERROR: expected built guest binary at ${binary}" >&2
-			exit 1
-		fi
-	done
-fi
 
 run_parent="${ROOT}/.cache/openshell-e2e/runs"
 mkdir -p "${run_parent}"
 run_dir="$(mktemp -d "${run_parent%/}/run.XXXXXX")"
 child_pid=
-child_process_group=0
 runtime_log=
-portable_jwt_dir="${ROOT}/.cache/openshell-e2e/gateway-jwt"
+portable_jwt_dir=
 keep=0
 if [ "${OPENSHELL_E2E_KEEP:-0}" = 1 ]; then
 	keep=1
@@ -324,15 +291,11 @@ fi
 # shellcheck disable=SC2329
 stop_child() {
 	local pid=$1
-	local process_group=$2
-	local signal_target=${pid}
+	local signal_target="-${pid}"
 
 	if [ -z "${pid}" ] || ! kill -0 "${pid}" 2>/dev/null; then
 		return
 	fi
-	if [ "${process_group}" -eq 1 ]; then
-		signal_target="-${pid}"
-	fi
 	kill -TERM -- "${signal_target}" 2>/dev/null || true
 	for _ in $(seq 1 30); do
 		if ! kill -0 "${pid}" 2>/dev/null; then
@@ -349,24 +312,25 @@ stop_child() {
 # Invoked by EXIT, INT, and TERM traps.
 # shellcheck disable=SC2329
 cleanup() {
-	local incoming_status=$?
-	local status=${1:-${incoming_status}}
+	local status=$?
 
 	trap - EXIT INT TERM
-	stop_child "${child_pid}" "${child_process_group}"
+	stop_child "${child_pid}"
 	if [ "${status}" -ne 0 ] && [ -n "${runtime_log}" ] && [ -f "${runtime_log}" ]; then
 		echo "=== ${mode} gateway log ===" >&2
 		cat "${runtime_log}" >&2
 		echo "=== end ${mode} gateway log ===" >&2
 	fi
 	if [ "${keep}" -eq 1 ]; then
-		echo "Kept E2E wrapper state at ${run_dir}" >&2
+		echo "Kept E2E runner state at ${run_dir}" >&2
 	else
-		rm -f \
-			"${portable_jwt_dir}/signing.pem" \
-			"${portable_jwt_dir}/public.pem" \
-			"${portable_jwt_dir}/kid"
-		rmdir "${portable_jwt_dir}" 2>/dev/null || true
+		if [ -n "${portable_jwt_dir}" ]; then
+			rm -f \
+				"${portable_jwt_dir}/signing.pem" \
+				"${portable_jwt_dir}/public.pem" \
+				"${portable_jwt_dir}/kid"
+			rmdir "${portable_jwt_dir}" 2>/dev/null || true
+		fi
 		rm -rf "${run_dir}"
 	fi
 	exit "${status}"
@@ -378,12 +342,15 @@ trap 'exit 143' TERM
 
 jwt_source_dir="${run_dir}/gateway-jwt"
 e2e_generate_gateway_jwt "${jwt_source_dir}"
-mkdir -p "${portable_jwt_dir}"
-install -m 0600 "${jwt_source_dir}/signing.pem" "${portable_jwt_dir}/signing.pem"
-install -m 0600 "${jwt_source_dir}/public.pem" "${portable_jwt_dir}/public.pem"
-install -m 0600 "${jwt_source_dir}/kid" "${portable_jwt_dir}/kid"
+if [ "${mode}" = host ]; then
+	portable_jwt_dir="${ROOT}/.cache/openshell-e2e/gateway-jwt"
+	mkdir -p "${portable_jwt_dir}"
+	install -m 0600 "${jwt_source_dir}/signing.pem" "${portable_jwt_dir}/signing.pem"
+	install -m 0600 "${jwt_source_dir}/public.pem" "${portable_jwt_dir}/public.pem"
+	install -m 0600 "${jwt_source_dir}/kid" "${portable_jwt_dir}/kid"
+fi
 
-host_port="$(python3 -c 'import socket; s=socket.socket(); s.bind(("127.0.0.1", 0)); print(s.getsockname()[1]); s.close()')"
+host_port="$(e2e_pick_port)"
 guest_port=
 if [ "${mode}" = vm ]; then
 	guest_port=8080
@@ -396,45 +363,33 @@ mkdir -p "${XDG_CONFIG_HOME}" "${XDG_DATA_HOME}" "${XDG_STATE_HOME}"
 
 gateway_name="openshell-e2e-${mode}-${host_port}"
 gateway_endpoint="http://127.0.0.1:${host_port}"
-e2e_register_plaintext_gateway \
-	"${XDG_CONFIG_HOME}" \
-	"${gateway_name}" \
-	"${gateway_endpoint}" \
-	"${host_port}"
-
-with_contract="${with_configurations[*]}"
 export OPENSHELL_GATEWAY_ENDPOINT="${gateway_endpoint}"
 export OPENSHELL_GATEWAY="${gateway_name}"
 export OPENSHELL_BIN="${host_cli_bin}"
-export OPENSHELL_E2E_MODE="${mode}"
-export OPENSHELL_E2E_VM="${vm}"
-export OPENSHELL_E2E_WITH="${with_contract}"
-export OPENSHELL_E2E_GATEWAY_CONFIG="${gateway_config}"
-export OPENSHELL_E2E_HOST_PORT="${host_port}"
-export OPENSHELL_E2E_GUEST_PORT="${guest_port}"
-
-staged_supervisor="${ROOT}/.cache/openshell-e2e/bin/openshell-sandbox"
-mkdir -p "$(dirname "${staged_supervisor}")"
-ln -sfn "${linux_sandbox_bin}" "${staged_supervisor}"
 
 if [ "${mode}" = host ]; then
 	e2e_align_docker_host_with_cli_context
+	staged_supervisor="${ROOT}/.cache/openshell-e2e/bin/openshell-sandbox"
+	mkdir -p "$(dirname "${staged_supervisor}")"
+	ln -sfn "${linux_sandbox_bin}" "${staged_supervisor}"
+
 	runtime_log="${run_dir}/gateway.log"
 	echo "==> Starting host gateway at ${gateway_endpoint}"
 	(
 		cd "${ROOT}"
-		exec "${host_gateway_bin}" \
+		exec python3 -c \
+			'import os, sys; os.setsid(); os.execvp(sys.argv[1], sys.argv[1:])' \
+			"${host_gateway_bin}" \
 			--config "${gateway_config}" \
 			--bind-address 127.0.0.1 \
 			--port "${host_port}" \
 			--disable-tls
 	) >"${runtime_log}" 2>&1 &
 	child_pid=$!
-	child_process_group=0
 else
 	runtime_log="${run_dir}/vm.log"
 	guest_launcher="${run_dir}/launch-gateway.sh"
-	guest_launcher_path="/home/openshell/.cache/openshell-e2e/bin/launch-gateway"
+	guest_launcher_path=/home/openshell/.cache/openshell-e2e/bin/launch-gateway
 	config_payload="$(base64 <"${gateway_config}" | tr -d '\r\n')"
 	jwt_signing_payload="$(base64 <"${jwt_source_dir}/signing.pem" | tr -d '\r\n')"
 	jwt_public_payload="$(base64 <"${jwt_source_dir}/public.pem" | tr -d '\r\n')"
@@ -477,7 +432,6 @@ EOF
 		vm_args+=(--with "${configuration}")
 	done
 	vm_args+=(
-		--copy "${guest_cli_bin}:/usr/local/bin/openshell"
 		--copy "${guest_gateway_bin}:/usr/local/bin/openshell-gateway"
 		--copy "${linux_sandbox_bin}:/home/openshell/.cache/openshell-e2e/bin/openshell-sandbox"
 		--copy "${guest_launcher}:${guest_launcher_path}"
@@ -496,9 +450,7 @@ EOF
 			"${vm_args[@]}"
 	) >"${runtime_log}" 2>&1 &
 	child_pid=$!
-	child_process_group=1
 fi
-export OPENSHELL_E2E_RUNTIME_LOG="${runtime_log}"
 
 probe_gateway() {
 	python3 - "${OPENSHELL_BIN}" "${1}" <<'PY'
@@ -535,8 +487,11 @@ wait_for_gateway() {
 			break
 		fi
 		if ! kill -0 "${child_pid}" 2>/dev/null; then
-			wait "${child_pid}"
-			process_status=$?
+			if wait "${child_pid}"; then
+				process_status=0
+			else
+				process_status=$?
+			fi
 			child_pid=
 			echo "ERROR: ${mode} gateway process exited before becoming ready" >&2
 			if [ "${process_status}" -eq 0 ]; then
@@ -561,22 +516,12 @@ wait_for_gateway() {
 	return 1
 }
 
-set +e
 wait_for_gateway
-readiness_status=$?
-set -e
-if [ "${readiness_status}" -ne 0 ]; then
-	cleanup "${readiness_status}"
-fi
 
 echo "==> Running E2E suite: ${suite_name}"
 cd "${ROOT}"
-set +e
 cargo test \
 	--manifest-path e2e/rust/Cargo.toml \
 	--features e2e \
 	--test "${suite_name}" \
 	-- --nocapture
-suite_status=$?
-set -e
-cleanup "${suite_status}"