From 1babec1b85545c1da3cbebdc4904a25520c128b9 Mon Sep 17 00:00:00 2001
From: nanxun <nanxun@nxsir.cn>
Date: Wed, 12 Aug 2026 17:33:05 +0800
Subject: [PATCH] fix(fnos): reuse content-addressed core runtime

---
 fnos/cmd/runtime_bootstrap     | 38 +++++++++++++++++++++++++++-------
 scripts/build-fnos.sh          |  3 ++-
 scripts/smoke-fnos-package.sh  |  2 +-
 scripts/verify-fnos-package.sh |  5 +++++
 tests/test_fnos_lifecycle.py   | 27 ++++++++++++++++++++++--
 5 files changed, 63 insertions(+), 12 deletions(-)

diff --git a/fnos/cmd/runtime_bootstrap b/fnos/cmd/runtime_bootstrap
index bea2ffb..7f117e6 100755
--- a/fnos/cmd/runtime_bootstrap
+++ b/fnos/cmd/runtime_bootstrap
@@ -18,11 +18,15 @@ fail() {
 [ -d "$WHEEL_ROOT" ] || fail "ImageFind 运行环境初始化失败：安装包缺少核心 wheelhouse。"
 [ -f "$CORE_REQUIREMENTS" ] || fail "ImageFind 运行环境初始化失败：安装包缺少核心依赖锁文件。"
 
-VERSION=$(sed -n '1p' "$VERSION_FILE")
-case "$VERSION" in
+PACKAGE_VERSION=$(sed -n '1p' "$VERSION_FILE")
+case "$PACKAGE_VERSION" in
     ''|*[!0-9.]* ) fail "ImageFind 运行环境初始化失败：版本信息无效。" ;;
 esac
-PACKAGE_ID=$(sed -n '2p' "$VERSION_FILE")
+APP_VERSION=$(sed -n '2p' "$VERSION_FILE")
+case "$APP_VERSION" in
+    ''|*[!0-9.]* ) fail "ImageFind 运行环境初始化失败：应用版本信息无效。" ;;
+esac
+PACKAGE_ID=$(sed -n '3p' "$VERSION_FILE")
 case "$PACKAGE_ID" in
     ''|*[!0-9a-f]* ) fail "ImageFind 运行环境初始化失败：安装包指纹无效。" ;;
 esac
@@ -39,8 +43,9 @@ PYTHON_VERSION=$($SYSTEM_PYTHON -c 'import sys; print(f"{sys.version_info.major}
 [ "$PYTHON_VERSION" = "3.12" ] || fail "ImageFind 需要 Python 3.12，当前为 $PYTHON_VERSION。"
 
 mkdir -p "$RUNTIME_ROOT"
-TARGET="$RUNTIME_ROOT/core-$VERSION"
-STAGE="$RUNTIME_ROOT/.core-$VERSION.tmp.$$"
+RUNTIME_ID=$(printf '%s' "$PACKAGE_ID" | cut -c1-16)
+TARGET="$RUNTIME_ROOT/core-$APP_VERSION-$RUNTIME_ID"
+STAGE="$RUNTIME_ROOT/.core-$APP_VERSION-$RUNTIME_ID.tmp.$$"
 CURRENT="$RUNTIME_ROOT/current"
 
 runtime_valid() {
@@ -51,7 +56,7 @@ runtime_valid() {
         IMAGEFIND_DATA_DIR="$DATA_ROOT/data" "$1/bin/python" -c \
             'import imagefind; print(imagefind.__version__)' 2>"$RUNTIME_ROOT/version-check.err"
     ) || return 1
-    [ "$RUNTIME_VERSION" = "$VERSION" ]
+    [ "$RUNTIME_VERSION" = "$APP_VERSION" ]
 }
 
 if ! runtime_valid "$TARGET"; then
@@ -83,7 +88,7 @@ if ! runtime_valid "$TARGET"; then
             fail "ImageFind 核心运行环境校验失败。"
         }
     if [ -e "$TARGET" ]; then
-        BACKUP="$RUNTIME_ROOT/core-$VERSION.invalid.$(date +%s)"
+        BACKUP="$RUNTIME_ROOT/core-$APP_VERSION-$RUNTIME_ID.invalid.$(date +%s)"
         mv "$TARGET" "$BACKUP"
     fi
     mv "$STAGE" "$TARGET"
@@ -122,5 +127,22 @@ if ! runtime_valid "$CURRENT"; then
     else
         ACTUAL_VERSION='Python 不可执行'
     fi
-    fail "ImageFind 核心运行环境切换失败：期望版本 $VERSION，current=$CURRENT_LINK，实际版本=$ACTUAL_VERSION。"
+    fail "ImageFind 核心运行环境切换失败：期望版本 $APP_VERSION，current=$CURRENT_LINK，实际版本=$ACTUAL_VERSION。"
 fi
+
+# Keep the active runtime plus one most-recent rollback candidate. fnOS package
+# build numbers change on every release, so retaining every historical venv
+# would otherwise grow the application data directory without bound.
+ROLLBACK_KEPT=0
+for OLD_RUNTIME in $(ls -dt "$RUNTIME_ROOT"/core-* 2>/dev/null || true); do
+    [ -d "$OLD_RUNTIME" ] || continue
+    [ "$OLD_RUNTIME" = "$TARGET" ] && continue
+    case "$OLD_RUNTIME" in
+        *.invalid.*) rm -rf -- "$OLD_RUNTIME"; continue ;;
+    esac
+    if [ "$ROLLBACK_KEPT" -eq 0 ]; then
+        ROLLBACK_KEPT=1
+    else
+        rm -rf -- "$OLD_RUNTIME"
+    fi
+done
diff --git a/scripts/build-fnos.sh b/scripts/build-fnos.sh
index 783842c..8e3c117 100755
--- a/scripts/build-fnos.sh
+++ b/scripts/build-fnos.sh
@@ -50,6 +50,7 @@ PYTHON_BIN=$(command -v "$PYTHON_BIN") || {
     printf 'Python with hatchling is required to assemble the release.\n' >&2
     exit 1
 }
+APP_VERSION=$("$PYTHON_BIN" -c 'import pathlib,sys,tomllib; print(tomllib.loads(pathlib.Path(sys.argv[1]).read_text())["project"]["version"])' "$PROJECT_ROOT/pyproject.toml")
 command -v npm >/dev/null 2>&1 || { printf 'npm is required.\n' >&2; exit 1; }
 "$PYTHON_BIN" -c 'import hatchling' >/dev/null 2>&1 || {
     printf 'hatchling is required in the release Python environment.\n' >&2
@@ -89,7 +90,7 @@ done
 PACKAGE_ID=$("$PYTHON_BIN" -c \
     'import hashlib,sys; h=hashlib.sha256(); [h.update(open(path,"rb").read()) for path in sys.argv[1:]]; print(h.hexdigest())' \
     "$STAGED_APP_WHEEL" "$STAGE/app/runtime/runtime-core.txt")
-printf '%s\n%s\n' "$PACKAGE_VERSION" "$PACKAGE_ID" >"$STAGE/app/runtime/VERSION"
+printf '%s\n%s\n%s\n' "$PACKAGE_VERSION" "$APP_VERSION" "$PACKAGE_ID" >"$STAGE/app/runtime/VERSION"
 "$PYTHON_BIN" "$PROJECT_ROOT/scripts/make_icons.py" "$STAGE"
 chmod 0755 "$STAGE/cmd/"*
 (
diff --git a/scripts/smoke-fnos-package.sh b/scripts/smoke-fnos-package.sh
index 73c9131..1280d24 100755
--- a/scripts/smoke-fnos-package.sh
+++ b/scripts/smoke-fnos-package.sh
@@ -17,7 +17,7 @@ SMOKE_PASSWORD="imagefind-smoke-password"
 SMOKE_POSTGRES_TOKEN="${IMAGEFIND_SMOKE_POSTGRES_TOKEN:-}"
 COOKIE_JAR="$DATA_ROOT/run/smoke-cookie.txt"
 BACKUP_FILE="$DATA_ROOT/run/smoke.ifbackup"
-EXPECTED_VERSION=$(sed -n '1p' "$APP_ROOT/runtime/VERSION")
+EXPECTED_VERSION=$(sed -n '2p' "$APP_ROOT/runtime/VERSION")
 
 APP_ROOT=$(CDPATH= cd -- "$APP_ROOT" && pwd)
 [ "${#SMOKE_POSTGRES_TOKEN}" -ge 20 ] || {
diff --git a/scripts/verify-fnos-package.sh b/scripts/verify-fnos-package.sh
index 86564d2..9c66c44 100755
--- a/scripts/verify-fnos-package.sh
+++ b/scripts/verify-fnos-package.sh
@@ -26,6 +26,11 @@ grep -q '^runtime/wheels/.*\.whl$' "$work_dir/files.txt"
 grep -q '^bin/libOpenCL.so.1$' "$work_dir/files.txt"
 runtime_version=$(tar -xOzf "$work_dir/app.tgz" runtime/VERSION | sed -n '1p')
 test "$runtime_version" = "$version"
+runtime_app_version=$(tar -xOzf "$work_dir/app.tgz" runtime/VERSION | sed -n '2p')
+runtime_package_id=$(tar -xOzf "$work_dir/app.tgz" runtime/VERSION | sed -n '3p')
+test -n "$runtime_app_version"
+test "${#runtime_package_id}" -eq 64
+grep -q "^runtime/imagefind-${runtime_app_version}-.*\\.whl$" "$work_dir/files.txt"
 if grep -Eq '^(data|models|cache|postgresql)/' "$work_dir/files.txt"; then
     printf 'persistent data must not be included in ImageFind app.tgz\n' >&2
     exit 1
diff --git a/tests/test_fnos_lifecycle.py b/tests/test_fnos_lifecycle.py
index e0bc636..e6f3c7c 100644
--- a/tests/test_fnos_lifecycle.py
+++ b/tests/test_fnos_lifecycle.py
@@ -19,7 +19,7 @@ def _mock_app(tmp_path: Path) -> tuple[Path, Path, Path]:
     app_root = tmp_path / "app"
     runtime = app_root / "runtime"
     (runtime / "wheels").mkdir(parents=True)
-    (runtime / "VERSION").write_text("0.4.3\n" + "a" * 64 + "\n")
+    (runtime / "VERSION").write_text("0.4.30001\n0.4.3\n" + "a" * 64 + "\n")
     (runtime / "runtime-core.txt").write_text("# mocked offline core\n")
     (runtime / "imagefind-0.4.3-py3-none-any.whl").write_bytes(b"mock wheel")
     (runtime / "wheels" / "dependency.whl").write_bytes(b"mock dependency")
@@ -312,7 +312,7 @@ def test_runtime_bootstrap_atomically_replaces_existing_current(
 
     assert result.returncode == 0, result.stderr
     assert current.is_symlink()
-    assert current.resolve() == runtime_root / "core-0.4.3"
+    assert current.resolve() == runtime_root / "core-0.4.3-aaaaaaaaaaaaaaaa"
     version = subprocess.run(
         [str(current / "bin" / "python"), "-m", "imagefind.main", "--version"],
         env=env,
@@ -332,6 +332,29 @@ def test_runtime_bootstrap_atomically_replaces_existing_current(
         assert (previous[0] / "sentinel").read_text() == "preserve old runtime"
 
 
+def test_runtime_bootstrap_reuses_content_id_across_package_builds_and_prunes_old_runtimes(tmp_path: Path):
+    app_root, args_file, password_file = _mock_app(tmp_path)
+    env = _environment(tmp_path, app_root, args_file, password_file)
+    runtime_root = tmp_path / "var" / "runtime"
+    runtime_root.mkdir(parents=True)
+    for name in ("core-0.3.14-old", "core-0.3.16-newer"):
+        old_runtime = runtime_root / name
+        old_runtime.mkdir()
+        (old_runtime / "sentinel").write_text(name)
+
+    first = _run("runtime_bootstrap", env)
+    target = runtime_root / "core-0.4.3-aaaaaaaaaaaaaaaa"
+    assert first.returncode == 0, first.stderr
+    assert target.is_dir()
+    assert len([path for path in runtime_root.glob("core-*") if path.is_dir()]) == 2
+
+    (app_root / "runtime" / "VERSION").write_text("0.4.30002\n0.4.3\n" + "a" * 64 + "\n")
+    second = _run("runtime_bootstrap", env)
+    assert second.returncode == 0, second.stderr
+    assert (runtime_root / "current").resolve() == target
+    assert len([path for path in runtime_root.glob("core-*") if path.is_dir()]) == 2
+
+
 def test_config_callback_writes_validated_direct_access_atomically(tmp_path: Path):
     data_root = tmp_path / "var"
     env = {