dsh-engineering-ui 0.0.0-stage → 1.5.0

package/engineering/tools/verify_sync.py

@@ -0,0 +1,218 @@
+# -*- coding: utf-8 -*-
+r"""校验 DSH 安装副本与仓库是否【完全同步】（逐文件哈希比对）。
+
+DryRun 报 "0 synced" 只能说明 sync 脚本没找到要复制的 ——
+但它的 Should-Skip 会跳过若干类型（.bak/_probe_/二进制等）。
+因此这里做一次**独立的全量比对**，把差异逐条列出来，
+证明"关键代码确实一致"，而不是只信脚本的自述。
+
+比对对象（sync 脚本覆盖的四处落点）：
+  1. engineering/tools      -> ~/.dsh/.agent-presets/engineering/tools
+  2. engineering/skills     -> ~/.dsh/skills
+  3. engineering/docs       -> ~/.dsh/.agent-presets/engineering/docs
+  4. plugins/dsh-eng-ui     -> ~/.dsh/profiles/*/node_modules/dsh-engineering-ui
+  5. preset 层根文件        -> ~/.dsh/.agent-presets/engineering/
+"""
+import hashlib
+import os
+import sys
+
+try:
+    sys.stdout.reconfigure(encoding="utf-8")
+except Exception:
+    pass
+
+# ── 【换机修复】不写死作者本机桌面路径 ──────────────────────────────────
+# 旧写法 REPO = r"C:\Users\j1877\Desktop\DSH-SW-and-CAD-main"
+# 换台电脑 / 别人 clone 后必然比对失败（报"仓库不存在"或比到别处）。
+# 现在：argv[1] > 环境变量 DSH_REPO_ROOT > 由本文件位置向上两级推导
+#       （tools/ -> engineering/ -> 仓库根）。
+def _find_repo():
+    import sys as _sys
+    if len(_sys.argv) > 1 and os.path.isdir(_sys.argv[1]):
+        return os.path.abspath(_sys.argv[1])
+    env = (os.environ.get("DSH_REPO_ROOT") or "").strip()
+    if env and os.path.isdir(env):
+        return os.path.abspath(env)
+    here = os.path.dirname(os.path.abspath(__file__))      # .../engineering/tools
+    return os.path.dirname(os.path.dirname(here))          # .../           
+
+
+REPO = _find_repo()
+DSH = os.path.expanduser(r"~\.dsh")
+DATA = os.path.join(DSH, ".agent-presets", "engineering")
+
+# 与 sync 脚本保持一致的跳过规则
+RUNTIME_NAMES = {
+    'mode_state.json', 'workflow_state.json', 'sw_state.json',
+    'platform_status.json', 'artifacts_registry.json', 'TASK_FINISHED.json',
+    'connection_state.json', 'design_params.json', 'defense_bypass.json',
+    'violations.json', 'guard.json', 'runtime.json',
+    # -- [换机修复] 本机设置（交付目录 / 各软件路径）按运行时数据处理 --
+    'settings.json',
+    # -- [2026-10-28] 与 sync 脚本同步扩充的运行时数据 --
+    'platform_agents.json', 'part_registry.json', 'custom_materials.json',
+    # 插件/AI 写入的现场文件（不属于代码）
+    'persona.json', 'sw_guard.log', 'sw_guard.pid', 'sw_monitor.pid',
+    # -- [Webots 接入] 与 sync-to-dsh.ps1 同步扩充的运行时数据 --
+    #   仿真现场（状态投影 / 作业账本 / 实例锁）绝不跟着代码走：
+    #   覆盖 = 丢正在跑的作业与证据；跨机同步更是毫无意义。
+    'webots_state.json', 'webots_jobs.sqlite3', 'webots_jobs.sqlite3-wal',
+    'webots_jobs.sqlite3-shm', 'webots_instance.lock',
+}
+RUNTIME_DIRS = {'reports', 'heartbeats', 'load_cases', 'physics_runs',
+                'mode_logs', '.defense', '__pycache__', 'output',
+                # webots_runs/<run_id>/ 是报告与证据落点（model_manifest /
+                #   conversion_report / experiment_report / logs / evidence），
+                #   属于运行现场，不是代码。
+                'webots_runs'}
+BIN_EXT = ('.sldprt', '.sldasm', '.slddrw', '.dwg', '.dxf', '.step', '.stp',
+           '.igs', '.stl', '.pdf', '.png', '.jpg', '.jpeg', '.zip', '.7z')
+
+
+def skip(name, parts):
+    # -- 【换机修复·降噪】按任务的运行时快照也算现场数据 ------------------
+    # 实测：安装副本里有 artifacts_registry.20261019_202722_88C07F68.json
+    #   这类【带时间戳的历史台账快照】，属于会话现场数据，本来就不该
+    #   从仓库同步过去，也不该被算成差异（否则每次比对都红一片，
+    #   真正的代码差异被淹掉）。
+    if name.startswith("artifacts_registry.") and name.endswith(".json"):
+        return True
+    if name in RUNTIME_NAMES:
+        return True
+    if '.bak' in name or name.endswith('.pyc') or name.startswith('~$'):
+        return True
+    if name.startswith('_probe_'):
+        return True
+    # 调试/一次性脚本（_tmp_ 前缀）与日志/pid 都不属于代码
+    if name.startswith('_tmp_'):
+        return True
+    if name.endswith('.log') or name.endswith('.pid'):
+        return True
+    if '%' in name:
+        return True
+    if name.lower().endswith(BIN_EXT):
+        return True
+    for p in parts:
+        if p in RUNTIME_DIRS:
+            return True
+        if p.startswith('_probe_') or '%' in p:
+            return True
+        if '.bak' in str(p):
+            return True
+    return False
+
+
+def md5(p):
+    h = hashlib.md5()
+    with open(p, 'rb') as f:
+        for chunk in iter(lambda: f.read(1 << 20), b''):
+            h.update(chunk)
+    return h.hexdigest()
+
+
+def compare(src_root, dst_root, label):
+    """返回 (only_in_src, only_in_dst, differ, same_count)"""
+    only_src, only_dst, differ, same = [], [], [], 0
+    if not os.path.isdir(src_root):
+        return None
+    if not os.path.isdir(dst_root):
+        return ([], [], [], 0, "目标目录不存在: %s" % dst_root)
+    seen = set()
+    for dp, dns, fns in os.walk(src_root):
+        rel_dir = os.path.relpath(dp, src_root)
+        parts_dir = [] if rel_dir == '.' else rel_dir.split(os.sep)
+        for fn in fns:
+            rel = os.path.join(rel_dir, fn) if rel_dir != '.' else fn
+            parts = parts_dir + [fn]
+            if skip(fn, parts):
+                continue
+            seen.add(os.path.normcase(rel))
+            s = os.path.join(dp, fn)
+            d = os.path.join(dst_root, rel)
+            if not os.path.exists(d):
+                only_src.append(rel)
+            elif md5(s) != md5(d):
+                differ.append(rel)
+            else:
+                same += 1
+    for dp, dns, fns in os.walk(dst_root):
+        rel_dir = os.path.relpath(dp, dst_root)
+        parts_dir = [] if rel_dir == '.' else rel_dir.split(os.sep)
+        for fn in fns:
+            rel = os.path.join(rel_dir, fn) if rel_dir != '.' else fn
+            parts = parts_dir + [fn]
+            if skip(fn, parts):
+                continue
+            if os.path.normcase(rel) not in seen:
+                only_dst.append(rel)
+    return (only_src, only_dst, differ, same, None)
+
+
+PAIRS = [
+    (os.path.join(REPO, "engineering", "tools"),
+     os.path.join(DATA, "tools"), "tools"),
+    (os.path.join(REPO, "engineering", "skills"),
+     os.path.join(DSH, "skills"), "skills"),
+    (os.path.join(REPO, "engineering", "docs"),
+     os.path.join(DATA, "docs"), "docs"),
+    (os.path.join(REPO, "engineering", "plugins", "dsh-engineering-ui"),
+     os.path.join(DSH, "profiles", "web", "node_modules", "dsh-engineering-ui"),
+     "plugin@web"),
+    (os.path.join(REPO, "engineering", "plugins", "dsh-engineering-ui"),
+     os.path.join(DSH, "profiles", "desktop", "node_modules", "dsh-engineering-ui"),
+     "plugin@desktop"),
+]
+
+print("=" * 78)
+print("DSH 安装副本 vs 仓库：逐文件哈希比对")
+print("=" * 78)
+total_bad = 0
+for src, dst, label in PAIRS:
+    r = compare(src, dst, label)
+    if r is None:
+        print("\n[%s] 源目录不存在，跳过" % label)
+        continue
+    if len(r) == 5 and r[4]:
+        print("\n[%s] %s" % (label, r[4]))
+        total_bad += 1
+        continue
+    only_src, only_dst, differ, same = r[:4]
+    bad = len(only_src) + len(differ)
+    total_bad += bad
+    print("\n[%s]" % label)
+    print("  一致 = %d" % same)
+    print("  仅仓库有（安装缺） = %d" % len(only_src))
+    print("  内容不同          = %d" % len(differ))
+    print("  仅安装有（多余）   = %d" % len(only_dst))
+    for x in only_src[:8]:
+        print("      [缺] %s" % x)
+    for x in differ[:8]:
+        print("      [异] %s" % x)
+    for x in only_dst[:5]:
+        print("      [余] %s" % x)
+
+# preset 层根文件
+print("\n[preset 根文件]")
+for fn in ('agent.cordis.yml', 'preset.yml', 'install-plugin.ps1', 'README.md'):
+    s = os.path.join(REPO, "engineering", fn)
+    d = os.path.join(DATA, fn)
+    if not os.path.exists(s):
+        continue
+    if not os.path.exists(d):
+        print("   [缺] %s" % fn)
+        total_bad += 1
+    elif md5(s) != md5(d):
+        print("   [异] %s" % fn)
+        total_bad += 1
+    else:
+        print("   [同] %s" % fn)
+
+print()
+print("=" * 78)
+if total_bad == 0:
+    print("★ 结论：安装副本与仓库【完全同步】，无差异 ✅")
+else:
+    print("★ 结论：仍有 %d 处差异，需要再次同步 ❌" % total_bad)
+print("=" * 78)
+sys.exit(1 if total_bad else 0)