chore: snapshot backup before rainycy push (20260624-032434)

Auto-committed by MiMo for migration to git.rainycy.top

main.py

@@ -58,6 +58,7 @@ class LinuxDoPreviewPlugin(Star):
         # 登录状态：跨 fetch 复用在同一 StealthySession 中
         self._auth_check_done = False
         self._logged_in = False
+        self._auth_lock = threading.Lock()
 
     async def terminate(self):
         _EXECUTOR.shutdown(wait=False)
@@ -203,173 +204,175 @@ class LinuxDoPreviewPlugin(Star):
                 return None
 
             page = ctx.new_page()
-            page.set_viewport_size({"width": 1280, "height": 900})
-
-            # ── 导航：等 networkidle 确保 JS 动态内容加载完成 ──
-            page.goto(url, wait_until="networkidle", timeout=timeout_ms)
-
-            # ── 等待 Discourse 帖子内容渲染 ──
             try:
-                page.wait_for_selector("#post_1", timeout=min(timeout_ms, 10000))
+                page.set_viewport_size({"width": 1280, "height": 900})
-            except Exception:
-                page.wait_for_timeout(3000)  # 回退：固定等待
 
-            # ── 隐藏非楼主内容，只保留第一篇帖子完整展示 ──
+                # ── 导航：等 networkidle 确保 JS 动态内容加载完成 ──
-            page.evaluate("""() => {
+                page.goto(url, wait_until="networkidle", timeout=timeout_ms)
-                const hide = (sel) => {
-                    const el = document.querySelector(sel);
-                    if (el) el.style.display = 'none';
-                };
-                hide('.d-header');                      // 顶部导航栏
-                hide('.sidebar-wrapper');               // 左侧边栏
-                hide('.topic-navigation-wrapper');      // 帖子导航条
-                hide('.footer-nav.visible');            // 底部导航
-                hide('.post-stream');                   // 隐藏整个帖子流（后面单独显示楼主）
 
-                // 隐藏所有回复帖子，只保留楼主
+                # ── 等待 Discourse 帖子内容渲染 ──
-                const posts = document.querySelectorAll('.topic-post');
+                try:
-                posts.forEach((post, i) => { if (i > 0) post.style.display = 'none'; });
+                    page.wait_for_selector("#post_1", timeout=min(timeout_ms, 10000))
+                except Exception:
+                    page.wait_for_timeout(3000)  # 回退：固定等待
 
-                // 滚动到顶部
+                # ── 隐藏非楼主内容，只保留第一篇帖子完整展示 ──
-                window.scrollTo(0, 0);
+                page.evaluate("""() => {
-            }""")
+                    const hide = (sel) => {
+                        const el = document.querySelector(sel);
+                        if (el) el.style.display = 'none';
+                    };
+                    hide('.d-header');                      // 顶部导航栏
+                    hide('.sidebar-wrapper');               // 左侧边栏
+                    hide('.topic-navigation-wrapper');      // 帖子导航条
+                    hide('.footer-nav.visible');            // 底部导航
+                    hide('.post-stream');                   // 隐藏整个帖子流（后面单独显示楼主）
 
-            # ── 展开 Discourse 截断的长帖 ──
+                    // 隐藏所有回复帖子，只保留楼主
-            page.evaluate("""() => {
+                    const posts = document.querySelectorAll('.topic-post');
-                // 移除所有展开按钮和截断遮罩
+                    posts.forEach((post, i) => { if (i > 0) post.style.display = 'none'; });
-                const removeSelectors = [
-                    '.expand-post',
-                    '.gap-bottom',
-                    '.gap',
-                    '.large-post-container .show-more',
-                    '.topic-body .show-more',
-                    '.cooked .show-more',
-                    '.lightbox',
-                ];
-                removeSelectors.forEach(sel => {
-                    document.querySelectorAll(sel).forEach(el => el.remove());
-                });
 
-                // 移除所有 max-height / overflow 限制
+                    // 滚动到顶部
-                const unclampSelectors = [
+                    window.scrollTo(0, 0);
-                    '.cooked',
+                }""")
-                    '.topic-body',
+
-                    '#post_1 .cooked',
+                # ── 展开 Discourse 截断的长帖 ──
-                    '#post_1 .topic-body',
+                page.evaluate("""() => {
-                    '#post_1 .contents',
+                    // 移除所有展开按钮和截断遮罩
-                    '.large-post-container',
+                    const removeSelectors = [
-                ];
+                        '.expand-post',
-                unclampSelectors.forEach(sel => {
+                        '.gap-bottom',
-                    document.querySelectorAll(sel).forEach(el => {
+                        '.gap',
+                        '.large-post-container .show-more',
+                        '.topic-body .show-more',
+                        '.cooked .show-more',
+                        '.lightbox',
+                    ];
+                    removeSelectors.forEach(sel => {
+                        document.querySelectorAll(sel).forEach(el => el.remove());
+                    });
+
+                    // 移除所有 max-height / overflow 限制
+                    const unclampSelectors = [
+                        '.cooked',
+                        '.topic-body',
+                        '#post_1 .cooked',
+                        '#post_1 .topic-body',
+                        '#post_1 .contents',
+                        '.large-post-container',
+                    ];
+                    unclampSelectors.forEach(sel => {
+                        document.querySelectorAll(sel).forEach(el => {
+                            el.style.maxHeight = 'none';
+                            el.style.overflow = 'visible';
+                            el.style.height = 'auto';
+                        });
+                    });
+
+                    // 展开 Discourse 长帖截断（data-* 属性方式）
+                    document.querySelectorAll('[data-expanded]').forEach(el => {
+                        el.setAttribute('data-expanded', 'true');
+                    });
+                    // 移除 truncated 标记
+                    document.querySelectorAll('.truncated').forEach(el => {
+                        el.classList.remove('truncated');
+                    });
+                }""")
+
+                # ── 点击可能存在的展开按钮 ──
+                try:
+                    expand_buttons = page.query_selector_all(
+                        '#post_1 .expand-post, #post_1 .show-more, '
+                        '#post_1 button[class*="expand"], '
+                        '#post_1 a[class*="expand"]'
+                    )
+                    for btn in expand_buttons:
+                        try:
+                            btn.click()
+                            page.wait_for_timeout(300)
+                        except Exception:
+                            pass
+                except Exception:
+                    pass
+
+                # ── 再次展开，防止点击按钮后重新截断 ──
+                page.evaluate("""() => {
+                    ['#post_1 .cooked', '#post_1 .topic-body', '#post_1 .contents'].forEach(sel => {
+                        document.querySelectorAll(sel).forEach(el => {
+                            el.style.maxHeight = 'none';
+                            el.style.overflow = 'visible';
+                            el.style.height = 'auto';
+                        });
+                    });
+                    // 确保图片容器也不截断
+                    document.querySelectorAll('#post_1 .lightbox-wrapper').forEach(el => {
                         el.style.maxHeight = 'none';
                         el.style.overflow = 'visible';
-                        el.style.height = 'auto';
                     });
-                });
+                }""")
 
-                // 展开 Discourse 长帖截断（data-* 属性方式）
+                # ── 滚动楼主帖子，触发懒加载图片 ──
-                document.querySelectorAll('[data-expanded]').forEach(el => {
+                post1_box = page.evaluate("""() => {
-                    el.setAttribute('data-expanded', 'true');
+                    const p1 = document.querySelector('#post_1');
-                });
+                    if (!p1) return null;
-                // 移除 truncated 标记
+                    const rect = p1.getBoundingClientRect();
-                document.querySelectorAll('.truncated').forEach(el => {
+                    return { top: rect.top + window.scrollY, height: rect.height };
-                    el.classList.remove('truncated');
+                }""")
-                });
+                if post1_box:
-            }""")
+                    post_top = int(post1_box.get('top', 0))
+                    post_height = int(post1_box.get('height', 0))
+                    for y in range(post_top, post_top + post_height, 400):
+                        page.evaluate(f"window.scrollTo(0, {y})")
+                        page.wait_for_timeout(200)
+                else:
+                    # 回退：滚动整个页面
+                    total_height = page.evaluate("document.body.scrollHeight")
+                    for y in range(0, total_height, 400):
+                        page.evaluate(f"window.scrollTo(0, {y})")
+                        page.wait_for_timeout(200)
 
-            # ── 点击可能存在的展开按钮 ──
+                # ── 等待图片加载完成 ──
-            try:
+                page.evaluate("""() => {
-                expand_buttons = page.query_selector_all(
+                    return new Promise(resolve => {
-                    '#post_1 .expand-post, #post_1 .show-more, '
+                        const imgs = document.querySelectorAll('#post_1 img');
-                    '#post_1 button[class*="expand"], '
+                        let loaded = 0;
-                    '#post_1 a[class*="expand"]'
+                        const total = imgs.length;
-                )
+                        if (total === 0) return resolve();
-                for btn in expand_buttons:
+                        imgs.forEach(img => {
-                    try:
+                            if (img.complete) {
-                        btn.click()
-                        page.wait_for_timeout(300)
-                    except Exception:
-                        pass
-            except Exception:
-                pass
-
-            # ── 再次展开，防止点击按钮后重新截断 ──
-            page.evaluate("""() => {
-                ['#post_1 .cooked', '#post_1 .topic-body', '#post_1 .contents'].forEach(sel => {
-                    document.querySelectorAll(sel).forEach(el => {
-                        el.style.maxHeight = 'none';
-                        el.style.overflow = 'visible';
-                        el.style.height = 'auto';
-                    });
-                });
-                // 确保图片容器也不截断
-                document.querySelectorAll('#post_1 .lightbox-wrapper').forEach(el => {
-                    el.style.maxHeight = 'none';
-                    el.style.overflow = 'visible';
-                });
-            }""")
-
-            # ── 滚动楼主帖子，触发懒加载图片 ──
-            post1_box = page.evaluate("""() => {
-                const p1 = document.querySelector('#post_1');
-                if (!p1) return null;
-                const rect = p1.getBoundingClientRect();
-                return { top: rect.top + window.scrollY, height: rect.height };
-            }""")
-            if post1_box:
-                post_top = int(post1_box.get('top', 0))
-                post_height = int(post1_box.get('height', 0))
-                for y in range(post_top, post_top + post_height, 400):
-                    page.evaluate(f"window.scrollTo(0, {y})")
-                    page.wait_for_timeout(200)
-            else:
-                # 回退：滚动整个页面
-                total_height = page.evaluate("document.body.scrollHeight")
-                for y in range(0, total_height, 400):
-                    page.evaluate(f"window.scrollTo(0, {y})")
-                    page.wait_for_timeout(200)
-
-            # ── 等待图片加载完成 ──
-            page.evaluate("""() => {
-                return new Promise(resolve => {
-                    const imgs = document.querySelectorAll('#post_1 img');
-                    let loaded = 0;
-                    const total = imgs.length;
-                    if (total === 0) return resolve();
-                    imgs.forEach(img => {
-                        if (img.complete) {
-                            loaded++;
-                            if (loaded >= total) resolve();
-                        } else {
-                            img.onload = img.onerror = () => {
                                 loaded++;
                                 if (loaded >= total) resolve();
-                            };
+                            } else {
-                        }
+                                img.onload = img.onerror = () => {
+                                    loaded++;
+                                    if (loaded >= total) resolve();
+                                };
+                            }
+                        });
+                        // 最多等 3 秒
+                        setTimeout(resolve, 3000);
                     });
-                    // 最多等 3 秒
+                }""")
-                    setTimeout(resolve, 3000);
-                });
-            }""")
 
-            # ── 滚动回顶部 ──
+                # ── 滚动回顶部 ──
-            page.evaluate("window.scrollTo(0, 0)")
+                page.evaluate("window.scrollTo(0, 0)")
-            page.wait_for_timeout(500)
+                page.wait_for_timeout(500)
 
-            # ── 截图：全页模式，隐藏导航栏后内容干净 ──
+                # ── 截图：全页模式，隐藏导航栏后内容干净 ──
-            full_page = self.config.get("screenshot_full_page", True)
+                full_page = self.config.get("screenshot_full_page", True)
-            page.screenshot(
+                page.screenshot(
-                path=str(save_path),
+                    path=str(save_path),
-                full_page=full_page,
+                    full_page=full_page,
-                timeout=timeout_ms,
+                    timeout=timeout_ms,
-            )
+                )
 
-            sz = save_path.stat().st_size
+                sz = save_path.stat().st_size
-            logger.info(
+                logger.info(
-                f"[LinuxDoPreview] 截图保存: {save_path.name} ({sz / 1024:.1f} KB)"
+                    f"[LinuxDoPreview] 截图保存: {save_path.name} ({sz / 1024:.1f} KB)"
-            )
+                )
-            page.close()
+                return save_path
-            return save_path
+            finally:
+                page.close()
 
         except Exception as e:
             logger.warning(f"[LinuxDoPreview] 截图失败: {type(e).__name__}: {e}")
@@ -377,54 +380,56 @@ class LinuxDoPreviewPlugin(Star):
 
     # ─────────── 文本提取 ───────────
 
+    @staticmethod
+    def _build_json_url(url: str) -> str:
+        """将 topic URL 转换为对应的 JSON API URL"""
+        json_url = url.rstrip('/')
+        if not json_url.endswith('.json'):
+            parts = json_url.split('/')
+            t_idx = -1
+            for i, p in enumerate(parts):
+                if p == 't':
+                    t_idx = i
+                    break
+            if t_idx >= 0 and len(parts) > t_idx + 2:
+                json_url = '/'.join(parts[:t_idx + 3])
+            json_url += '.json'
+        return json_url
+
     def _extract_content_from_json(self, session, url: str) -> str:
         """通过 Discourse JSON API 获取完整的楼主帖子内容
-        
+
         Discourse 的 .json 端点返回结构化数据，包含完整的 cooked HTML，
         不受页面截断、懒加载或 Cloudflare 渲染问题的影响。
         """
         try:
-            # 构造 JSON URL：topic-url.json 或 topic-url/1.json
+            json_url = self._build_json_url(url)
-            json_url = url.rstrip('/')
+
-            if not json_url.endswith('.json'):
-                # 对于帖子链接如 /t/topic-slug/12345/5，取前两段
-                parts = json_url.split('/')
-                # 找到 /t/ 后的部分
-                t_idx = -1
-                for i, p in enumerate(parts):
-                    if p == 't':
-                        t_idx = i
-                        break
-                if t_idx >= 0 and len(parts) > t_idx + 2:
-                    # 重建为 /t/slug/id 格式
-                    json_url = '/'.join(parts[:t_idx + 3])
-                json_url += '.json'
-            
             logger.info(f"[LinuxDoPreview] JSON API 请求: {json_url}")
             resp = session.fetch(json_url)
             if resp.status != 200:
                 logger.info(f"[LinuxDoPreview] JSON API 返回 {resp.status}")
                 return ""
-            
+
             import json
             data = json.loads(resp.body.decode("utf-8", errors="replace"))
-            
+
             # 从 post_stream 中提取第一个帖子（楼主）
             post_stream = data.get("post_stream", {})
             posts = post_stream.get("posts", [])
             if not posts:
                 return ""
-            
+
             first_post = posts[0]
             cooked_html = first_post.get("cooked", "")
             if not cooked_html:
                 return ""
-            
+
             # 使用 lxml 解析 HTML 并提取纯文本
             if _lh is not None:
                 tree = _lh.fromstring(cooked_html)
                 return _clean_text(tree.text_content())
-            
+
             # 回退：正则去标签
             text = re.sub(r"<[^>]+>", " ", cooked_html)
             text = re.sub(r"\s+", " ", text).strip()
@@ -567,25 +572,34 @@ class LinuxDoPreviewPlugin(Star):
         if self._has_session_cookie():
             cookie_value = (self.config.get("linuxdo_session_cookie", "") or "").strip()
             if not self._inject_session_cookie(session, cookie_value):
-                self._auth_check_done = True
+                with self._auth_lock:
-                self._logged_in = False
+                    self._auth_check_done = True
+                    self._logged_in = False
                 return False
             # 校验结果只算一次（Cookie 有效性跨请求稳定）
-            if not self._auth_check_done:
+            with self._auth_lock:
-                self._logged_in = self._check_login_state(session)
+                need_check = not self._auth_check_done
-                self._auth_check_done = True
+            if need_check:
-                if self._logged_in:
+                logged_in = self._check_login_state(session)
+                with self._auth_lock:
+                    self._logged_in = logged_in
+                    self._auth_check_done = True
+                if logged_in:
                     logger.info("[LinuxDoPreview] Cookie 登录验证成功")
                 else:
                     logger.warning(
                         "[LinuxDoPreview] 会话 Cookie 无效或已过期，将匿名访问。"
                         "请在浏览器重新获取 Cookie（推荐 _t，长效）后填入配置。"
                     )
-            return self._logged_in
+            with self._auth_lock:
+                return self._logged_in
 
         # ── 仅用户名/密码：受 hCaptcha 限制，无法自动登录（仅提示一次） ──
-        if self._has_auto_login() and not self._auth_check_done:
+        with self._auth_lock:
-            self._auth_check_done = True
+            need_warn = self._has_auto_login() and not self._auth_check_done
+            if need_warn:
+                self._auth_check_done = True
+        if need_warn:
             logger.warning(
                 "[LinuxDoPreview] linux.do 登录启用了 hCaptcha 人机验证，账号密码"
                 "自动登录不可用。请在浏览器登录 linux.do 后，F12 → Application → "
@@ -594,27 +608,18 @@ class LinuxDoPreviewPlugin(Star):
             )
 
         # 都没配置 / 自动登录不可用 → 匿名
-        self._logged_in = False
+        with self._auth_lock:
+            self._logged_in = False
         return False
 
     def _fetch_topic_data(self, session, url: str) -> dict | None:
         """通过 Discourse JSON API 获取完整的主题数据
-        
+
         返回的 dict 包含帖子原始数据（cooked HTML、作者、标签、统计等），
         可同时供文本提取和自定义 HTML 渲染使用。
         """
         try:
-            json_url = url.rstrip('/')
+            json_url = self._build_json_url(url)
-            if not json_url.endswith('.json'):
-                parts = json_url.split('/')
-                t_idx = -1
-                for i, p in enumerate(parts):
-                    if p == 't':
-                        t_idx = i
-                        break
-                if t_idx >= 0 and len(parts) > t_idx + 2:
-                    json_url = '/'.join(parts[:t_idx + 3])
-                json_url += '.json'
 
             logger.info(f"[LinuxDoPreview] 拉取 topic JSON: {json_url}")
             resp = session.fetch(json_url)
@@ -749,7 +754,6 @@ class LinuxDoPreviewPlugin(Star):
         posts_count = topic_data.get("posts_count", 0)
         views = topic_data.get("views", 0)
         like_count = topic_data.get("like_count", 0)
-        created_at = topic_data.get("created_at", "")
         tags = topic_data.get("tags", []) or []
 
         post_stream = topic_data.get("post_stream", {}) or {}
@@ -774,12 +778,14 @@ class LinuxDoPreviewPlugin(Star):
         if "{size}" in author_avatar:
             author_avatar = author_avatar.replace("{size}", "120")
         post_created = first.get("created_at", "") or ""
-        post_like = first.get("like_count", 0)
         cooked_html = first.get("cooked", "") or ""
 
         # 把 Discourse 相对资源 URL 补全为绝对 URL
         cooked_html = self._normalize_cooked_urls(cooked_html)
 
+        # 消毒 HTML：移除危险标签和事件属性
+        cooked_html = self._sanitize_html(cooked_html)
+
         # 发布时间格式化
         created_text = ""
         if post_created:
@@ -981,6 +987,27 @@ class LinuxDoPreviewPlugin(Star):
             return f"{n/1000:.1f}k"
         return str(n)
 
+    @staticmethod
+    def _sanitize_html(html_str: str) -> str:
+        """消毒 HTML：移除 <script>、<iframe>、<object>、<embed> 及 on* 事件属性"""
+        if not html_str:
+            return ""
+        html_str = re.sub(
+            r'<\s*(script|iframe|object|embed|applet|form|input|textarea|button|select)\b[^>]*>.*?<\s*/\s*\1\s*>',
+            '', html_str, flags=re.IGNORECASE | re.DOTALL
+        )
+        html_str = re.sub(
+            r'<\s*(script|iframe|object|embed|applet|form|input|textarea|button|select)\b[^>]*/?\s*>',
+            '', html_str, flags=re.IGNORECASE
+        )
+        html_str = re.sub(r'\s+on\w+\s*=\s*"[^"]*"', '', html_str, flags=re.IGNORECASE)
+        html_str = re.sub(r"\s+on\w+\s*=\s*'[^']*'", '', html_str, flags=re.IGNORECASE)
+        html_str = re.sub(r'\s+on\w+\s*=\s*[^\s>]+', '', html_str, flags=re.IGNORECASE)
+        html_str = re.sub(
+            r'href\s*=\s*["\']?\s*javascript:', 'href="#', html_str, flags=re.IGNORECASE
+        )
+        return html_str
+
     @staticmethod
     def _normalize_cooked_urls(cooked_html: str) -> str:
         """将 cooked 中的相对资源 URL 转绝对 URL，剥离轻臾框包裹与 meta 信息
@@ -1000,73 +1027,63 @@ class LinuxDoPreviewPlugin(Star):
         if not cooked_html:
             return ""
         try:
-            import re as _re
             # 1) 绝对化 src/href（相对与协议无关 URL）
-            cooked_html = _re.sub(
+            cooked_html = re.sub(
                 r'(src|href)="(//[^"]+)"',
                 r'\1="https:\2',
                 cooked_html,
             )
-            cooked_html = _re.sub(
+            cooked_html = re.sub(
                 r'(src|href)="(/uploads/[^"]+)"',
                 r'\1="https://linux.do\2',
                 cooked_html,
             )
 
-            # 2) 整块剥离 lightbox-wrapper：仅保留内部 <img>，丢弃其余
+            # 2) 使用 lxml 处理嵌套 div 结构（lightbox-wrapper、meta 等）
-            def _pick_imgs(block: str) -> str:
+            if _lh is not None:
-                imgs = _re.findall(r'<img\b[^>]*>', block, flags=_re.IGNORECASE)
+                try:
-                return "".join(imgs)
+                    tree = _lh.fromstring(f"<root>{cooked_html}</root>")
 
-            cooked_html = _re.sub(
+                    # lightbox-wrapper：只保留内部 <img>，丢弃其余
-                r'<div[^>]*class="[^"]*lightbox-wrapper[^"]*"[^>]*>(.*?)</div>',
+                    for wrapper in tree.cssselect(".lightbox-wrapper"):
-                lambda m: _pick_imgs(m.group(1)),
+                        imgs = wrapper.cssselect("img")
-                cooked_html,
+                        parent = wrapper.getparent()
-                flags=_re.DOTALL,
+                        if parent is None:
-            )
+                            continue
+                        idx = list(parent).index(wrapper)
+                        for i, img in enumerate(imgs):
+                            parent.insert(idx + i, img)
+                        parent.remove(wrapper)
 
-            # 3) 退路：直接裸的 <a class="lightbox"> 包裹，剥 a、保留 img
+                    # 裸 <a class="lightbox">：剥 a、保留子元素
-            cooked_html = _re.sub(
+                    for a in tree.cssselect("a.lightbox"):
+                        parent = a.getparent()
+                        if parent is None:
+                            continue
+                        idx = list(parent).index(a)
+                        for i, child in enumerate(list(a)):
+                            parent.insert(idx + i, child)
+                        parent.remove(a)
+
+                    # 删除 meta 信息块、文件名、代码块工具栏、download 按钮
+                    for sel in [".meta", ".filename", ".codeblock-buttons",
+                                ".pre-actions", "a.download"]:
+                        for el in tree.cssselect(sel):
+                            el.getparent().remove(el)
+
+                    cooked_html = "".join(
+                        str(_lh.tostring(child, encoding="unicode"))
+                        for child in tree
+                    )
+                except Exception:
+                    pass
+
+            # 3) 正则回退：处理 lxml 未覆盖的残留
+            cooked_html = re.sub(
                 r'<a [^>]*class="[^"]*\blightbox\b[^"]*"[^>]*>(.*?)</a>',
                 r'\1',
                 cooked_html,
-                flags=_re.DOTALL,
+                flags=re.DOTALL,
-            )
-
-            # 4) 删除所有残留的 meta 信息块（文件尺寸、文件名、下载按钮等）
-            cooked_html = _re.sub(
-                r'<div[^>]*class="[^"]*\bmeta\b[^"]*"[^>]*>.*?</div>',
-                '',
-                cooked_html,
-                flags=_re.DOTALL,
-            )
-            cooked_html = _re.sub(
-                r'<span[^>]*class="[^"]*\bfilename\b[^"]*"[^>]*>.*?</span>',
-                '',
-                cooked_html,
-                flags=_re.DOTALL,
-            )
-
-            # 5) 删除代码块顶部的工具栏（copy/undo 按钮）防止占位
-            cooked_html = _re.sub(
-                r'<div[^>]*class="[^"]*\bcodeblock-buttons\b[^"]*"[^>]*>.*?</div>',
-                '',
-                cooked_html,
-                flags=_re.DOTALL,
-            )
-            cooked_html = _re.sub(
-                r'<pre[^>]*>\s*<div[^>]*class="[^"]*\bpre-actions\b[^"]*"[^>]*>.*?</div>',
-                '<pre>',
-                cooked_html,
-                flags=_re.DOTALL,
-            )
-
-            # 6) 删除 download 按钮、悬浮提示等装饰
-            cooked_html = _re.sub(
-                r'<a[^>]*class="[^"]*\bdownload[^"]*"[^>]*>.*?</a>',
-                '',
-                cooked_html,
-                flags=_re.DOTALL,
             )
 
         except Exception:
@@ -1087,66 +1104,68 @@ class LinuxDoPreviewPlugin(Star):
             if not ctx:
                 return None
             page = ctx.new_page()
-            page.set_viewport_size({"width": 820, "height": 1200})
-
-            # 设置内容，等待图片资源加载
-            page.set_content(html, wait_until="domcontentloaded", timeout=timeout_ms)
-
-            # 主动等所有 <img> 加载完成（最多 3s），并剔除加载失败的图
-            page.evaluate("""() => new Promise(resolve => {
-                const imgs = document.querySelectorAll('img');
-                if (!imgs.length) return resolve();
-                let done = 0;
-                const tick = (img) => {
-                    done++;
-                    // 图加载失败：移除 <img> 避免占位巨大空白
-                    if (img.complete && img.naturalWidth === 0) {
-                        img.remove();
-                    }
-                    if (done >= imgs.length) resolve();
-                };
-                imgs.forEach(img => {
-                    if (img.complete) tick(img);
-                    else {
-                        img.addEventListener('load', () => tick(img), { once: true });
-                        img.addEventListener('error', () => tick(img), { once: true });
-                    }
-                });
-                setTimeout(resolve, 3000);
-            })""")
-
-            page.wait_for_timeout(300)
-
-            # ── 自适应截图：总是优先对 .card 元素截图，按内容实际边界拍 ──
-            # 元素截图零空白、零截断，不受 viewport 高度限制。
-            # `screenshot_full_page` 仅作为后备回退：元素截图失败时才使用。
-            card_locator = page.locator(".card")
-            full_page = self.config.get("screenshot_full_page", True)
             try:
-                if card_locator.count() > 0:
+                page.set_viewport_size({"width": 820, "height": 1200})
-                    card_locator.first.screenshot(
+
-                        path=str(save_path),
+                # 设置内容，等待图片资源加载
-                        timeout=timeout_ms,
+                page.set_content(html, wait_until="domcontentloaded", timeout=timeout_ms)
-                    )
+
-                else:
+                # 主动等所有 <img> 加载完成（最多 3s），并剔除加载失败的图
+                page.evaluate("""() => new Promise(resolve => {
+                    const imgs = document.querySelectorAll('img');
+                    if (!imgs.length) return resolve();
+                    let done = 0;
+                    const tick = (img) => {
+                        done++;
+                        // 图加载失败：移除 <img> 避免占位巨大空白
+                        if (img.complete && img.naturalWidth === 0) {
+                            img.remove();
+                        }
+                        if (done >= imgs.length) resolve();
+                    };
+                    imgs.forEach(img => {
+                        if (img.complete) tick(img);
+                        else {
+                            img.addEventListener('load', () => tick(img), { once: true });
+                            img.addEventListener('error', () => tick(img), { once: true });
+                        }
+                    });
+                    setTimeout(resolve, 3000);
+                })""")
+
+                page.wait_for_timeout(300)
+
+                # ── 自适应截图：总是优先对 .card 元素截图，按内容实际边界拍 ──
+                # 元素截图零空白、零截断，不受 viewport 高度限制。
+                # `screenshot_full_page` 仅作为后备回退：元素截图失败时才使用。
+                card_locator = page.locator(".card")
+                full_page = self.config.get("screenshot_full_page", True)
+                try:
+                    if card_locator.count() > 0:
+                        card_locator.first.screenshot(
+                            path=str(save_path),
+                            timeout=timeout_ms,
+                        )
+                    else:
+                        page.screenshot(
+                            path=str(save_path),
+                            full_page=full_page,
+                            timeout=timeout_ms,
+                        )
+                except Exception:
+                    # 回退：若元素截图失败（少见），退到全页截图
                     page.screenshot(
                         path=str(save_path),
                         full_page=full_page,
                         timeout=timeout_ms,
                     )
-            except Exception:
+                sz = save_path.stat().st_size
-                # 回退：若元素截图失败（少见），退到全页截图
+                logger.info(
-                page.screenshot(
+                    f"[LinuxDoPreview] 渲染截图: {save_path.name} ({sz / 1024:.1f} KB)"
-                    path=str(save_path),
-                    full_page=full_page,
-                    timeout=timeout_ms,
                 )
-            sz = save_path.stat().st_size
+                return save_path
-            logger.info(
+            finally:
-                f"[LinuxDoPreview] 渲染截图: {save_path.name} ({sz / 1024:.1f} KB)"
+                page.close()
-            )
-            page.close()
-            return save_path
         except Exception as e:
             logger.warning(f"[LinuxDoPreview] HTML 渲染失败: {type(e).__name__}: {e}")
             return None