dsh-pocket 2.1.3 → 2.1.4

package/README.en.md

@@ -191,7 +191,7 @@ Such tools take over all traffic and often cut cloudflared's tunnel-edge connect
 ```sh
 npm install
 node client/build.mjs   # rebuild after editing client/
-npm test                # proxy / auth / compression / tunnel / service / RPC / settings (91 tests)
+npm test                # proxy / auth / compression / tunnel / service / RPC / settings (93 tests)
 ```
 
 **Want to try your changes locally without publishing?** Point the installed plugin at your local checkout with a symlink and restart dsh web. Full steps (including switching back to the npm release) are in [LOCAL-DEV.md](./LOCAL-DEV.md).

package/README.md

@@ -192,7 +192,7 @@ npx @deepseek-ai/dsh web
 ```sh
 npm install
 node client/build.mjs   # 改 client/ 后重新打包
-npm test                # 代理 / 认证 / 压缩 / 隧道 / 服务 / RPC / 设置（91 测试）
+npm test                # 代理 / 认证 / 压缩 / 隧道 / 服务 / RPC / 设置（93 测试）
 ```
 
 **改完想在本机先试？** 不用发版：把插件换成指向本地仓库的软链，重启 dsh web 就是本地代码。完整步骤（含怎么换回 npm 官方版本）见 [LOCAL-DEV.md](./LOCAL-DEV.md)。

package/client/client.js

@@ -535,14 +535,19 @@ var MOBILE_CSS = `
      it to 500), and when the layer outranks the drawer, the backdrop paints
      ABOVE the drawer and swallows every tap \u2014 the drawer opens but no row
      can be pressed (every tap just closes it). The drawer must therefore
-     outrank any such raise: 600 clears the known 500 while staying under the
-     fixed-position banners/toasts (z 9999) that float at the viewport level. */
+     outrank any such raise.
+     1200 (was 600) clears the mobile layers shipped by
+     @linxin666/dsh-web-ui-all \u2014 its sidebar pane is z-index 1100, its
+     details pane 1000 and its full-screen frame ::after mask 1050 (issue
+     #67: that mask sat on top of the 600 drawer and ate every tap). Still
+     far under the fixed-position banners/toasts (z 9999) that float at the
+     viewport level. */
   [data-mobile-nav="frame"] > :first-child {
     position: absolute !important;
     inset: 0 auto 0 0 !important;
     width: max-content !important;
     max-width: 92vw !important;
-    z-index: 600 !important;
+    z-index: 1200 !important;
     transform: translateX(-110%);
     transition: transform .28s var(--ds-ease-in-out, ease-in-out);
     background: var(--dsw-alias-bg-base, #ffffff);
@@ -571,6 +576,32 @@ var MOBILE_CSS = `
     transform: none !important;
   }
 
+  /* Kill a competing full-screen mask (issue #67).
+     @linxin666/dsh-web-ui-all ships its own mobile drawer, and part of it is
+
+       [data-dsh-frame]:not([data-sidebar-collapsed])::after {
+         content: ""; position: fixed; inset: 0; z-index: 1050;
+         background: rgb(0 0 0 / 24%);
+       }
+
+     The pseudo-element belongs to the frame we already mark, and the frame
+     carries only "position: relative" with z-index auto \u2014 no stacking context
+     \u2014 so this mask competes with the drawer in the parent stacking context
+     and, at 1050, paints over it. It covers the whole viewport, so every tap
+     on a session row lands on the mask instead: the drawer opens but nothing
+     inside it can be pressed, and the page behind cannot be scrolled.
+     Removing it is safe: the mobile stylesheet already renders its own
+     backdrop, and tapping outside the drawer is handled in JS.
+
+     The attribute selector is repeated on purpose. Their rule has the same
+     specificity (0,2,1) once ours is written the obvious way, and plugin
+     stylesheets are injected in load order, so a tie would be decided by
+     whichever plugin happened to load last. Doubling the attribute makes it
+     (0,3,1) and deterministic. */
+  [data-mobile-nav="frame"][data-mobile-nav="frame"]:not([data-sidebar-collapsed])::after {
+    content: none !important;
+  }
+
   /* Drag handles are useless on touch and would float over the drawer. */
   [data-side="sidebar"],
   [data-side="details"] {

package/client/mobile/mobile.css.ts

@@ -210,14 +210,19 @@ export const MOBILE_CSS = `
      it to 500), and when the layer outranks the drawer, the backdrop paints
      ABOVE the drawer and swallows every tap — the drawer opens but no row
      can be pressed (every tap just closes it). The drawer must therefore
-     outrank any such raise: 600 clears the known 500 while staying under the
-     fixed-position banners/toasts (z 9999) that float at the viewport level. */
+     outrank any such raise.
+     1200 (was 600) clears the mobile layers shipped by
+     @linxin666/dsh-web-ui-all — its sidebar pane is z-index 1100, its
+     details pane 1000 and its full-screen frame ::after mask 1050 (issue
+     #67: that mask sat on top of the 600 drawer and ate every tap). Still
+     far under the fixed-position banners/toasts (z 9999) that float at the
+     viewport level. */
   [data-mobile-nav="frame"] > :first-child {
     position: absolute !important;
     inset: 0 auto 0 0 !important;
     width: max-content !important;
     max-width: 92vw !important;
-    z-index: 600 !important;
+    z-index: 1200 !important;
     transform: translateX(-110%);
     transition: transform .28s var(--ds-ease-in-out, ease-in-out);
     background: var(--dsw-alias-bg-base, #ffffff);
@@ -246,6 +251,32 @@ export const MOBILE_CSS = `
     transform: none !important;
   }
 
+  /* Kill a competing full-screen mask (issue #67).
+     @linxin666/dsh-web-ui-all ships its own mobile drawer, and part of it is
+
+       [data-dsh-frame]:not([data-sidebar-collapsed])::after {
+         content: ""; position: fixed; inset: 0; z-index: 1050;
+         background: rgb(0 0 0 / 24%);
+       }
+
+     The pseudo-element belongs to the frame we already mark, and the frame
+     carries only "position: relative" with z-index auto — no stacking context
+     — so this mask competes with the drawer in the parent stacking context
+     and, at 1050, paints over it. It covers the whole viewport, so every tap
+     on a session row lands on the mask instead: the drawer opens but nothing
+     inside it can be pressed, and the page behind cannot be scrolled.
+     Removing it is safe: the mobile stylesheet already renders its own
+     backdrop, and tapping outside the drawer is handled in JS.
+
+     The attribute selector is repeated on purpose. Their rule has the same
+     specificity (0,2,1) once ours is written the obvious way, and plugin
+     stylesheets are injected in load order, so a tie would be decided by
+     whichever plugin happened to load last. Doubling the attribute makes it
+     (0,3,1) and deterministic. */
+  [data-mobile-nav="frame"][data-mobile-nav="frame"]:not([data-sidebar-collapsed])::after {
+    content: none !important;
+  }
+
   /* Drag handles are useless on touch and would float over the drawer. */
   [data-side="sidebar"],
   [data-side="details"] {

package/package.json

@@ -80,5 +80,5 @@
     "access": "public",
     "registry": "https://registry.npmjs.org/"
   },
-  "version": "2.1.3"
+  "version": "2.1.4"
 }