bunite-core 0.14.0 → 0.16.0

package/src/native/win-webview2/bunite_webview2_ffi.cpp

@@ -1,6 +1,10 @@
 #include "webview2_internal.h"
 
+#include <algorithm>
+#include <array>
 #include <cstring>
+#include <memory>
+#include <vector>
 #include <wincrypt.h>  // CryptBinaryToStringA — base64 encoding for screenshot payload.
 
 using namespace bunite_webview2;
@@ -12,7 +16,7 @@ void setViewInputPassthrough(ViewHost* v, bool passthrough);
 
 extern "C" {
 
-BUNITE_EXPORT int32_t bunite_abi_version(void) { return 9; }
+BUNITE_EXPORT int32_t bunite_abi_version(void) { return 11; }
 
 BUNITE_EXPORT void bunite_set_log_level(int32_t level) {
   if (level < 0) level = 0;
@@ -379,7 +383,8 @@ BUNITE_EXPORT void bunite_view_reload(uint32_t view_id) {
 BUNITE_EXPORT void bunite_view_remove(uint32_t view_id) { destroyView(view_id); }
 
 // Input dispatch — CDP via CallDevToolsProtocolMethod (Playwright pattern).
-// MouseEvent.isTrusted is false (CDP-synthesized); capability honest.
+// Edge runtime injects below DevTools surface — DOM `isTrusted=true`
+// (see `bunite_view_capabilities` note).
 namespace {
 
 const char* cdpMouseButton(int32_t b) {
@@ -392,6 +397,23 @@ void cdpCall(ViewHost* v, const wchar_t* method, const std::string& json) {
       method, utf8ToWide(json).c_str(), nullptr);
 }
 
+// Result-aware CDP call. `cb(ok, json)` runs on the UI thread once the result
+// arrives; on failure `ok=false` and json is an error description.
+void cdpCallWithResult(ViewHost* v, const wchar_t* method, const std::string& json,
+                       std::function<void(bool, std::string)> cb) {
+  if (!v || !v->webview) { if (cb) cb(false, "view not ready"); return; }
+  auto lifetime = g_runtime.lifetime;
+  v->webview->CallDevToolsProtocolMethod(
+      method, utf8ToWide(json).c_str(),
+      Microsoft::WRL::Callback<ICoreWebView2CallDevToolsProtocolMethodCompletedHandler>(
+          [lifetime, cb](HRESULT hr, LPCWSTR result) -> HRESULT {
+            if (!lifetime || !lifetime->alive.load()) return S_OK;
+            if (FAILED(hr) || !result) { cb(false, "CDP call failed"); return S_OK; }
+            cb(true, wideToUtf8(result));
+            return S_OK;
+          }).Get());
+}
+
 }  // namespace
 
 BUNITE_EXPORT void bunite_view_click(uint32_t view_id, double x, double y,
@@ -428,6 +450,10 @@ BUNITE_EXPORT void bunite_view_press(uint32_t view_id, int32_t windows_vk_code,
   std::string key_str = key ? key : "";
   std::string code_str = code ? code : "";
 
+  // CDP modifier mask: 1=alt, 2=ctrl, 4=meta, 8=shift. Suppress text when any
+  // non-shift modifier is set so shortcuts like Ctrl+A don't insert "a".
+  const bool has_non_shift_modifier = (modifiers & ~static_cast<uint32_t>(8)) != 0;
+
   auto buildPart = [&](const char* type, bool include_text) {
     std::string out = "{\"type\":\"";
     out += type;
@@ -437,7 +463,7 @@ BUNITE_EXPORT void bunite_view_press(uint32_t view_id, int32_t windows_vk_code,
     if (!code_str.empty()) out += ",\"code\":\"" + escapeJsonString(code_str) + "\"";
     // CDP `location`: 0 standard, 1 left mod, 2 right mod, 3 numpad.
     if (location > 0) out += ",\"location\":" + std::to_string(location);
-    if (include_text && !char_str.empty())
+    if (include_text && !char_str.empty() && !has_non_shift_modifier)
       out += ",\"text\":\"" + escapeJsonString(char_str) + "\"";
     out += "}";
     return out;
@@ -485,7 +511,765 @@ BUNITE_EXPORT uint32_t bunite_view_capabilities(uint32_t view_id) {
          BUNITE_CAP_NATIVE_INPUT_TRUSTED |
          BUNITE_CAP_CLICK | BUNITE_CAP_TYPE | BUNITE_CAP_PRESS | BUNITE_CAP_SCROLL |
          BUNITE_CAP_MOUSE | BUNITE_CAP_DIALOGS | BUNITE_CAP_CONSOLE |
-         BUNITE_CAP_SCREENSHOT | BUNITE_CAP_FORMAT_PNG | BUNITE_CAP_FORMAT_JPEG;
+         BUNITE_CAP_SCREENSHOT | BUNITE_CAP_FORMAT_PNG | BUNITE_CAP_FORMAT_JPEG |
+         BUNITE_CAP_AX | BUNITE_CAP_BOUNDING_RECT | BUNITE_CAP_FRAMES |
+         BUNITE_CAP_DOWNLOADS | BUNITE_CAP_POPUPS |
+         BUNITE_CAP_RESOLVE_AND_CLICK;
+}
+
+BUNITE_EXPORT void bunite_view_set_download_policy(uint32_t view_id, int32_t policy, const char* download_dir) {
+  ViewHost* v = getView(view_id);
+  if (!v) return;
+  if (policy < 0 || policy > 2) policy = 2;
+  v->download_policy.store(policy);
+  v->download_dir = download_dir ? download_dir : "";
+}
+
+BUNITE_EXPORT void bunite_view_popup_accept(uint32_t new_view_id, uint32_t host_window_id,
+                                              double x, double y, double w, double h) {
+  ViewHost* v = getView(new_view_id);
+  if (!v || !v->controller || !v->container_hwnd) return;
+  WindowHost* host_window = nullptr;
+  {
+    std::lock_guard<std::mutex> g(g_runtime.object_mutex);
+    auto wit = g_runtime.windows_by_id.find(host_window_id);
+    if (wit == g_runtime.windows_by_id.end()) return;
+    host_window = wit->second;
+  }
+  if (!host_window || !host_window->hwnd) return;
+  v->window = host_window;
+  host_window->views.push_back(v);
+  SetParent(v->container_hwnd, host_window->hwnd);
+  MoveWindow(v->container_hwnd, static_cast<int>(x), static_cast<int>(y),
+             static_cast<int>(w), static_cast<int>(h), TRUE);
+  ShowWindow(v->container_hwnd, SW_SHOW);
+  RECT bounds{0, 0, static_cast<LONG>(w), static_cast<LONG>(h)};
+  v->controller->put_Bounds(bounds);
+  v->controller->put_IsVisible(TRUE);
+  emitWebviewEvent(new_view_id, "view-ready", "");
+}
+
+BUNITE_EXPORT void bunite_view_popup_dismiss(uint32_t new_view_id) {
+  ViewHost* v = nullptr;
+  {
+    std::lock_guard<std::mutex> g(g_runtime.object_mutex);
+    auto it = g_runtime.views_by_id.find(new_view_id);
+    if (it == g_runtime.views_by_id.end()) return;
+    v = it->second;
+    g_runtime.views_by_id.erase(it);
+  }
+  if (!v) return;
+  if (v->controller) v->controller->Close();
+  if (v->container_hwnd) DestroyWindow(v->container_hwnd);
+  delete v;
+}
+
+namespace {
+void emitAxError(uint32_t view_id, uint32_t request_id, const char* code, const std::string& message) {
+  std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                        ",\"ok\":false,\"code\":\"" + code +
+                        "\",\"message\":\"" + escapeJsonString(message) + "\"}";
+  emitWebviewEvent(view_id, "accessibility-result", payload);
+}
+}  // namespace
+
+namespace {
+// Extract integer field from a CDP JSON response — used to pluck
+// `executionContextId` without pulling in a JSON parser.
+bool extractJsonInt(const std::string& json, const std::string& key, int& out) {
+  std::string needle = "\"" + key + "\":";
+  auto p = json.find(needle);
+  if (p == std::string::npos) return false;
+  p += needle.size();
+  while (p < json.size() && (json[p] == ' ' || json[p] == '\t')) ++p;
+  auto end = p;
+  while (end < json.size() && (json[end] == '-' || (json[end] >= '0' && json[end] <= '9'))) ++end;
+  if (end == p) return false;
+  out = std::stoi(json.substr(p, end - p));
+  return true;
+}
+void emitListFramesErrorWv2(uint32_t view_id, uint32_t request_id, const char* code, const std::string& message) {
+  std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                        ",\"ok\":false,\"code\":\"" + code +
+                        "\",\"message\":\"" + escapeJsonString(message) + "\"}";
+  emitWebviewEvent(view_id, "list-frames-result", payload);
+}
+}  // namespace
+
+BUNITE_EXPORT void bunite_view_list_frames(uint32_t view_id, uint32_t request_id) {
+  ViewHost* v = getView(view_id);
+  if (!v || !v->webview) { emitListFramesErrorWv2(view_id, request_id, "not_supported", "view not ready"); return; }
+  cdpCallWithResult(v, L"Page.getFrameTree", "{}",
+      [view_id, request_id](bool ok, std::string result) {
+        if (!ok) { emitListFramesErrorWv2(view_id, request_id, "runtime_error", "getFrameTree failed: " + result); return; }
+        std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                              ",\"ok\":true,\"raw\":" + result + "}";
+        emitWebviewEvent(view_id, "list-frames-result", payload);
+      });
+}
+
+BUNITE_EXPORT void bunite_view_evaluate_in_frame(uint32_t view_id, uint32_t request_id,
+                                                  const char* script_c, const char* frame_id_c) {
+  std::string script = script_c ? script_c : "";
+  std::string frameId = frame_id_c ? frame_id_c : "";
+  if (frameId.empty()) {
+    bunite_view_evaluate(view_id, request_id, script_c);
+    return;
+  }
+  ViewHost* v = getView(view_id);
+  if (!v || !v->webview) {
+    std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                          ",\"ok\":false,\"code\":\"not_supported\",\"message\":\"view not ready\"}";
+    emitWebviewEvent(view_id, "evaluate-result", payload);
+    return;
+  }
+  std::string isoParams = "{\"frameId\":\"" + escapeJsonString(frameId) + "\",\"worldName\":\"bunite-eval\"}";
+  cdpCallWithResult(v, L"Page.createIsolatedWorld", isoParams,
+      [view_id, request_id, script](bool ok, std::string isoResult) {
+        if (!ok) {
+          std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                                ",\"ok\":false,\"code\":\"runtime_error\","
+                                "\"message\":\"createIsolatedWorld failed\"}";
+          emitWebviewEvent(view_id, "evaluate-result", payload);
+          return;
+        }
+        int contextId = 0;
+        if (!extractJsonInt(isoResult, "executionContextId", contextId)) {
+          std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                                ",\"ok\":false,\"code\":\"runtime_error\","
+                                "\"message\":\"missing executionContextId\"}";
+          emitWebviewEvent(view_id, "evaluate-result", payload);
+          return;
+        }
+        // Re-lookup — the async gap could have torn down the view.
+        ViewHost* v2 = getView(view_id);
+        if (!v2 || !v2->webview) {
+          std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                                ",\"ok\":false,\"code\":\"not_supported\","
+                                "\"message\":\"view destroyed\"}";
+          emitWebviewEvent(view_id, "evaluate-result", payload);
+          return;
+        }
+        std::string evalParams = "{\"expression\":\"" + escapeJsonString(script) +
+                                 "\",\"contextId\":" + std::to_string(contextId) +
+                                 ",\"returnByValue\":true,\"awaitPromise\":true}";
+        cdpCallWithResult(v2, L"Runtime.evaluate", evalParams,
+            [view_id, request_id](bool ok2, std::string evalResult) {
+              if (!ok2) {
+                std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                                      ",\"ok\":false,\"code\":\"runtime_error\","
+                                      "\"message\":\"Runtime.evaluate failed\"}";
+                emitWebviewEvent(view_id, "evaluate-result", payload);
+                return;
+              }
+              // Normalize CDP shape to the flat `{requestId, ok, value/code/message}`
+              // that the host's evaluate-result handler expects. exceptionDetails
+              // at the top level is preceded by `},` — distinguishes from a value
+              // that happens to contain the literal token inside a string.
+              auto excPos = evalResult.find("},\"exceptionDetails\"");
+              if (excPos == std::string::npos) excPos = evalResult.find("}, \"exceptionDetails\"");
+              if (excPos != std::string::npos) {
+                std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                                      ",\"ok\":false,\"code\":\"runtime_error\","
+                                      "\"message\":\"evaluate threw\"}";
+                emitWebviewEvent(view_id, "evaluate-result", payload);
+                return;
+              }
+              // Extract result.value as a JSON substring. CDP returns
+              // `{"result":{"type":"...","value":<json>}}`. We find the `"value":`
+              // token inside the inner object and slice until the matching close.
+              auto resPos = evalResult.find("\"result\":");
+              if (resPos == std::string::npos) {
+                std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                                      ",\"ok\":true,\"value\":\"null\"}";
+                emitWebviewEvent(view_id, "evaluate-result", payload);
+                return;
+              }
+              auto valKey = evalResult.find("\"value\":", resPos);
+              if (valKey == std::string::npos) {
+                // type === "undefined" — no value field.
+                std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                                      ",\"ok\":true,\"value\":\"null\"}";
+                emitWebviewEvent(view_id, "evaluate-result", payload);
+                return;
+              }
+              size_t start = valKey + 8;  // past `"value":`
+              // Skip whitespace.
+              while (start < evalResult.size() && (evalResult[start] == ' ' || evalResult[start] == '\t')) ++start;
+              // Find balanced end of the JSON value.
+              size_t end = start;
+              if (start < evalResult.size()) {
+                char c = evalResult[start];
+                if (c == '"') {
+                  ++end;
+                  while (end < evalResult.size() && evalResult[end] != '"') {
+                    if (evalResult[end] == '\\' && end + 1 < evalResult.size()) ++end;
+                    ++end;
+                  }
+                  if (end < evalResult.size()) ++end;
+                } else if (c == '{' || c == '[') {
+                  int depth = 0;
+                  bool inStr = false;
+                  while (end < evalResult.size()) {
+                    char ch = evalResult[end];
+                    if (inStr) {
+                      if (ch == '\\' && end + 1 < evalResult.size()) ++end;
+                      else if (ch == '"') inStr = false;
+                    } else if (ch == '"') inStr = true;
+                    else if (ch == '{' || ch == '[') ++depth;
+                    else if (ch == '}' || ch == ']') { --depth; if (depth == 0) { ++end; break; } }
+                    ++end;
+                  }
+                } else {
+                  // Number / true / false / null — read until non-token char.
+                  while (end < evalResult.size()) {
+                    char ch = evalResult[end];
+                    if (ch == ',' || ch == '}' || ch == ' ' || ch == '\t' || ch == '\n') break;
+                    ++end;
+                  }
+                }
+              }
+              std::string valueJson = evalResult.substr(start, end - start);
+              std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                                    ",\"ok\":true,\"value\":\"" + escapeJsonString(valueJson) + "\"}";
+              emitWebviewEvent(view_id, "evaluate-result", payload);
+            });
+      });
+}
+
+}  // extern "C" — exit so the helpers below can return std::string.
+
+namespace {
+
+void emitResolveAndClickErrorWv2(uint32_t view_id, uint32_t request_id, const char* code, const std::string& message) {
+  std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                        ",\"ok\":false,\"code\":\"" + code +
+                        "\",\"message\":\"" + escapeJsonString(message) + "\"}";
+  emitWebviewEvent(view_id, "resolve-and-click-result", payload);
+}
+
+const char* cdpButtonNameWv2(int32_t b) {
+  switch (b) { case 1: return "middle"; case 2: return "right"; default: return "left"; }
+}
+
+std::string extractJsonValueRaw(const std::string& s, const std::string& key) {
+  std::string needle = "\"" + key + "\":";
+  size_t p = s.find(needle);
+  if (p == std::string::npos) return {};
+  p += needle.size();
+  while (p < s.size() && (s[p] == ' ' || s[p] == '\t')) ++p;
+  if (p >= s.size()) return {};
+  size_t end = p;
+  char c = s[p];
+  if (c == '"') {
+    ++end;
+    while (end < s.size() && s[end] != '"') { if (s[end] == '\\' && end + 1 < s.size()) ++end; ++end; }
+    if (end < s.size()) ++end;
+  } else if (c == '{' || c == '[') {
+    int depth = 0; bool inStr = false;
+    while (end < s.size()) {
+      char ch = s[end];
+      if (inStr) { if (ch == '\\' && end + 1 < s.size()) ++end; else if (ch == '"') inStr = false; }
+      else if (ch == '"') inStr = true;
+      else if (ch == '{' || ch == '[') ++depth;
+      else if (ch == '}' || ch == ']') { --depth; if (depth == 0) { ++end; break; } }
+      ++end;
+    }
+  } else {
+    while (end < s.size()) {
+      char ch = s[end];
+      if (ch == ',' || ch == '}' || ch == ' ' || ch == '\t' || ch == '\n') break;
+      ++end;
+    }
+  }
+  return s.substr(p, end - p);
+}
+
+double extractJsonDouble(const std::string& s, const std::string& key, double dflt = 0.0) {
+  std::string raw = extractJsonValueRaw(s, key);
+  if (raw.empty()) return dflt;
+  try { return std::stod(raw); } catch (...) { return dflt; }
+}
+
+std::string extractJsonStringDecoded(const std::string& s, const std::string& key) {
+  std::string raw = extractJsonValueRaw(s, key);
+  if (raw.size() < 2 || raw.front() != '"' || raw.back() != '"') return {};
+  std::string out; out.reserve(raw.size());
+  for (size_t i = 1; i + 1 < raw.size(); ++i) {
+    if (raw[i] == '\\' && i + 1 < raw.size() - 1) {
+      char nxt = raw[++i];
+      switch (nxt) {
+        case '"': out += '"'; break;
+        case '\\': out += '\\'; break;
+        case 'n': out += '\n'; break;
+        case 'r': out += '\r'; break;
+        case 't': out += '\t'; break;
+        case '/': out += '/'; break;
+        default: out += nxt; break;
+      }
+    } else out += raw[i];
+  }
+  return out;
+}
+
+std::string escapeForJsStringWv2(const std::string& s) {
+  std::string out; out.reserve(s.size() + 2);
+  for (char c : s) {
+    if (c == '"' || c == '\\') { out.push_back('\\'); out.push_back(c); }
+    else if (c == '\n') out += "\\n";
+    else if (c == '\r') out += "\\r";
+    else if (c == '\t') out += "\\t";
+    else out.push_back(c);
+  }
+  return out;
+}
+
+std::string buildResolveScriptWv2(const std::string& selector) {
+  // Frame-local rect + innerWidth/innerHeight for bilinear mapping when the
+  // frame is transformed (rotate/scale). Main frame uses iw/ih harmlessly.
+  std::string sel_lit = "\"" + escapeForJsStringWv2(selector) + "\"";
+  return
+    "(function(){"
+      "var el=document.querySelector(" + sel_lit + ");"
+      "if(!el)return{ok:false,code:\"not_found\"};"
+      "el.scrollIntoView({block:\"nearest\",inline:\"nearest\",behavior:\"instant\"});"
+      "var r=el.getBoundingClientRect();"
+      "var vis=r.width>0&&r.height>0&&r.bottom>0&&r.right>0"
+      "&&r.top<innerHeight&&r.left<innerWidth;"
+      "if(!vis)return{ok:false,code:\"not_visible\"};"
+      "return{ok:true,x:r.x,y:r.y,w:r.width,h:r.height,"
+      "cx:r.x+r.width/2,cy:r.y+r.height/2,"
+      "iw:innerWidth,ih:innerHeight};"
+    "})()";
+}
+
+void dispatchCdpClickWv2(ViewHost* v, double cx, double cy,
+                          int32_t button, int32_t click_count, uint32_t modifiers) {
+  if (click_count < 1) click_count = 1;
+  const char* btn = cdpButtonNameWv2(button);
+  for (int i = 1; i <= click_count; ++i) {
+    std::string base = "\"x\":" + std::to_string(cx) + ",\"y\":" + std::to_string(cy) +
+                       ",\"button\":\"" + btn + "\",\"clickCount\":" + std::to_string(i) +
+                       ",\"modifiers\":" + std::to_string(modifiers);
+    cdpCall(v, L"Input.dispatchMouseEvent", "{\"type\":\"mousePressed\"," + base + "}");
+    cdpCall(v, L"Input.dispatchMouseEvent", "{\"type\":\"mouseReleased\"," + base + "}");
+  }
+}
+
+// CDP call routed to a child target session (flatten:true). Requires
+// ICoreWebView2_11.
+void cdpCallForSession(ViewHost* v, const std::string& session_id, const wchar_t* method,
+                        const std::string& params_json,
+                        std::function<void(bool, std::string)> cb) {
+  if (!v || !v->webview) { if (cb) cb(false, "view not ready"); return; }
+  ComPtr<ICoreWebView2_11> wv11;
+  if (FAILED(v->webview.As(&wv11)) || !wv11) { if (cb) cb(false, "ICoreWebView2_11 unavailable"); return; }
+  auto lifetime = g_runtime.lifetime;
+  wv11->CallDevToolsProtocolMethodForSession(
+      utf8ToWide(session_id).c_str(), method, utf8ToWide(params_json).c_str(),
+      Microsoft::WRL::Callback<ICoreWebView2CallDevToolsProtocolMethodCompletedHandler>(
+          [lifetime, cb](HRESULT hr, LPCWSTR result) -> HRESULT {
+            if (!lifetime || !lifetime->alive.load()) return S_OK;
+            if (FAILED(hr) || !result) { cb(false, "CDP-for-session call failed"); return S_OK; }
+            cb(true, wideToUtf8(result));
+            return S_OK;
+          }).Get());
+}
+
+// Register one-time listeners for Target.attachedToTarget / detachedFromTarget.
+// Populates view->oopif_sessions as OOPIF child sessions attach.
+void armOopifEvents(ViewHost* v) {
+  if (v->oopif_event_tokens_registered) return;
+  using namespace Microsoft::WRL;
+  ComPtr<ICoreWebView2DevToolsProtocolEventReceiver> attached_r, detached_r;
+  if (FAILED(v->webview->GetDevToolsProtocolEventReceiver(L"Target.attachedToTarget", &attached_r))
+   || FAILED(v->webview->GetDevToolsProtocolEventReceiver(L"Target.detachedFromTarget", &detached_r))) {
+    return;
+  }
+  uint32_t view_id = v->id;
+  attached_r->add_DevToolsProtocolEventReceived(
+      Callback<ICoreWebView2DevToolsProtocolEventReceivedEventHandler>(
+          [view_id](ICoreWebView2*, ICoreWebView2DevToolsProtocolEventReceivedEventArgs* args) -> HRESULT {
+            LPWSTR raw = nullptr;
+            if (FAILED(args->get_ParameterObjectAsJson(&raw)) || !raw) return S_OK;
+            std::string p = wideToUtf8(raw); CoTaskMemFree(raw);
+            std::string session_id = extractJsonStringDecoded(p, "sessionId");
+            std::string info = extractJsonValueRaw(p, "targetInfo");
+            std::string type = extractJsonStringDecoded(info, "type");
+            std::string target_id = extractJsonStringDecoded(info, "targetId");
+            if (session_id.empty() || target_id.empty() || type != "iframe") return S_OK;
+            auto* v = getView(view_id);
+            if (!v) return S_OK;
+            std::lock_guard<std::mutex> lk(v->oopif_sessions_mutex);
+            v->oopif_sessions[target_id] = session_id;
+            return S_OK;
+          }).Get(),
+      &v->target_attached_token);
+  detached_r->add_DevToolsProtocolEventReceived(
+      Callback<ICoreWebView2DevToolsProtocolEventReceivedEventHandler>(
+          [view_id](ICoreWebView2*, ICoreWebView2DevToolsProtocolEventReceivedEventArgs* args) -> HRESULT {
+            LPWSTR raw = nullptr;
+            if (FAILED(args->get_ParameterObjectAsJson(&raw)) || !raw) return S_OK;
+            std::string p = wideToUtf8(raw); CoTaskMemFree(raw);
+            std::string session_id = extractJsonStringDecoded(p, "sessionId");
+            if (session_id.empty()) return S_OK;
+            auto* v = getView(view_id);
+            if (!v) return S_OK;
+            std::lock_guard<std::mutex> lk(v->oopif_sessions_mutex);
+            for (auto it = v->oopif_sessions.begin(); it != v->oopif_sessions.end(); ) {
+              if (it->second == session_id) it = v->oopif_sessions.erase(it);
+              else ++it;
+            }
+            return S_OK;
+          }).Get(),
+      &v->target_detached_token);
+  v->oopif_event_tokens_registered = true;
+}
+
+void finishResolveAndClickWv2(uint32_t view_id, uint32_t request_id, double x, double y,
+                               double w, double h, double cx, double cy,
+                               int32_t button, int32_t click_count, uint32_t modifiers) {
+  ViewHost* v = getView(view_id);
+  if (!v || !v->webview) { emitResolveAndClickErrorWv2(view_id, request_id, "runtime_error", "view destroyed"); return; }
+  dispatchCdpClickWv2(v, cx, cy, button, click_count, modifiers);
+  // Edge runtime CDP → DOM trust=true (empirical; matches existing click cap).
+  std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                        ",\"ok\":true,\"rect\":{\"x\":" + std::to_string(x) +
+                        ",\"y\":" + std::to_string(y) +
+                        ",\"width\":" + std::to_string(w) +
+                        ",\"height\":" + std::to_string(h) + "},"
+                        "\"isTrustedEvent\":true}";
+  emitWebviewEvent(view_id, "resolve-and-click-result", payload);
+}
+
+struct FrameResolveOkWv2 { double x, y, w, h, cx, cy, iw, ih; };
+
+// Parse a Runtime.evaluate response (regardless of session); on success forwards
+// frame-local fields to `onOk`. Script's failure branch routes through error emit.
+void parseEvalAndContinueWv2(uint32_t view_id, uint32_t request_id, bool ok, const std::string& evalResult,
+                              std::function<void(const FrameResolveOkWv2&)> onOk) {
+  if (!ok) {
+    BUNITE_INFO("webview2/eval: Runtime.evaluate failed view=%u request=%u body=%.300s%s",
+                view_id, request_id, evalResult.c_str(),
+                evalResult.size() > 300 ? "..." : "");
+    emitResolveAndClickErrorWv2(view_id, request_id, "runtime_error", "Runtime.evaluate failed"); return;
+  }
+  if (evalResult.find("\"exceptionDetails\"") != std::string::npos) {
+    emitResolveAndClickErrorWv2(view_id, request_id, "runtime_error", "evaluate threw"); return;
+  }
+  std::string value = extractJsonValueRaw(evalResult, "value");
+  if (value.empty()) { emitResolveAndClickErrorWv2(view_id, request_id, "runtime_error", "evaluate returned no value"); return; }
+  std::string okRaw = extractJsonValueRaw(value, "ok");
+  if (okRaw != "true") {
+    std::string code = extractJsonStringDecoded(value, "code");
+    if (code.empty()) code = "runtime_error";
+    emitResolveAndClickErrorWv2(view_id, request_id, code.c_str(), "");
+    return;
+  }
+  onOk(FrameResolveOkWv2{
+      extractJsonDouble(value, "x"), extractJsonDouble(value, "y"),
+      extractJsonDouble(value, "w"), extractJsonDouble(value, "h"),
+      extractJsonDouble(value, "cx"), extractJsonDouble(value, "cy"),
+      extractJsonDouble(value, "iw"), extractJsonDouble(value, "ih"),
+  });
+}
+
+void evalInFrameWv2(uint32_t view_id, uint32_t request_id, const std::string& frameId,
+                     const std::string& script,
+                     std::function<void(const FrameResolveOkWv2&)> onOk) {
+  ViewHost* v = getView(view_id);
+  if (!v || !v->webview) { emitResolveAndClickErrorWv2(view_id, request_id, "runtime_error", "view destroyed"); return; }
+  std::string session_id;
+  {
+    std::lock_guard<std::mutex> lk(v->oopif_sessions_mutex);
+    auto it = v->oopif_sessions.find(frameId);
+    if (it != v->oopif_sessions.end()) session_id = it->second;
+  }
+  if (!session_id.empty()) {
+    std::string evalParams = "{\"expression\":\"" + escapeJsonString(script) +
+                             "\",\"returnByValue\":true,\"awaitPromise\":true}";
+    cdpCallForSession(v, session_id, L"Runtime.evaluate", evalParams,
+        [view_id, request_id, onOk](bool ok, std::string r) {
+          parseEvalAndContinueWv2(view_id, request_id, ok, r, onOk);
+        });
+    return;
+  }
+  // In-process: createIsolatedWorld + Runtime.evaluate via main session.
+  std::string isoParams = "{\"frameId\":\"" + escapeJsonString(frameId) + "\",\"worldName\":\"bunite-rac\"}";
+  cdpCallWithResult(v, L"Page.createIsolatedWorld", isoParams,
+      [view_id, request_id, script, onOk](bool ok, std::string isoResult) {
+        if (!ok) { emitResolveAndClickErrorWv2(view_id, request_id, "runtime_error", "createIsolatedWorld failed"); return; }
+        int contextId = 0;
+        if (!extractJsonInt(isoResult, "executionContextId", contextId)) {
+          emitResolveAndClickErrorWv2(view_id, request_id, "runtime_error", "missing executionContextId"); return;
+        }
+        ViewHost* v2 = getView(view_id);
+        if (!v2 || !v2->webview) { emitResolveAndClickErrorWv2(view_id, request_id, "runtime_error", "view destroyed"); return; }
+        std::string evalParams = "{\"expression\":\"" + escapeJsonString(script) +
+                                 "\",\"contextId\":" + std::to_string(contextId) +
+                                 ",\"returnByValue\":true,\"awaitPromise\":true}";
+        cdpCallWithResult(v2, L"Runtime.evaluate", evalParams,
+            [view_id, request_id, onOk](bool ok2, std::string r) {
+              parseEvalAndContinueWv2(view_id, request_id, ok2, r, onOk);
+            });
+      });
+}
+
+inline void bilinearMapWv2(const std::array<double, 8>& q, double iw, double ih,
+                            double fx, double fy, double& px, double& py) {
+  const double u = (iw > 0) ? (fx / iw) : 0.0;
+  const double v = (ih > 0) ? (fy / ih) : 0.0;
+  px = (1-u)*(1-v)*q[0] + u*(1-v)*q[2] + u*v*q[4] + (1-u)*v*q[6];
+  py = (1-u)*(1-v)*q[1] + u*(1-v)*q[3] + u*v*q[5] + (1-u)*v*q[7];
+}
+
+// Recursive frame path lookup in Page.getFrameTree response (text-based parser).
+// Returns [main_frame_id, ..., target_frame_id]; empty if target not found.
+std::vector<std::string> findFramePathWv2(const std::string& node, const std::string& target) {
+  std::string frame = extractJsonValueRaw(node, "frame");
+  std::string this_id = extractJsonStringDecoded(frame, "id");
+  if (this_id.empty()) return {};
+  if (this_id == target) return {this_id};
+  std::string children = extractJsonValueRaw(node, "childFrames");
+  if (children.size() < 2 || children.front() != '[') return {};
+  // Walk child array — each element is a JSON object {frame, childFrames?}.
+  size_t pos = 1;
+  while (pos < children.size() && children[pos] != ']') {
+    while (pos < children.size() && (children[pos] == ' ' || children[pos] == ',')) ++pos;
+    if (pos >= children.size() || children[pos] != '{') break;
+    int depth = 0;
+    size_t end = pos;
+    bool inStr = false;
+    while (end < children.size()) {
+      char ch = children[end];
+      if (inStr) { if (ch == '\\' && end + 1 < children.size()) ++end; else if (ch == '"') inStr = false; }
+      else if (ch == '"') inStr = true;
+      else if (ch == '{') ++depth;
+      else if (ch == '}') { --depth; if (depth == 0) { ++end; break; } }
+      ++end;
+    }
+    auto sub = findFramePathWv2(children.substr(pos, end - pos), target);
+    if (!sub.empty()) { sub.insert(sub.begin(), this_id); return sub; }
+    pos = end;
+  }
+  return {};
+}
+
+bool parseQuad8(const std::string& content, std::array<double, 8>& out) {
+  if (content.size() < 2 || content.front() != '[' || content.back() != ']') return false;
+  size_t pos = 1;
+  for (int i = 0; i < 8; ++i) {
+    while (pos < content.size() && (content[pos] == ' ' || content[pos] == ',')) ++pos;
+    size_t end = pos;
+    while (end < content.size() && content[end] != ',' && content[end] != ']') ++end;
+    if (end == pos) return false;
+    try { out[i] = std::stod(content.substr(pos, end - pos)); } catch (...) { return false; }
+    pos = end;
+  }
+  return true;
+}
+
+// Issue CDP on a specific OOPIF session, or main session if `session_id` empty.
+void cdpForChain(ViewHost* v, const std::string& session_id, const wchar_t* method,
+                  const std::string& params_json,
+                  std::function<void(bool, std::string)> cb) {
+  if (session_id.empty()) cdpCallWithResult(v, method, params_json, std::move(cb));
+  else cdpCallForSession(v, session_id, method, params_json, std::move(cb));
+}
+
+struct ChainStateWv2 {
+  uint32_t view_id;
+  uint32_t request_id;
+  std::string targetFrameId;
+  std::string script;
+  int32_t button, click_count;
+  uint32_t modifiers;
+  std::vector<std::string> chain;             // [main, ..., target]
+  std::vector<std::array<double, 8>> link_quads;
+  std::vector<std::pair<double, double>> ancestor_inner;  // chain[1..N-2]'s iw/ih
+};
+
+void composeAndDispatchWv2(std::shared_ptr<ChainStateWv2> s, const FrameResolveOkWv2& fr);
+void fetchTargetEvalWv2(std::shared_ptr<ChainStateWv2> s);
+void fetchAncestorInnerWv2(std::shared_ptr<ChainStateWv2> s, size_t i);
+void fetchLinkWv2(std::shared_ptr<ChainStateWv2> s, size_t link_idx);
+
+std::string sessionForChainIdxWv2(uint32_t view_id, const std::vector<std::string>& chain, size_t idx) {
+  if (idx == 0) return {};
+  ViewHost* v = getView(view_id);
+  if (!v) return {};
+  std::lock_guard<std::mutex> lk(v->oopif_sessions_mutex);
+  auto it = v->oopif_sessions.find(chain[idx]);
+  return (it != v->oopif_sessions.end()) ? it->second : std::string{};
+}
+
+void fetchLinkWv2(std::shared_ptr<ChainStateWv2> s, size_t link_idx) {
+  if (link_idx + 1 >= s->chain.size()) { fetchAncestorInnerWv2(s, 1); return; }
+  const std::string parent_session = sessionForChainIdxWv2(s->view_id, s->chain, link_idx);
+  const std::string& child_frameId = s->chain[link_idx + 1];
+  ViewHost* v = getView(s->view_id);
+  if (!v) { emitResolveAndClickErrorWv2(s->view_id, s->request_id, "runtime_error", "view destroyed"); return; }
+  std::string ownerParams = "{\"frameId\":\"" + escapeJsonString(child_frameId) + "\"}";
+  cdpForChain(v, parent_session, L"DOM.getFrameOwner", ownerParams,
+      [s, link_idx, parent_session](bool ok, std::string r) {
+        if (!ok) { emitResolveAndClickErrorWv2(s->view_id, s->request_id, "not_found", "getFrameOwner failed"); return; }
+        int backendNodeId = 0;
+        if (!extractJsonInt(r, "backendNodeId", backendNodeId) || !backendNodeId) {
+          emitResolveAndClickErrorWv2(s->view_id, s->request_id, "not_found", "no backendNodeId"); return;
+        }
+        ViewHost* v2 = getView(s->view_id);
+        if (!v2) { emitResolveAndClickErrorWv2(s->view_id, s->request_id, "runtime_error", "view destroyed"); return; }
+        std::string boxParams = "{\"backendNodeId\":" + std::to_string(backendNodeId) + "}";
+        cdpForChain(v2, parent_session, L"DOM.getBoxModel", boxParams,
+            [s, link_idx](bool ok2, std::string rb) {
+              if (!ok2) { emitResolveAndClickErrorWv2(s->view_id, s->request_id, "not_visible", "iframe has no box"); return; }
+              std::string model = extractJsonValueRaw(rb, "model");
+              std::string content = extractJsonValueRaw(model, "content");
+              std::array<double, 8> quad{};
+              if (!parseQuad8(content, quad)) { emitResolveAndClickErrorWv2(s->view_id, s->request_id, "runtime_error", "bad quad"); return; }
+              s->link_quads.push_back(quad);
+              fetchLinkWv2(s, link_idx + 1);
+            });
+      });
+}
+
+void fetchAncestorInnerWv2(std::shared_ptr<ChainStateWv2> s, size_t i) {
+  if (i + 1 >= s->chain.size()) { fetchTargetEvalWv2(s); return; }
+  const std::string sid = sessionForChainIdxWv2(s->view_id, s->chain, i);
+  ViewHost* v = getView(s->view_id);
+  if (!v) { emitResolveAndClickErrorWv2(s->view_id, s->request_id, "runtime_error", "view destroyed"); return; }
+  std::string params = "{\"expression\":\"JSON.stringify({iw:innerWidth,ih:innerHeight})\",\"returnByValue\":true,\"awaitPromise\":true}";
+  cdpForChain(v, sid, L"Runtime.evaluate", params,
+      [s, i](bool ok, std::string r) {
+        if (!ok) { emitResolveAndClickErrorWv2(s->view_id, s->request_id, "runtime_error", "ancestor eval failed"); return; }
+        // Result: {"result":{"type":"string","value":"<json>"}}
+        std::string value = extractJsonValueRaw(r, "value");
+        if (value.size() < 2) { emitResolveAndClickErrorWv2(s->view_id, s->request_id, "runtime_error", "ancestor eval no value"); return; }
+        // value is a JSON string literal — extractJsonValueRaw returns it with quotes; decode.
+        std::string inner;
+        for (size_t p = 1; p + 1 < value.size(); ++p) {
+          if (value[p] == '\\' && p + 2 < value.size()) {
+            char nxt = value[++p];
+            switch (nxt) { case '"': inner += '"'; break; case '\\': inner += '\\'; break;
+                            case 'n': inner += '\n'; break; default: inner += nxt; }
+          } else inner += value[p];
+        }
+        double iw = extractJsonDouble(inner, "iw"), ih = extractJsonDouble(inner, "ih");
+        s->ancestor_inner.push_back({iw, ih});
+        fetchAncestorInnerWv2(s, i + 1);
+      });
+}
+
+void fetchTargetEvalWv2(std::shared_ptr<ChainStateWv2> s) {
+  evalInFrameWv2(s->view_id, s->request_id, s->targetFrameId, s->script,
+      [s](const FrameResolveOkWv2& fr) { composeAndDispatchWv2(s, fr); });
+}
+
+void composeAndDispatchWv2(std::shared_ptr<ChainStateWv2> s, const FrameResolveOkWv2& fr) {
+  auto mapCorner = [&](double fx, double fy, double& px, double& py) {
+    double cur_x = fx, cur_y = fy;
+    double cur_iw = fr.iw, cur_ih = fr.ih;
+    for (size_t i = s->link_quads.size(); i-- > 0; ) {
+      double mx, my;
+      bilinearMapWv2(s->link_quads[i], cur_iw, cur_ih, cur_x, cur_y, mx, my);
+      cur_x = mx; cur_y = my;
+      if (i == 0) break;
+      cur_iw = s->ancestor_inner[i - 1].first;
+      cur_ih = s->ancestor_inner[i - 1].second;
+    }
+    px = cur_x; py = cur_y;
+  };
+  double pcx, pcy; mapCorner(fr.cx, fr.cy, pcx, pcy);
+  double cx0, cy0, cx1, cy1, cx2, cy2, cx3, cy3;
+  mapCorner(fr.x,             fr.y,             cx0, cy0);
+  mapCorner(fr.x + fr.w,      fr.y,             cx1, cy1);
+  mapCorner(fr.x + fr.w,      fr.y + fr.h,      cx2, cy2);
+  mapCorner(fr.x,             fr.y + fr.h,      cx3, cy3);
+  const double min_x = std::min(std::min(cx0, cx1), std::min(cx2, cx3));
+  const double max_x = std::max(std::max(cx0, cx1), std::max(cx2, cx3));
+  const double min_y = std::min(std::min(cy0, cy1), std::min(cy2, cy3));
+  const double max_y = std::max(std::max(cy0, cy1), std::max(cy2, cy3));
+  finishResolveAndClickWv2(s->view_id, s->request_id,
+      min_x, min_y, max_x - min_x, max_y - min_y, pcx, pcy,
+      s->button, s->click_count, s->modifiers);
+}
+
+// Walk ancestor chain via Page.getFrameTree, compose bilinear transforms across
+// nested OOPIF/same-origin frames, dispatch click in main-page coords.
+void runFrameTargetedWv2(uint32_t view_id, uint32_t request_id, const std::string& frameId,
+                          const std::string& script,
+                          int32_t button, int32_t click_count, uint32_t modifiers) {
+  ViewHost* v = getView(view_id);
+  if (!v || !v->webview) { emitResolveAndClickErrorWv2(view_id, request_id, "runtime_error", "view destroyed"); return; }
+  cdpCallWithResult(v, L"Page.getFrameTree", "{}",
+      [view_id, request_id, frameId, script, button, click_count, modifiers](bool ok, std::string r) {
+        if (!ok) { emitResolveAndClickErrorWv2(view_id, request_id, "runtime_error", "getFrameTree failed"); return; }
+        std::string root = extractJsonValueRaw(r, "frameTree");
+        std::vector<std::string> chain = findFramePathWv2(root, frameId);
+        if (chain.size() < 2) { emitResolveAndClickErrorWv2(view_id, request_id, "not_found", "frame not in tree"); return; }
+        auto s = std::make_shared<ChainStateWv2>();
+        s->view_id = view_id; s->request_id = request_id;
+        s->targetFrameId = frameId; s->script = script;
+        s->button = button; s->click_count = click_count; s->modifiers = modifiers;
+        s->chain = std::move(chain);
+        fetchLinkWv2(s, 0);
+      });
+}
+}  // namespace
+
+extern "C" {
+
+BUNITE_EXPORT void bunite_view_resolve_and_click(
+    uint32_t view_id, uint32_t request_id,
+    const char* selector_c, const char* frame_id_c,
+    int32_t button, int32_t click_count, uint32_t modifiers) {
+  ViewHost* v = getView(view_id);
+  if (!v || !v->webview) { emitResolveAndClickErrorWv2(view_id, request_id, "runtime_error", "view not ready"); return; }
+  std::string selector = selector_c ? selector_c : "";
+  std::string frameId = frame_id_c ? frame_id_c : "";
+  std::string script = buildResolveScriptWv2(selector);
+
+  if (frameId.empty()) {
+    std::string evalParams = "{\"expression\":\"" + escapeJsonString(script) + "\",\"returnByValue\":true,\"awaitPromise\":true}";
+    cdpCallWithResult(v, L"Runtime.evaluate", evalParams,
+        [view_id, request_id, button, click_count, modifiers](bool ok, std::string r) {
+          parseEvalAndContinueWv2(view_id, request_id, ok, r,
+              [view_id, request_id, button, click_count, modifiers](const FrameResolveOkWv2& fr) {
+                finishResolveAndClickWv2(view_id, request_id,
+                                          fr.x, fr.y, fr.w, fr.h, fr.cx, fr.cy,
+                                          button, click_count, modifiers);
+              });
+        });
+    return;
+  }
+
+  // Frame-targeted: lazy event subscription + setAutoAttach so OOPIF child
+  // sessions populate v->oopif_sessions before frame eval routes.
+  armOopifEvents(v);
+  if (!v->oopif_autoattach_armed.exchange(true)) {
+    cdpCallWithResult(v, L"Target.setAutoAttach",
+        "{\"autoAttach\":true,\"flatten\":true,\"waitForDebuggerOnStart\":false}",
+        [view_id, request_id, frameId, script, button, click_count, modifiers](bool ok, std::string) {
+          if (!ok) { emitResolveAndClickErrorWv2(view_id, request_id, "runtime_error", "setAutoAttach failed"); return; }
+          runFrameTargetedWv2(view_id, request_id, frameId, script, button, click_count, modifiers);
+        });
+    return;
+  }
+  runFrameTargetedWv2(view_id, request_id, frameId, script, button, click_count, modifiers);
+}
+
+BUNITE_EXPORT void bunite_view_accessibility_snapshot(uint32_t view_id, uint32_t request_id,
+                                                       int32_t /*interesting_only*/) {
+  // CDP getFullAXTree accepts only depth/frameId — interesting-only filter is TS-side.
+  ViewHost* v = getView(view_id);
+  if (!v || !v->webview) { emitAxError(view_id, request_id, "not_supported", "view not ready"); return; }
+  cdpCallWithResult(v, L"Accessibility.getFullAXTree", "{}",
+      [view_id, request_id](bool ok, std::string result) {
+        if (!ok) { emitAxError(view_id, request_id, "runtime_error", "getFullAXTree failed: " + result); return; }
+        std::string payload = "{\"requestId\":" + std::to_string(request_id) +
+                              ",\"ok\":true,\"tree\":" + result + "}";
+        emitWebviewEvent(view_id, "accessibility-result", payload);
+      });
 }
 
 BUNITE_EXPORT void bunite_view_screenshot(uint32_t view_id, uint32_t request_id,