fix(desktop-browser): return popup focus to the user's page and wait for a shown view

Author: waleedlatif1

apps/desktop/e2e/browser-focus.spec.ts

@@ -251,6 +251,18 @@ test('browser focus and shortcuts stay with the surface the user is using', asyn
       await expect.poll(focusedPageUrls).toEqual([`${site}/four`])
     })
 
+    await check('a popup the agent opens hands focus back to the page the user is in', async () => {
+      const before = await tabCount()
+      const snapshot = await execute('browser_snapshot', {})
+      const ref = /"Open popup" \[ref=(\d+)\]/.exec(
+        String((snapshot as { outline?: string }).outline)
+      )?.[1]
+      expect(ref, 'snapshot lists the popup link').toBeTruthy()
+      await execute('browser_click', { elementId: Number(ref) })
+      await expect.poll(tabCount).toBe(before + 1)
+      await expect.poll(focusedPageUrls).toEqual([`${site}/four`])
+    })
+
     await check('reload keys typed in the page reload only that page', async () => {
       for (const [keyCode, modifiers] of [
         ['R', [PRIMARY]],

apps/desktop/src/main/browser-agent/session.ts

@@ -1857,7 +1857,9 @@ export function focusPageForUser(contents: WebContents): void {
 function applyPendingUserFocus(view: WebContentsView): void {
   const tab = activeTab()
   if (!tab?.pendingUserFocus || tab.view !== view || view.webContents.isDestroyed()) return
-  if (!isPanelVisible() || getBrowserScopeId() !== getActiveBrowserScopeId()) return
+  // A renderer modal can hide the view while the panel keeps its bounds.
+  if (!view.getVisible() || !isPanelVisible()) return
+  if (getBrowserScopeId() !== getActiveBrowserScopeId()) return
   tab.pendingUserFocus = false
   view.webContents.focus()
 }
@@ -1880,8 +1882,9 @@ function publishPageIssue(tab: AgentTab, focusRecovery = false): void {
   // Recovery moves focus to the renderer's issue page only when the failed page
   // held it; the user typing in chat keeps their caret.
   const pageHadFocus =
-    currentScope.focusedBrowserTabId === tab.id ||
-    (!tab.view.webContents.isDestroyed() && tab.view.webContents.isFocused())
+    !currentScope.browserChromeFocused &&
+    (currentScope.focusedBrowserTabId === tab.id ||
+      (!tab.view.webContents.isDestroyed() && tab.view.webContents.isFocused()))
   if (
     focusRecovery &&
     pageHadFocus &&
@@ -2307,15 +2310,31 @@ function initializeTabView(
       const tab = tabs.find((entry) => entry.view.webContents === contents)
       // A tab the user cannot see never keeps keyboard focus. Chromium focuses
       // a page opened without an opener (a target=_blank link) while creating
-      // it, before the tab is even listed. Hand focus back once that focus
-      // call has returned, or Chromium finishes it over the top.
+      // it, before the tab is even listed. Hand focus back to the visible page
+      // the user was in, or else to Sim, once that focus call has returned, or
+      // Chromium finishes it over the top.
       if (!tab || tab.id !== currentScope.activeTabId) {
+        const active = activeTab()
+        const returnTo =
+          active &&
+          !currentScope.browserChromeFocused &&
+          currentScope.focusedBrowserTabId === active.id
+            ? active
+            : null
         setImmediate(
           bindToBrowserScope(scopeId, () => {
             const current = tabs.find((entry) => entry.view.webContents === contents)
             const win = panelWindow()
             if (current?.id === currentScope.activeTabId || !win || win.isDestroyed()) return
-            if (!contents.isDestroyed() && contents.isFocused()) win.webContents.focus()
+            if (contents.isDestroyed() || !contents.isFocused()) return
+            if (
+              returnTo?.id === currentScope.activeTabId &&
+              !returnTo.view.webContents.isDestroyed()
+            ) {
+              returnTo.view.webContents.focus()
+            } else {
+              win.webContents.focus()
+            }
           })
         )
         return

apps/desktop/src/test/electron-mock.ts

@@ -250,7 +250,11 @@ function createWebContentsMock() {
 export class WebContentsView {
   webContents = createWebContentsMock()
   setBackgroundColor = vi.fn()
-  setVisible = vi.fn()
+  private visible = true
+  setVisible = vi.fn((visible: boolean) => {
+    this.visible = visible
+  })
+  getVisible = vi.fn(() => this.visible)
   private bounds = { x: 0, y: 0, width: 0, height: 0 }
   setBounds = vi.fn((bounds: { x: number; y: number; width: number; height: number }) => {
     this.bounds = { ...bounds }

apps/sim/app/workspace/[workspaceId]/home/components/mothership-view/components/resource-content/components/browser-session/browser-page-issue.tsx

@@ -127,7 +127,8 @@ export function browserPageIssueCopy(issue: BrowserPageIssue): BrowserPageIssueC
 
 function isEditingOutside(heading: HTMLElement | null): boolean {
   const active = document.activeElement
-  if (!(active instanceof HTMLElement) || !document.hasFocus()) return false
+  // activeElement survives a window blur, so a caret left in chat still counts.
+  if (!(active instanceof HTMLElement)) return false
   const section = heading?.closest('section')
   if (section?.contains(active)) return false
   return (