|
| 1 | +import { mkdirSync, mkdtempSync, rmSync, writeFileSync } from 'node:fs' |
| 2 | +import { createServer, type Server } from 'node:http' |
| 3 | +import { tmpdir } from 'node:os' |
| 4 | +import { dirname, join } from 'node:path' |
| 5 | +import { fileURLToPath } from 'node:url' |
| 6 | +import { |
| 7 | + type ElectronApplication, |
| 8 | + _electron as electron, |
| 9 | + expect, |
| 10 | + type Page, |
| 11 | + test, |
| 12 | +} from '@playwright/test' |
| 13 | +import type { BrowserToolName } from '@sim/browser-protocol' |
| 14 | +import type { SimDesktopApi } from '@sim/desktop-bridge' |
| 15 | +import { getErrorMessage } from '@sim/utils/errors' |
| 16 | + |
| 17 | +const DESKTOP_DIR = fileURLToPath(new URL('..', import.meta.url)) |
| 18 | +const SCOPE = 'browser-focus-e2e' |
| 19 | +const PRIMARY = process.platform === 'darwin' ? 'meta' : 'control' |
| 20 | +const SHELL_FIXTURE = |
| 21 | + '<!doctype html><title>Sim fixture</title><h1>Browser focus fixture</h1><textarea id="composer" aria-label="Composer"></textarea>' |
| 22 | +const page = (name: string) => |
| 23 | + `<!doctype html><title>${name}</title><a href="/popup" target="_blank">Open popup</a><input aria-label="Field">` |
| 24 | + |
| 25 | +type Bridge = typeof globalThis & { simDesktop: SimDesktopApi; shellMarker?: string } |
| 26 | + |
| 27 | +/** |
| 28 | + * The browser shares one window with Sim, so focus and shortcuts decide whether |
| 29 | + * a keystroke lands in chat, in the page, or reloads all of Sim. These checks |
| 30 | + * drive the real shell and native tab views. |
| 31 | + */ |
| 32 | +test('browser focus and shortcuts stay with the surface the user is using', async () => { |
| 33 | + const reportPath = |
| 34 | + process.env.DESKTOP_BROWSER_FOCUS_REPORT_PATH ?? test.info().outputPath('browser-focus.json') |
| 35 | + const checks: { |
| 36 | + name: string |
| 37 | + status: 'passed' | 'failed' |
| 38 | + durationMs: number |
| 39 | + error?: string |
| 40 | + }[] = [] |
| 41 | + const check = async (name: string, run: () => Promise<void>) => { |
| 42 | + const started = Date.now() |
| 43 | + try { |
| 44 | + await test.step(name, run) |
| 45 | + checks.push({ name, status: 'passed', durationMs: Date.now() - started }) |
| 46 | + } catch (error) { |
| 47 | + checks.push({ |
| 48 | + name, |
| 49 | + status: 'failed', |
| 50 | + durationMs: Date.now() - started, |
| 51 | + error: getErrorMessage(error), |
| 52 | + }) |
| 53 | + throw error |
| 54 | + } |
| 55 | + } |
| 56 | + const calls = new Map<string, { chatId: string; toolName: BrowserToolName; args: unknown }>() |
| 57 | + const server: Server = createServer(async (request, response) => { |
| 58 | + const path = new URL(request.url ?? '/', 'http://localhost').pathname |
| 59 | + if (path === '/api/desktop/tool/authorize') { |
| 60 | + let body = '' |
| 61 | + for await (const chunk of request) body += chunk.toString() |
| 62 | + const authorization = calls.get(JSON.parse(body).toolCallId) |
| 63 | + response.writeHead(authorization ? 200 : 403, { 'Content-Type': 'application/json' }) |
| 64 | + response.end(JSON.stringify(authorization ?? {})) |
| 65 | + return |
| 66 | + } |
| 67 | + // The app origin serves the shell; the same server on localhost is the web. |
| 68 | + const isSite = request.headers.host?.startsWith('localhost') === true |
| 69 | + response.writeHead(200, { |
| 70 | + 'Content-Type': 'text/html', |
| 71 | + ...(isSite ? {} : { 'Set-Cookie': 'better-auth.session_token=fixture; HttpOnly; Path=/' }), |
| 72 | + }) |
| 73 | + response.end(isSite ? page(path.slice(1)) : SHELL_FIXTURE) |
| 74 | + }) |
| 75 | + const userData = mkdtempSync(join(tmpdir(), 'sim-browser-focus-e2e-')) |
| 76 | + let app: ElectronApplication | undefined |
| 77 | + let passed = false |
| 78 | + try { |
| 79 | + await new Promise<void>((resolve) => server.listen(0, '127.0.0.1', resolve)) |
| 80 | + const address = server.address() |
| 81 | + if (!address || typeof address === 'string') throw new Error('Missing fixture address') |
| 82 | + const origin = `http://127.0.0.1:${address.port}` |
| 83 | + /** Pages outside the app origin browse in the agent partition, like any third-party site. */ |
| 84 | + const site = origin.replace('127.0.0.1', 'localhost') |
| 85 | + const shellApp = await electron.launch({ |
| 86 | + args: [process.env.SIM_DESKTOP_E2E_MAIN ?? '.'], |
| 87 | + cwd: DESKTOP_DIR, |
| 88 | + env: { ...process.env, SIM_DESKTOP_ORIGIN: origin, SIM_DESKTOP_USER_DATA: userData }, |
| 89 | + }) |
| 90 | + app = shellApp |
| 91 | + const shell: Page = await shellApp.firstWindow() |
| 92 | + await shellApp.evaluate(({ app, BrowserWindow }) => { |
| 93 | + const host = BrowserWindow.getAllWindows()[0] |
| 94 | + host.webContents.setBackgroundThrottling(false) |
| 95 | + app.focus({ steal: true }) |
| 96 | + host.focus() |
| 97 | + }) |
| 98 | + await expect(shell.getByRole('heading')).toHaveText('Browser focus fixture') |
| 99 | + await shell.evaluate(async (scope) => { |
| 100 | + const api = (globalThis as Bridge).simDesktop.browserAgent |
| 101 | + await api.activateScope(scope) |
| 102 | + const updateBounds = () => |
| 103 | + api.setPanelBounds( |
| 104 | + { x: 0, y: 120, width: innerWidth, height: innerHeight - 120 }, |
| 105 | + null, |
| 106 | + scope |
| 107 | + ) |
| 108 | + updateBounds() |
| 109 | + ;(globalThis as Bridge & { boundsTimer?: number }).boundsTimer = window.setInterval( |
| 110 | + updateBounds, |
| 111 | + 200 |
| 112 | + ) |
| 113 | + ;(globalThis as Bridge).shellMarker = 'alive' |
| 114 | + }, SCOPE) |
| 115 | + |
| 116 | + let callCount = 0 |
| 117 | + const execute = async (tool: BrowserToolName, args: Record<string, unknown>) => { |
| 118 | + const callId = `browser-focus-${++callCount}` |
| 119 | + calls.set(callId, { chatId: SCOPE, toolName: tool, args }) |
| 120 | + const result = await shell.evaluate( |
| 121 | + ({ callId, tool, args, scope }) => |
| 122 | + (globalThis as Bridge).simDesktop.browserAgent.executeTool(callId, tool, args, scope), |
| 123 | + { callId, tool, args, scope: SCOPE } |
| 124 | + ) |
| 125 | + expect(result.ok).toBe(true) |
| 126 | + return result.ok ? result.result : undefined |
| 127 | + } |
| 128 | + const panelAction = (action: Record<string, unknown>) => |
| 129 | + shell.evaluate( |
| 130 | + ({ action, scope }) => |
| 131 | + (globalThis as Bridge).simDesktop.browserAgent.panelAction( |
| 132 | + action as unknown as Parameters<SimDesktopApi['browserAgent']['panelAction']>[0], |
| 133 | + scope |
| 134 | + ), |
| 135 | + { action, scope: SCOPE } |
| 136 | + ) |
| 137 | + const setPanelFocused = (focused: boolean) => |
| 138 | + shell.evaluate( |
| 139 | + ({ focused, scope }) => |
| 140 | + (globalThis as Bridge).simDesktop.browserAgent.setPanelFocused(focused, scope), |
| 141 | + { focused, scope: SCOPE } |
| 142 | + ) |
| 143 | + const tabCount = () => |
| 144 | + shell.evaluate( |
| 145 | + async (scope) => |
| 146 | + (await (globalThis as Bridge).simDesktop.browserAgent.activateScope(scope)).tabs.length, |
| 147 | + SCOPE |
| 148 | + ) |
| 149 | + const shellAlive = () => |
| 150 | + shell.evaluate(() => (globalThis as Bridge).shellMarker === 'alive').catch(() => false) |
| 151 | + const composerFocused = () => |
| 152 | + shell.evaluate(() => document.hasFocus() && document.activeElement?.id === 'composer') |
| 153 | + const focusedPageUrls = () => |
| 154 | + shellApp.evaluate(({ BrowserWindow, WebContentsView }) => |
| 155 | + BrowserWindow.getAllWindows()[0] |
| 156 | + .contentView.children.filter( |
| 157 | + (view) => view instanceof WebContentsView && view.webContents.isFocused() |
| 158 | + ) |
| 159 | + .map((view) => (view as Electron.WebContentsView).webContents.getURL()) |
| 160 | + ) |
| 161 | + const visiblePageUrl = () => |
| 162 | + shellApp.evaluate(({ BrowserWindow, WebContentsView }) => { |
| 163 | + const view = BrowserWindow.getAllWindows()[0].contentView.children.find( |
| 164 | + (child) => child instanceof WebContentsView && child.getVisible() |
| 165 | + ) as Electron.WebContentsView | undefined |
| 166 | + return view?.webContents.getURL() ?? null |
| 167 | + }) |
| 168 | + /** Counts main-frame loads of the visible page from now on. */ |
| 169 | + const countVisiblePageLoads = () => |
| 170 | + shellApp.evaluate(({ BrowserWindow, WebContentsView }) => { |
| 171 | + const view = BrowserWindow.getAllWindows()[0].contentView.children.find( |
| 172 | + (child) => child instanceof WebContentsView && child.getVisible() |
| 173 | + ) as Electron.WebContentsView |
| 174 | + const counter = globalThis as typeof globalThis & { pageLoads?: number; counted?: boolean } |
| 175 | + counter.pageLoads = 0 |
| 176 | + if (!counter.counted) { |
| 177 | + counter.counted = true |
| 178 | + view.webContents.on('did-finish-load', () => { |
| 179 | + counter.pageLoads = (counter.pageLoads ?? 0) + 1 |
| 180 | + }) |
| 181 | + } |
| 182 | + }) |
| 183 | + const pageLoads = () => |
| 184 | + shellApp.evaluate(() => (globalThis as typeof globalThis & { pageLoads?: number }).pageLoads) |
| 185 | + const focusVisiblePage = () => |
| 186 | + shellApp.evaluate(({ BrowserWindow, WebContentsView }) => { |
| 187 | + const view = BrowserWindow.getAllWindows()[0].contentView.children.find( |
| 188 | + (child) => child instanceof WebContentsView && child.getVisible() |
| 189 | + ) as Electron.WebContentsView |
| 190 | + view.webContents.focus() |
| 191 | + }) |
| 192 | + const pressInPage = (keyCode: string, modifiers: string[]) => |
| 193 | + shellApp.evaluate( |
| 194 | + ({ BrowserWindow, WebContentsView }, { keyCode, modifiers }) => { |
| 195 | + const view = BrowserWindow.getAllWindows()[0].contentView.children.find( |
| 196 | + (child) => child instanceof WebContentsView && child.getVisible() |
| 197 | + ) as Electron.WebContentsView |
| 198 | + const input = { keyCode, modifiers } as Electron.KeyboardInputEvent |
| 199 | + view.webContents.sendInputEvent({ ...input, type: 'keyDown' }) |
| 200 | + view.webContents.sendInputEvent({ ...input, type: 'keyUp' }) |
| 201 | + }, |
| 202 | + { keyCode, modifiers } |
| 203 | + ) |
| 204 | + /** Clicks an application-menu item the way its accelerator would. */ |
| 205 | + const clickMenu = (label: string) => |
| 206 | + shellApp.evaluate(({ BrowserWindow, Menu }, label) => { |
| 207 | + const win = BrowserWindow.getAllWindows()[0] |
| 208 | + const find = (items: Electron.MenuItem[]): Electron.MenuItem | null => { |
| 209 | + for (const item of items) { |
| 210 | + if (item.label === label) return item |
| 211 | + const found = item.submenu ? find(item.submenu.items) : null |
| 212 | + if (found) return found |
| 213 | + } |
| 214 | + return null |
| 215 | + } |
| 216 | + const item = find(Menu.getApplicationMenu()?.items ?? []) |
| 217 | + if (!item) throw new Error(`No menu item ${label}`) |
| 218 | + item.click(undefined, win, win.webContents) |
| 219 | + }, label) |
| 220 | + |
| 221 | + await check('the agent working never takes the caret from the composer', async () => { |
| 222 | + await shell.locator('#composer').click() |
| 223 | + await shell.keyboard.type('draft') |
| 224 | + for (const [tool, args] of [ |
| 225 | + ['browser_open_url', { url: `${site}/one` }], |
| 226 | + ['browser_navigate', { url: `${site}/two` }], |
| 227 | + ['browser_open_tab', { url: `${site}/three` }], |
| 228 | + ] as const) { |
| 229 | + await execute(tool, args) |
| 230 | + await expect.poll(visiblePageUrl).not.toBeNull() |
| 231 | + expect(await composerFocused()).toBe(true) |
| 232 | + expect(await focusedPageUrls()).toEqual([]) |
| 233 | + } |
| 234 | + // A link without an opener is focused by Chromium itself as it is created. |
| 235 | + const snapshot = await execute('browser_snapshot', {}) |
| 236 | + const ref = /"Open popup" \[ref=(\d+)\]/.exec( |
| 237 | + String((snapshot as { outline?: string }).outline) |
| 238 | + )?.[1] |
| 239 | + expect(ref, 'snapshot lists the popup link').toBeTruthy() |
| 240 | + await execute('browser_click', { elementId: Number(ref) }) |
| 241 | + await expect.poll(tabCount).toBe(3) |
| 242 | + await expect.poll(focusedPageUrls).toEqual([]) |
| 243 | + expect(await composerFocused()).toBe(true) |
| 244 | + await shell.keyboard.type(' continues') |
| 245 | + await expect(shell.locator('#composer')).toHaveValue('draft continues') |
| 246 | + }) |
| 247 | + |
| 248 | + await check('an omnibox navigation hands focus to the page it loads', async () => { |
| 249 | + await panelAction({ action: 'switch-tab', tabId: '1' }) |
| 250 | + await panelAction({ action: 'navigate', url: `${site}/four` }) |
| 251 | + await expect.poll(focusedPageUrls).toEqual([`${site}/four`]) |
| 252 | + }) |
| 253 | + |
| 254 | + await check('reload keys typed in the page reload only that page', async () => { |
| 255 | + for (const [keyCode, modifiers] of [ |
| 256 | + ['R', [PRIMARY]], |
| 257 | + ['R', [PRIMARY, 'shift']], |
| 258 | + ['F5', []], |
| 259 | + ] as const) { |
| 260 | + await countVisiblePageLoads() |
| 261 | + await focusVisiblePage() |
| 262 | + await pressInPage(keyCode, [...modifiers]) |
| 263 | + await expect.poll(pageLoads).toBe(1) |
| 264 | + } |
| 265 | + }) |
| 266 | + |
| 267 | + await check('the page keeps its shortcut claim after the chrome reports blur', async () => { |
| 268 | + await setPanelFocused(true) |
| 269 | + await focusVisiblePage() |
| 270 | + await setPanelFocused(false) |
| 271 | + await countVisiblePageLoads() |
| 272 | + await clickMenu('Reload') |
| 273 | + await expect.poll(pageLoads).toBe(1) |
| 274 | + expect(await shellAlive()).toBe(true) |
| 275 | + }) |
| 276 | + |
| 277 | + await check('Back and Forward move through the focused page history', async () => { |
| 278 | + await panelAction({ action: 'navigate', url: `${site}/five` }) |
| 279 | + await expect.poll(visiblePageUrl).toBe(`${site}/five`) |
| 280 | + await focusVisiblePage() |
| 281 | + await clickMenu('Back') |
| 282 | + await expect.poll(visiblePageUrl).toBe(`${site}/four`) |
| 283 | + await focusVisiblePage() |
| 284 | + await clickMenu('Forward') |
| 285 | + await expect.poll(visiblePageUrl).toBe(`${site}/five`) |
| 286 | + }) |
| 287 | + |
| 288 | + await check('browser shortcuts work while renderer chrome hides the page', async () => { |
| 289 | + const before = await tabCount() |
| 290 | + await shell.evaluate((scope) => { |
| 291 | + const bridge = globalThis as Bridge & { boundsTimer?: number } |
| 292 | + window.clearInterval(bridge.boundsTimer) |
| 293 | + bridge.simDesktop.browserAgent.setPanelBounds(null, null, scope) |
| 294 | + }, SCOPE) |
| 295 | + await setPanelFocused(true) |
| 296 | + await clickMenu('New Tab') |
| 297 | + await expect.poll(tabCount).toBe(before + 1) |
| 298 | + }) |
| 299 | + passed = true |
| 300 | + } finally { |
| 301 | + mkdirSync(dirname(reportPath), { recursive: true }) |
| 302 | + writeFileSync(reportPath, JSON.stringify({ passed, checks }, null, 2)) |
| 303 | + await app?.close() |
| 304 | + await new Promise<void>((resolve) => server.close(() => resolve())) |
| 305 | + rmSync(userData, { recursive: true, force: true }) |
| 306 | + } |
| 307 | +}) |
0 commit comments