dsh-builtin-browser 0.4.0 → 0.4.2

package/src/browser-electron/history-store.ts

@@ -13,10 +13,11 @@
  * @module dsh-browser/browser-electron/history-store
  */
 
-import { appendFileSync, existsSync, mkdirSync, readFileSync, writeFileSync } from 'node:fs'
+import { appendFileSync, existsSync, mkdirSync, readFileSync } from 'node:fs'
 import { homedir } from 'node:os'
 import { dirname, join } from 'node:path'
 import type { VisitedPage } from '../browser/types.js'
+import { writeFileAtomic } from './settings-store.js'
 
 export type { VisitedPage }
 
@@ -51,6 +52,22 @@ export function visitedHistoryPath(): string {
  * a corrupted line must never hide the rest of the history.
  */
 export class HistoryStore {
+  /**
+   * File size at which the next rewrite happens.
+   *
+   * Advanced after each rewrite so the cost is paid once per PRUNE_SLACK appends
+   * rather than on every append: asking "is it over the cap?" was true from the first
+   * over-cap write onwards, which is what made every later navigation rewrite the file.
+   */
+  private pruneThreshold: number
+  /**
+   * Appends since the last rewrite.
+   *
+   * prune() runs after every navigation, and the answer is almost always "nothing to
+   * do" — but working that out from the file costs a full parse. Counting instead makes
+   * the common case a comparison.
+   */
+  private appendedSincePrune = 0
   private readonly maxEntries: number
   private readonly maxAgeMs: number
 
@@ -63,6 +80,8 @@ export class HistoryStore {
     limits: HistoryLimits = {},
   ) {
     this.maxEntries = limits.maxEntries ?? DEFAULT_MAX_ENTRIES
+    // The first rewrite is due once the file has grown PRUNE_SLACK past the cap.
+    this.pruneThreshold = this.maxEntries + this.slack()
     this.maxAgeMs = (limits.maxAgeDays ?? DEFAULT_MAX_AGE_DAYS) * 24 * 60 * 60 * 1000
   }
 
@@ -84,6 +103,7 @@ export class HistoryStore {
     } catch {
       // Best-effort: history is not worth failing a navigation over.
     }
+    this.appendedSincePrune++
   }
 
   /**
@@ -123,7 +143,22 @@ export class HistoryStore {
    * the count cap. Cheap no-op until the file grows past the cap, so callers can
    * run it after every append.
    */
+  /**
+   * How far past the cap the file may grow before it is rewritten.
+   *
+   * Scaled to the cap rather than fixed: a fixed 500 was meaningless when the cap is
+   * small (a caller asking for three entries would never reach 503), and the slack is
+   * there to batch rewrites, not to overrule the caller's bound.
+   * @returns the number of extra entries tolerated before a rewrite.
+   */
+  private slack(): number {
+    return Math.min(PRUNE_SLACK, Math.max(1, Math.floor(this.maxEntries / 2)))
+  }
+
   prune(): void {
+    // Cheap first: past the cap but without enough new entries to justify a rewrite.
+    // Reading the file to reach this conclusion was the remaining per-navigation cost.
+    if (this.appendedSincePrune === 0) return
     const all = this.readAll()
     const cutoff = Date.now() - this.maxAgeMs
     const fresh = all.filter(page => page.at >= cutoff)
@@ -131,12 +166,24 @@ export class HistoryStore {
     // Only rewrite when something actually fell off, and only once the excess
     // is worth a full rewrite.
     if (kept.length === all.length) return
-    if (all.length - kept.length < PRUNE_SLACK && fresh.length <= this.maxEntries) return
+    // Only rewrite once the file has grown PRUNE_SLACK past what we last wrote. Asking
+    // "is the file over the cap?" instead was true from the first over-cap append
+    // onwards, so every single navigation after that rewrote the whole file.
+    if (all.length < this.pruneThreshold) return
     try {
       const body = kept.map(page => `${JSON.stringify(page)}\n`).join('')
-      writeFileSync(this.file, body)
+      // Atomic, like the settings document. A plain writeFileSync truncates first, and this is
+      // the whole history in one call: a TerminateProcess mid-write — the browser host is
+      // killed exactly that way on Windows — left a partial file, and the reader drops a
+      // half-written TAIL rather than the middle, so everything after the tear was lost.
+      writeFileAtomic(this.file, body)
+      // The file now holds only what was kept, so the next rewrite is due once
+      // PRUNE_SLACK more entries have been appended.
+      this.pruneThreshold = kept.length + this.slack()
+      this.appendedSincePrune = 0
     } catch {
-      // Best-effort: an unpruned file still reads correctly.
+      // Best-effort: an unpruned file still reads correctly, and the atomic write leaves the
+      // previous one in place when it fails.
     }
   }
 

package/src/browser-electron/host-main.ts

@@ -50,26 +50,40 @@ try {
 }
 
 /**
- * Read the spawn token. The parent sends it two ways and the child prefers
- * the first that yields a value:
+ * Read the spawn token. The parent sends it two ways:
  *   1. stdin (first line) — the secure channel on Unix/macOS;
  *   2. DSH_BROWSER_RPC_TOKEN env — the fallback on Windows, where Electron is
  *      a GUI-subsystem process and never receives piped stdin (the parent
  *      sets it in the spawn env for exactly this reason).
- * stdin wins over env so the parent's token never has to touch argv.
+ *
+ * stdin WINS over env, which is what the ordering here has to express. It used to read env
+ * first and only consult stdin when that came up empty, so the stdin branch never ran while
+ * env was set — every spawn was told both ways, so the documented preference was silently
+ * the other way round. The env value is kept as the default for the common case where stdin
+ * delivers nothing; the interface below overwrites it when a line arrives.
  */
 let RPC_TOKEN = process.env.DSH_BROWSER_RPC_TOKEN ?? ''
-const stdinDelivered = RPC_TOKEN !== ''
-if (process.stdin !== null && process.stdin.readable) {
-  const stdinLines = createInterface({ input: process.stdin, terminal: false })
+/** True once a line actually arrived on stdin (as opposed to env being present). */
+let tokenFromStdin = false
+/**
+ * The one and only stdin reader.
+ *
+ * It is created here and reused by the hello path inside `whenReady` below: two
+ * `createInterface` calls on the same stream both consume its lines, so the second one saw
+ * nothing at all. A single reader means whoever needs the token from stdin waits on it.
+ */
+const stdinLines = process.stdin !== null && process.stdin.readable
+  ? createInterface({ input: process.stdin, terminal: false })
+  : undefined
+if (stdinLines !== undefined) {
   stdinLines.on('line', (line) => {
-    if (RPC_TOKEN === '') {
-      RPC_TOKEN = line.trim()
-      stdinLines.close()
-    }
+    if (tokenFromStdin) return
+    tokenFromStdin = true
+    const token = line.trim()
+    if (token !== '') RPC_TOKEN = token
   })
   stdinLines.on('close', () => {
-    if (RPC_TOKEN === '' && !stdinDelivered) {
+    if (RPC_TOKEN === '') {
       process.stderr.write('[dsh-browser host] warning: no token received on stdin or env\n')
     }
   })
@@ -797,11 +811,19 @@ async function handle(op: string, msg: { id: number; viewId?: string; windowId?:
         try {
           let image
           try {
-            image = await entry.webContentsView.webContents.capturePage()
+            image = await withTimeout(
+              entry.webContentsView.webContents.capturePage(),
+              COMMAND_TIMEOUT_MS,
+              'capturePage',
+            )
           } catch (error) {
             process.stderr.write(`[dsh-browser host] capturePage failed: ${String(error)}\n`)
             await new Promise(resolve => setTimeout(resolve, 400))
-            image = await entry.webContentsView.webContents.capturePage()
+            image = await withTimeout(
+              entry.webContentsView.webContents.capturePage(),
+              COMMAND_TIMEOUT_MS,
+              'capturePage',
+            )
           }
           // Downscale to fit the requested box, preserving aspect ratio.
           const size = image.getSize()
@@ -831,7 +853,11 @@ async function handle(op: string, msg: { id: number; viewId?: string; windowId?:
             try { win.window.contentView.removeChildView(toolbar) } catch { /* already gone */ }
           }
           try {
-            const shot = await entry.webContentsView.webContents.debugger.sendCommand('Page.captureScreenshot', {})
+            const shot = await withTimeout(
+              entry.webContentsView.webContents.debugger.sendCommand('Page.captureScreenshot', {}),
+              COMMAND_TIMEOUT_MS,
+              'Page.captureScreenshot',
+            )
             const data = (shot as { data?: unknown }).data
             if (typeof data === 'string' && data.length > 0) base64 = data
           } finally {
@@ -869,7 +895,12 @@ async function handle(op: string, msg: { id: number; viewId?: string; windowId?:
         // the parent's memory or hit the single-line cap. This also avoids
         // Electron's download pipeline entirely (CDP debugger attach can
         // interfere with will-download).
-        const result = await entry.webContentsView.webContents.debugger.sendCommand('Runtime.evaluate', {
+        // Bounded like every other op. The evaluate awaits a page-side fetch, so a slow or
+        // unresponsive server held the serial queue open forever: handle() never settled,
+        // every later op queued behind it, and the parent's plain timeout does not restart
+        // the host.
+        const result = await withTimeout(
+          entry.webContentsView.webContents.debugger.sendCommand('Runtime.evaluate', {
           // Stream the body through a reader so the size cap is enforced as
           // bytes arrive — never buffer an unbounded download into memory
           // before checking the limit (a huge URL would otherwise OOM the
@@ -910,7 +941,10 @@ async function handle(op: string, msg: { id: number; viewId?: string; windowId?:
           })()`,
           awaitPromise: true,
           returnByValue: true,
-        })
+          }) as Promise<{ result?: { value?: unknown }; exceptionDetails?: { exception?: { description?: string } } }>,
+          COMMAND_TIMEOUT_MS,
+          'download evaluate',
+        )
         const value = (result as { result?: { value?: unknown } }).result?.value
         if (typeof value !== 'string') {
           const detail = (result as { exceptionDetails?: { exception?: { description?: string } } }).exceptionDetails
@@ -939,7 +973,25 @@ async function handle(op: string, msg: { id: number; viewId?: string; windowId?:
         if (win === undefined || entry === undefined) throw new Error(`flushAuth: unknown view ${viewId}`)
         // Export the session's cookies so login state can be saved/restored
         // across browser hosts (or shared with another machine).
-        const cookies = await entry.webContentsView.webContents.session.cookies.get({})
+        //
+        // Scoped to this view's own page. The jar is SHARED on purpose: views here are built
+        // without a partition and every window drives the same browser, so signing in once
+        // works everywhere. What must not be shared is the HANDING OVER — an unfiltered read
+        // would put every domain every task has visited, plus whatever the human is signed
+        // into, in front of one task's caller. Least exposure, not isolation. An unknown page
+        // refuses rather than widening.
+        //
+        // Bounded like every other op: a renderer that stops answering here would hold the
+        // serial queue forever, and the parent's plain timeout does not restart the host.
+        const pageUrl = entry.webContentsView.webContents.getURL()
+        if (pageUrl === '' || pageUrl === 'about:blank') {
+          throw new Error('flushAuth: this view has no page loaded, so the cookie scope is unknown — reading everything would expose other sessions')
+        }
+        const cookies = await withTimeout(
+          entry.webContentsView.webContents.session.cookies.get({ url: pageUrl }),
+          COMMAND_TIMEOUT_MS,
+          'cookies.get',
+        )
         const exported = cookies.map(c => {
           // Electron types the cookie domain as optional; a missing domain
           // cannot be exported meaningfully, so default to '' (matches the
@@ -973,16 +1025,22 @@ async function handle(op: string, msg: { id: number; viewId?: string; windowId?:
         let restored = 0
         for (const c of cookies as Array<{ url?: string; name?: string; value?: string; domain?: string; path?: string; secure?: boolean; httpOnly?: boolean; expirationDate?: number }>) {
           if (typeof c.url !== 'string' || typeof c.name !== 'string' || typeof c.value !== 'string') continue
-          await entry.webContentsView.webContents.session.cookies.set({
-            url: c.url,
-            name: c.name,
-            value: c.value,
-            ...typeof c.domain === 'string' ? { domain: c.domain } : {},
-            ...typeof c.path === 'string' ? { path: c.path } : {},
-            ...typeof c.secure === 'boolean' ? { secure: c.secure } : {},
-            ...typeof c.httpOnly === 'boolean' ? { httpOnly: c.httpOnly } : {},
-            ...typeof c.expirationDate === 'number' ? { expirationDate: c.expirationDate } : {},
-          })
+          // Bounded per cookie: this runs once for every cookie, so a single hang held the
+          // serial queue open forever with no path to recovery.
+          await withTimeout(
+            entry.webContentsView.webContents.session.cookies.set({
+              url: c.url,
+              name: c.name,
+              value: c.value,
+              ...typeof c.domain === 'string' ? { domain: c.domain } : {},
+              ...typeof c.path === 'string' ? { path: c.path } : {},
+              ...typeof c.secure === 'boolean' ? { secure: c.secure } : {},
+              ...typeof c.httpOnly === 'boolean' ? { httpOnly: c.httpOnly } : {},
+              ...typeof c.expirationDate === 'number' ? { expirationDate: c.expirationDate } : {},
+            }),
+            COMMAND_TIMEOUT_MS,
+            'cookies.set',
+          )
           restored++
         }
         reply(msg.id, { ok: true, result: { restored } })
@@ -1050,16 +1108,18 @@ void app.whenReady().then(() => {
     socket.write(JSON.stringify({ id: 0, op: 'hello', token: RPC_TOKEN }) + '\n')
   }
   if (RPC_TOKEN !== '') {
+    // Either stdin already delivered a line (it wins) or env supplied the fallback.
     writeHello()
   } else {
-    // No token yet (env absent and stdin still pending): wait for stdin's
-    // first line with a bounded fallback, then send whatever we have — an
-    // empty hello is refused by the parent, which then fails cleanly.
+    // Neither yet: wait for stdin's first line with a bounded fallback, then send whatever
+    // we have — an empty hello is refused by the parent, which then fails cleanly. The
+    // reader is the module-level one, NOT a second createInterface on the same stream: two
+    // readers of one stdin both consume its lines, so the second one never saw the token.
     const fallback = setTimeout(() => writeHello(), 2_000)
-    if (process.stdin !== null && process.stdin.readable) {
-      const stdinLines = createInterface({ input: process.stdin, terminal: false })
-      stdinLines.once('line', () => { clearTimeout(fallback); writeHello() })
-      stdinLines.once('close', () => { clearTimeout(fallback); writeHello() })
+    if (stdinLines !== undefined) {
+      const send = (): void => { clearTimeout(fallback); writeHello() }
+      stdinLines.once('line', send)
+      stdinLines.once('close', send)
     }
   }
   const rl = createInterface({ input: socket })
@@ -1092,7 +1152,17 @@ void app.whenReady().then(() => {
   // without cleanup, exit so no zombie Electron window is left behind.
   socket.on('close', () => {
     process.stderr.write('[dsh-browser host] parent connection closed, exiting\n')
-    app.exit(0)
+    // A normal quit, not `app.exit(0)`. exit() skips before-quit/will-quit and does not wait
+    // for Chromium to finish writing the profile — the leveldb files holding cookies and site
+    // data. This path runs on every dispose, and a session that has just logged in is exactly
+    // when the profile holds unwritten state, so a hard exit could leave a half-written leveldb
+    // behind, which is what makes a browser profile fail to open on the next launch.
+    app.quit()
+    // Belt and braces: the parent is already gone, so there is nothing left to serve. If a
+    // listener or a window blocks the quit, forcing the exit after a grace period is still
+    // strictly better than an orphaned Electron window the user has to close by hand.
+    const force = setTimeout(() => app.exit(0), 5000)
+    force.unref()
   })
   // Keep the process alive until the parent closes the socket or kills us.
 })