[Bug]: Proxy auth without page.route() causes redirects to hang in Chromium

[andypple83]

Version

1.60.0-next (reproducible since 1.57.0)

Steps to reproduce

1. Launch Chromium with proxy credentials:

const browser = await chromium.launch({
  proxy: { server: 'http://myproxy:8080', username: 'user', password: 'secret' }
});

2. Create a page without registering any page.route() or context.route()
3. Navigate to a URL that returns a 302/303 redirect (e.g. Google search, any OAuth flow)

  const page = await browser.newPage();
  await page.goto('http://example.com/page-that-redirects'); // hangs forever

Expected behavior

The redirect is followed and the page loads normally, as it does in Playwright 1.56.0.

Actual behavior

page.goto hangs indefinitely. The redirect's Fetch.requestPaused event is stored in _requestIdToRequestPausedEvent waiting for a Network.requestWillBeSent that never arrives.

Adding a no-op page.route('**/*', route => route.continue()) makes it work, because a RouteImpl is created.

Additional context

• Works with page.route() registered (a RouteImpl is created)
• Works in Playwright 1.56.0
• Chromium-specific (Firefox/WebKit handle proxy auth differently)
• Related: context.route() causes click to hang on POST 303 redirect to accounts.google.com #39823, [Bug]: page.goto times out on share.google URLs using chromium #39299, [BUG] Private Network Requests Hang When Route Is Registered #12148

Environment

System:
    OS: macOS 26.3.1
    CPU: (14) arm64 Apple M4 Max
    Memory: 7.96 GB / 36.00 GB
  Binaries:
    Node: 25.9.0 - /opt/homebrew/bin/node
    Yarn: 1.22.22 - /opt/homebrew/bin/yarn
    npm: 11.12.1 - /opt/homebrew/bin/npm
    bun: 1.3.5 - /Users/andy/.bun/bin/bun
  Languages:
    Bash: 3.2.57 - /bin/bash

[Akashgite1]

Is the expected fix to ensure RouteImpl is initialized automatically?

[andypple83]

RouteImpl exists for user interception. Creating RouteImpl would add complexity, such as calling route.continue(). I have a simple fix to just auto-continue.

I need to verify this patch works with my local script before creating the PR.

[yury-s]

Please share complete repro. I was not reproduce it locally. Also both tests from your change pass with current implementation when I run them.

[yury-s]

Here is a test with redirect that works just fine too.

[andypple83]

Thanks @yury-s for looking into this. Very appreciated your help 👍

I can not reproduce with a local proxy. It only happens with a real external HTTP proxy, requesting the Google search page.

Reproduce

I use this script to reproduce:

  # test_proxy_hang.py
  # Requires: pip install playwright && python -m playwright install chromium
  # Run: python test_proxy_hang.py

  import asyncio
  from playwright.async_api import async_playwright

  # Replace with any real HTTP proxy that requires Basic auth
  PROXY = {"server": "http://<proxy-host>:<port>", "username": "<user>", "password": "<pass>"}

  async def main():
      url = "https://www.google.com/search?q=weather&sourceid=chrome&ie=UTF-8"
      total = 5

      async with async_playwright() as p:
          browser = await p.chromium.launch(headless=True)

          fail1 = 0
          print("--- No route ---")
          for i in range(total):
              ctx = await browser.new_context(proxy=PROXY)
              page = await ctx.new_page()
              try:
                  await page.goto(url, timeout=30000, wait_until="domcontentloaded")
                  print(f"Run {i+1}/{total}: OK")
              except Exception as e:
                  fail1 += 1
                  print(f"Run {i+1}/{total}: {type(e).__name__}")
              await ctx.close()

          fail2 = 0
          print("\n--- With route ---")
          for i in range(total):
              ctx = await browser.new_context(proxy=PROXY)
              page = await ctx.new_page()
              await page.route("**/*", lambda route: route.continue_())
              try:
                  await page.goto(url, timeout=30000, wait_until="domcontentloaded")
                  print(f"Run {i+1}/{total}: OK")
              except Exception as e:
                  fail2 += 1
                  print(f"Run {i+1}/{total}: {type(e).__name__}")
              await ctx.close()

          await browser.close()
      print(f"\nNo route:   {fail1}/{total} failed")
      print(f"With route: {fail2}/{total} failed")

  asyncio.run(main())

Results:

Results on Playwright 1.58.0, Chromium 145, macOS:
  --- No route ---
  Run 1/5: TimeoutError
  Run 2/5: TimeoutError
  Run 3/5: TimeoutError
  Run 4/5: TimeoutError
  Run 5/5: TimeoutError

  --- With route ---
  Run 1/5: OK
  Run 2/5: OK
  Run 3/5: OK
  Run 4/5: OK
  Run 5/5: OK

  No route:   5/5 failed
  With route: 0/5 failed

CDP Protocol traces

I use this script:

  # test_cdp_trace.py
  # Single-run script for capturing CDP protocol traces.
  #
  # Usage:
  #   DEBUG=pw:protocol python test_cdp_trace.py <url> 2>trace.log
  #   DEBUG=pw:protocol python test_cdp_trace.py <url> --route 2>trace.log
  #
  # Then extract relevant events:
  #   grep -E 'Fetch\.(requestPaused|continueRequest|authRequired|continueWithAuth)|Network\.requestWillBeSent[^E]' trace.log

  import asyncio
  import sys
  from playwright.async_api import async_playwright

  PROXY = {"server": "http://<proxy-host>:<port>", "username": "<user>", "password": "<pass>"}

  async def main():
      url = sys.argv[1] if len(sys.argv) > 1 else "https://www.google.com/search?q=weather&sourceid=chrome&ie=UTF-8"
      use_route = "--route" in sys.argv

      async with async_playwright() as p:
          browser = await p.chromium.launch(headless=True)
          ctx = await browser.new_context(proxy=PROXY)
          page = await ctx.new_page()
          if use_route:
              await page.route("**/*", lambda route: route.continue_())
              print(f"MODE: with route, url={url}", file=sys.stderr)
          else:
              print(f"MODE: no route, url={url}", file=sys.stderr)
          try:
              await page.goto(url, timeout=20000, wait_until="domcontentloaded")
              print(f"OK - {page.url}", file=sys.stderr)
          except Exception as e:
              print(f"{type(e).__name__}: {str(e).splitlines()[0][:80]}", file=sys.stderr)
          await browser.close()

  asyncio.run(main())

I tried to extract the CDP protocol traces with (Chromium 145)

# Google, no route (HANGS)
# DEBUG=pw:protocol python test_cdp_trace.py "https://www.google.com/search?q=weather" 2>trace_google_no_route.log

Google, no route (hangs) — Chrome creates a new interception job after proxy auth:
  <<< Network.requestWillBeSent       requestId=4AF08...
  <<< Fetch.requestPaused             job-1.0  networkId=4AF08...
  >>> Fetch.continueRequest            job-1.0                      (auto-continued)
  <<< Fetch.authRequired              job-1.0                      (proxy 407)
  >>> Fetch.continueWithAuth           job-1.0                      (credentials sent)
  <<< Fetch.requestPaused             job-2.0  networkId=4AF08...   (NEW job, no redirectedRequestId)
  *** HUNG — no Fetch.continueRequest sent, no Network.requestWillBeSent arrives ***

# Google, with route (WORKS)
DEBUG=pw:protocol python test_cdp_trace.py "https://www.google.com/search?q=weather" --route 2>trace_google_with_route.log

  Google, with route (works) — job-2.0 is continued by the route handler:
  <<< Network.requestWillBeSent       requestId=EBF86...
  <<< Fetch.requestPaused             job-1.0  networkId=EBF86...
  >>> Fetch.continueRequest            job-1.0                      (user handler continues)
  <<< Fetch.authRequired              job-1.0
  >>> Fetch.continueWithAuth           job-1.0
  <<< Fetch.requestPaused             job-2.0  networkId=EBF86...
  >>> Fetch.continueRequest            job-2.0                      ← continued by route handler
  ... navigation proceeds normally

# httpbin, no route (WORKS — for comparison)
DEBUG=pw:protocol python test_cdp_trace.py "https://httpbin.org/redirect-to?url=https://httpbin.org/get&status_code=302" 2>trace_httpbin_no_route.log

  httpbin, no route (works) — Chrome keeps the same interception job family:
  <<< Network.requestWillBeSent       requestId=1A433...
  <<< Fetch.requestPaused             job-1.0  networkId=1A433...
  >>> Fetch.continueRequest            job-1.0
  <<< Fetch.authRequired              job-1.0                      (proxy 407)
  >>> Fetch.continueWithAuth           job-1.0
  <<< Network.requestWillBeSent       requestId=1A433... REDIRECT   ← arrives
  <<< Fetch.requestPaused             job-1.1  redirectedFrom=job-1.0  ← same family
  >>> Fetch.continueRequest            job-1.1                      ← matched and continued

Note: I cannot reproduce this in Chromium 147.

[yury-s]

We cannot act on this without a full repro.

Note: I cannot reproduce this in Chromium 147.

If that's the case, it will be fixed in 1.60 release anyway. You can try it already with canary release.