[BUG] 'Request context disposed' was reported when use route.fetch()

[erma0]

Context:

• Playwright Version: 1.35.0
• Operating System: Windows 11
• Python version: 3.11.2
• Browser: Chromium-msedge
• Extra:

Code Snippet

from playwright.sync_api import Route, sync_playwright


def handle(route: Route):
    res = route.fetch()
    route.fulfill(response=res)


with sync_playwright() as playwright:
    browser = playwright.chromium.launch(channel='msedge', headless=False)
    context = browser.new_context()
    page = context.new_page()

    page.route("**/*", handle)
    page.goto("https://www.baidu.com/", wait_until='domcontentloaded')
    page.unroute("**/*") # not work

    page.close()
    context.close()
    browser.close()

Describe the bug
'Request context disposed' or 'Response context disposed'

I think it's because the page has been closed, but the route handle function is still running.

It appears to be related to microsoft/playwright-python#1402 microsoft/playwright-python#1433 microsoft/playwright-python#1458 and has been fixed, but this issue still persists.

How can I stop a running route handle function before the page is closed?

thanks.

[mxschmitt]

Reproducible with Node.js, moving upstream so it can be discussed there.

import { chromium } from 'playwright';

(async () => {
  const browser = await chromium.launch({
    headless: false
  });
  const context = await browser.newContext();
  const page = await context.newPage();
  await page.route('**/*', async route => {
    const response = await route.fetch();
    route.fulfill({
      response
    });
  });
  await page.goto('https://www.baidu.com/', {
    waitUntil: 'domcontentloaded'
  });
  // ---------------------
  await context.close();
  await browser.close();
})();

[dgozman]

Investigation notes:

• Playwright should not silently catch errors in user code to avoid masking problems.
• fetch() must throw when request cannot be made, because it returns a value.

Possible improvements, that could still lead to minor regression in existing test suites:

• unroute() to await all ongoing route handlers that it removes.
• page.close()/context.close() to automatically call unrouteAll() that would await ongoing handlers.
  Both improvements do not help with pages that were closed by the browser, for example with popupWindow.close().

[daelmaak]

This is still a problem, and I didn't find a solution to this except for not doing route.fetch() at all. Any update on this?

[nbaldzhiev]

I am getting Error: apiRequestContext.fetch: Request context disposed. errors, but in Webkit (playwright 1.38.0).
A related discussion I found is https://github.com/microsoft/playwright/discussions/12810.
The suggestion there is to add a page.waitForResponse call, but it doesn't always help in our case as we still see this failure from time to time.

Our code (redacted):

  await page.route(url, async route => {
    const response = await page.request.fetch(route.request())
    const body = await response.json()
    do-some-modifications-to-body
    route.fulfill({
      response,
      body: JSON.stringify(body)
    })
  })

  await page.waitForResponse(url, { timeout: 15 * 1000 })

[frakic]

A related discussion I found is https://github.com/microsoft/playwright/discussions/12810. The suggestion there is to add a page.waitForResponse call, but it doesn't always help in our case as we still see this failure from time to time.

The link's broken.

I'm getting this error on @playwright/test version 1.38.1, NodeJS v16.20.2 when trying to use context.request.fetch().

For some reason, I only experience this issue when I run my tests against an instance of my React app that is being run locally.

If I test against a deployed version of my React app, the command executes without a problem. I've even tried building the React app and serving it locally, but the issue persists. Kind of weird...

Anyway, I hope this gets fixed soon because it's slowing me down so much.

[btomaj]

I can confirm that neither page.waitForResponse() nor page.waitForLoadState("load" | "domcontentloaded" | "networkidle") solve this problem.

In my scenario, none of page, context, or browser are closed; my tests conclude before requests have been processed by page.route() in a fixture, causing subsequent tests to fail. This seems like a serious design flaw: the test causing the problem passes, causing the next test to fail when page.route() throws an error from the previous test.

The proposed solution to await all ongoing route handlers that unroute() removes doesn't solve for my scenario. Calling page.unroute() after await use() in a fixture throws the following error:

Error: page.unroute: Target closed

What's required to move forward and implement a fix?

[ovistoica]

Bump this. We have the same issue