Proxy authentication fails when using .get() from the Tab class

[nyx-rattapoom]

When I navigate with page.get() on a Tab object, proxy credentials are not sent, because seleniumbase.undetected.cdp_driver.connection.Connection.set_auth() is never invoked.
If I first call driver.get() (on the Browser object), the proxy authentication does work.

Example code (two work-arounds)

import asyncio
import time

from seleniumbase import cdp_driver

proxy_address = "http://xx.xx.xx.xx:yyyy"
proxy_user = "u"
proxy_password = "p"
proxy = f"{proxy_user}:{proxy_password}@{proxy_address}"


async def workaround_a():
    driver = await cdp_driver.start_async(lang="en", proxy=proxy)
    page = driver.main_tab
    # workaround
    if proxy_user and proxy_password:
        await driver.connection.set_auth(proxy_user, proxy_password, page)
        time.sleep(0.25)
    await ip_check(page)
    driver.stop()


async def workaround_b():
    driver = await cdp_driver.start_async(lang="en", proxy=proxy)
    # workaround
    if proxy_user and proxy_password:
        await driver.get("about:blank")
    page = driver.main_tab
    await ip_check(page)
    driver.stop()


async def ip_check(page):
    await page.get("https://api.ipify.org/")
    body_element = await page.select("body")
    ip_address = body_element.text
    if not ip_address or "ERR" in ip_address:
        raise Exception("Failed to determine IP Address!")
    print(f"My IP Address = {ip_address}")


if __name__ == "__main__":
    loop = asyncio.new_event_loop()
    loop.run_until_complete(workaround_a())
    loop.run_until_complete(workaround_b())

SeleniumBase version: 4.40.6

[mdmintz]

When using the async format, the page object must be created from a driver.get(URL) call.
Eg. page = await driver.get(url).
From there, you can use page.get(URL) to go to a new URL if you need to.

Sample code with proxy:

import asyncio
import time
from seleniumbase import cdp_driver

async def main():
    url = "https://api.ipify.org/"
    driver = await cdp_driver.start_async(proxy="USER:PASS@IP:PORT")
    page = await driver.get(url)
    body_element = await page.select("body")
    ip_address = body_element.text
    print(ip_address)

    time.sleep(1)
    await page.get("about:blank")
    time.sleep(1)

    await page.get(url)
    body_element = await page.select("body")
    ip_address = body_element.text
    print(ip_address)

    driver.stop()

if __name__ == "__main__":
    # Call an async function with awaited methods
    loop = asyncio.new_event_loop()
    loop.run_until_complete(main())

[nyx-rattapoom]

Which mean If I want to encapsulate driver (Browser) and allow to use only Page object,
I'm require to return driver.get("about:blank") instead to ensure proxy setup has initiated ?

I just think the proxy should be inited from start_async and not driver.get

[mdmintz]

Due to https://groups.google.com/a/chromium.org/g/chromium-extensions/c/aEHdhDZ-V0E/m/UWP4-k32AgAJ and https://chromium.googlesource.com/chromium/src/+/290ed8046692651ce76088914750cb659b65fb17%5E%21/chrome/browser/extensions/extension_service.cc, which removed the --load-extension switch on Chrome builds, setting a proxy must be done via CDP, rather than by extension if you're using the async format (or other pure CDP format).

That's why proxy takes effect here: page = await driver.get(url), when page is initialized via the driver.get(url) call.