[Bug]: clock.pause_at() hangs when the page has an active requestAnimationFrame loop

[galipruzansky]

Version

1.59.0 (playwright-python), Chromium 147.0.7727.15, macOS (also reproduces with the default headless shell)

Steps to reproduce

import asyncio
from playwright.async_api import async_playwright

PAGE = "<script>function tick(){requestAnimationFrame(tick)}requestAnimationFrame(tick)</script>"

async def main():
 async with async_playwright() as pw:
 browser = await pw.chromium.launch(headless=True)
 page = await browser.new_page()
 await page.clock.install()
 await page.set_content(PAGE)
 await asyncio.sleep(0.3)

 for cycle in (1, 2):
 fake_now = int(await page.evaluate("Date.now()"))
 await asyncio.wait_for(page.clock.pause_at(fake_now), timeout=10)
 print(f"cycle {cycle}: pause_at OK")
 await page.clock.resume()
 await asyncio.sleep(0.3)
 await browser.close()

asyncio.run(main())

Expected behavior

Both pause_at calls return promptly — each pauses at the page's current fake time.

Actual behavior

The first install -> pause_at -> resume cycle works; the second pause_at never resolves (hangs indefinitely; the wait_for above times out):

cycle 1: pause_at OK
cycle 2: pause_at HANGS (>10s)

Reproduces identically on the default headless shell and channel="chromium". Removing the requestAnimationFrame loop from the page makes both cycles pass.

Additional context

• [Feature]: Pause function on Clock API playwright#32807 was closed recommending pauseAt as the way to pause time. For any rAF-driven page (games, canvas/WebGL apps), repeated freeze/resume via pause_at deadlocks on the second freeze, so that recommendation currently has no working path for such pages.
• Possibly related to [Bug]: requestAnimationFrame does not execute callbacks after page.clock.fastForward() for the duration of time that was fast forwarded playwright#37635 (rAF starvation during fastForward): pause_at fast-forwards to the target time internally (_innerFastForwardTo), and a rAF callback re-queues itself on every fired frame.

[Skn0tt]

Moving this over to upstream since it's not specific to the port.

[Skn0tt]

Apologies, it is specific to the port! JS throws, and Python hangs, and that's a bug. I'll take a look at how we can make Python throw as well.

[Skn0tt]

A weird bug pause_at, under Python, takes second values, whereas Date.now() produces millisecond values. Somewhere within Playwright, it is multiplied again and then goes out of Javascript's safe integer range, and that's where the hang comes from. I'll make us detect that upstream and throw a nice error about it.

I'll also update our docs on this on the Python end.

On your end, add a / 1000 to your code, and it'll work!