ci: de-flake drive gate — drop the racy iframe probe, keep input/canvas checks

Re-running the enriched gate on the real binaries exposed a ~1-in-5 flake: the
iframe probe (nested data: / srcdoc) re-navigates, so Juggler's frame id changes
mid-check ("execution context destroyed" / "Frame was detached"). set_content
isn't an option either — this build rejects its document.write ("operation is
insecure").

Drop the iframe from the gate: a same-origin srcdoc iframe is a weak proxy for
the cross-origin issue #20 anyway. The page is now a plain subframe-free
goto(data:), and the mouse/keyboard/canvas-determinism/navigator-surface checks
(the firefox-2 class + stealth smoke) stay. 20/20 clean locally. The faithful
#20 sentinel (tests/test_cross_origin_iframe.py, two localhost origins) should
be wired as its own e2e gate job.
This commit is contained in:
feder-cr 2026-06-09 14:35:50 +02:00
parent 8d7b6eafdf
commit 90529ff181

View file

@ -10,17 +10,28 @@ It deliberately covers the failure modes that HISTORICALLY shipped green:
- juggler missing entirely TargetClosedError on launch (firefox-8)
- mouse/keyboard input broken click/move/type assertions (firefox-2 #9:
jugglerSendMouseEvent / synthesizeMouseEvent)
- cross-origin iframe broken content_frame() reachable (issue #20)
- canvas non-deterministic identical draw identical dataURL (stealth
seed must be per-session, not per-readback)
- headless navigator tells navigator.webdriver falsy, languages
non-empty, plugins is a real PluginArray
All of this is headless, NO screenshot GPU-free (can't false-fail on the
GPU-less hosted runners), and fully offline data: URLs only, NO network, NO
proxy, NO secrets safe in public CI. WebGL determinism is intentionally NOT
checked here (it needs SWGL and can false-fail headless); it lives in the local
proxy realness gate, alongside the fingerprint/WebRTC-vs-vanilla checks.
GPU-less hosted runners), and fully offline safe in public CI. WebGL
determinism is intentionally NOT checked here (it needs SWGL and can false-fail
headless); it lives in the local proxy realness gate.
NOT covered here on purpose:
- Cross-origin iframe (issue #20): a same-origin srcdoc/data iframe is a weak
proxy for it AND races Juggler's frame tracking (the frame re-navigates, its
id changes "Frame was detached" ~1-in-8). The faithful #20 sentinel is
`tests/test_cross_origin_iframe.py` (e2e, two localhost origins); wire that
as its own gate job rather than a fragile in-gate check.
Robustness (learned the hard way): the page is a SIMPLE
`goto("data:text/html,...")` with NO subframe. `set_content` throws "The
operation is insecure" on this build (its document.write is rejected), and a
nested `data:`/srcdoc iframe races the evaluates intermittent "execution
context destroyed by navigation" / "Frame was detached".
Usage: python ci_drive_gate.py /path/to/firefox[.exe | .app/Contents/MacOS/firefox]
Exit 0 + "DRIVE GATE OK ..." on success; non-zero with a reason on failure.
@ -31,18 +42,13 @@ import sys
from playwright.sync_api import sync_playwright
# Single offline page that wires up every probe: a clickable button, a text
# input, a same-document iframe, and a mousemove counter. data: URLs execute
# inline scripts and are same-origin, so this needs no server.
# Simple, subframe-free data: URL — proven stable across runners.
PAGE = (
"data:text/html,"
"<title>dt</title>"
"<h1 id=x>hello-drive</h1>"
"<button id=b onclick=\"window.__clicked=1\">go</button>"
"<input id=inp>"
"<iframe id=f src=\"data:text/html,<b id=ok>ok</b>\"></iframe>"
"<script>window.__moves=0;"
"addEventListener('mousemove',function(){window.__moves++})</script>"
)
# Identical 2D draw, evaluated twice in one session. The stealth canvas spoof is
@ -59,13 +65,16 @@ def main(exe: str) -> int:
with sync_playwright() as p:
browser = p.firefox.launch(executable_path=exe, headless=True)
page = browser.new_page()
page.goto(PAGE)
page.goto(PAGE) # default wait_until="load"; no subframe → settles cleanly
# Attach the mousemove counter explicitly (don't depend on inline-script timing).
page.evaluate("window.__moves = 0; window.addEventListener('mousemove', () => { window.__moves++; })")
ua = page.evaluate("navigator.userAgent")
webdriver = page.evaluate("navigator.webdriver")
text = page.evaluate("() => document.getElementById('x').textContent")
# firefox-2 / issue-#9 catcher: real mouse + keyboard over juggler.
page.wait_for_selector("#b")
page.mouse.move(20, 20)
page.mouse.move(120, 90) # exercises synthesizeMouseEvent path
page.click("#b") # mousedown/up/click → onclick fires
@ -75,13 +84,6 @@ def main(exe: str) -> int:
moves = page.evaluate("window.__moves")
typed = page.evaluate("() => document.getElementById('inp').value")
# issue-#20 catcher: the iframe must be reachable and runnable.
frame_el = page.query_selector("#f")
frame = frame_el.content_frame() if frame_el else None
iframe_ok = bool(frame) and frame.evaluate(
"() => document.getElementById('ok') && document.getElementById('ok').textContent"
) == "ok"
# stealth-determinism catcher: identical draw → identical dataURL.
canvas_a = page.evaluate(CANVAS_DRAW)
canvas_b = page.evaluate(CANVAS_DRAW)
@ -98,14 +100,13 @@ def main(exe: str) -> int:
assert clicked == 1, "page.click() did not fire onclick — mouse-event synthesis broken (firefox-2 class)"
assert moves >= 1, "page.mouse.move() produced no mousemove — jugglerSendMouseEvent regression"
assert typed == "ok", f"page.keyboard.type() failed: {typed!r}"
assert iframe_ok, "iframe content_frame() unreachable — fission.webContentIsolationStrategy regression (issue #20)"
assert canvas_a == canvas_b, "canvas non-deterministic across identical draws (stealth seed broken → bot tell)"
assert langs and langs > 0, "navigator.languages empty (headless tell)"
assert plugins, "navigator.plugins is not a PluginArray (headless tell)"
print(
f"DRIVE GATE OK | UA={ua} | webdriver={webdriver} | "
f"click+mousemove+keyboard+iframe+canvas-determinism+navsurface=ok"
f"click+mousemove+keyboard+canvas-determinism+navsurface=ok"
)
return 0