2032066 - script.addPreloadScript causes a console error message on navigation

Status: RESOLVED FIXED

(Remote Protocol :: WebDriver BiDi, defect, P3)

Description

[Holger Benl [:hbenl]]

The following test checks that adding a preload script and navigating doesn't add any console messages:

async def test_add_preload_script_no_log_messages(
    bidi_session, add_preload_script, subscribe_events
):
    await add_preload_script(function_declaration="() => 42")

    log = []
    async def on_event(method, data):
        log.append(data)
    remove_listener = bidi_session.add_event_listener("log.entryAdded", on_event)

    new_context = await bidi_session.browsing_context.create(type_hint="tab")
    await subscribe_events(
        events=["log.entryAdded"], contexts=[new_context["context"]]
    )

    await bidi_session.browsing_context.navigate(
        context=new_context["context"],
        url="data:text/html,",
        wait="complete",
    )

    assert log == []

    remove_listener()

It fails due to an unexpected permission denied error:
E assert [{'type': 'javascript', 'level': 'error', 'source': {'realm': 'bf31ba61-dc38-4355-aeaf-69fd1a18e028', 'context': 'c950fb41-bb2a-42ae-a3e5-77338f9f0f59'}, 'text': 'Error: Permission denied to access property "length"', 'timestamp': 1776262656353}] == []

The issue causes this Playwright test to fail: it runs a Playwright test with tracing enabled (tracing is implemented using a preload script) and then checks the test's console messages in Playwright's trace viewer.

Comment 1

[Henrik Skupin [:whimboo][⌚️UTC+2] (away July 24th - August 16th)]

Sadly there is no stack available. Do you see in the browser logs more details? Would be good to know where exactly the error is raised. Is that maybe a recent regression?

Comment 2

[Holger Benl [:hbenl]]

I couldn't find any more details in the browser logs.
The error first showed up in Playwright CI on 2026-02-16 and then permanently from 2026-02-20. On 2026-02-19, D284029 was merged and when I undo the changes from that patch, the error disappears and the test passes. Unfortunately this doesn't tell us where the error occurs and I have no idea why it happened on 2026-02-16.

Comment 3

[BugBot [:suhaib / :marco]]

Set release status flags based on info from the regressing bug 1985997

:Sasha, since you are the author of the regressor, bug 1985997, could you take a look? Also, could you set the severity field?

For more information, please visit BugBot documentation.

Comment 4

[Henrik Skupin [:whimboo][⌚️UTC+2] (away July 24th - August 16th)]

Thanks, Holger. If this only results in a console error from a non-critical code path, it may be acceptable. What I am wondering, though, is whether it could actually prevent the preload script from being added. That would be a much more serious regression, and in that case I would expect we might already have seen related bug reports given that 149 was released nearly 4 weeks ago. Hopefully that is not happening, or only in a very specific edge case.

Comment 5

[Holger Benl [:hbenl]]

I found another Playwright test that fails because of this issue: the test blocks service workers (using a preload script) and verifies that no error is thrown on navigation to about:blank but it receives the permission denied error.
The test started permafailing on 2026-02-13 but worked consistently before that date.

Comment 6

[Henrik Skupin [:whimboo][⌚️UTC+2] (away July 24th - August 16th)]

(In reply to Holger Benl [:hbenl] from comment #5)

The test started permafailing on 2026-02-13 but worked consistently before that date.

Thanks. Bug 1936770 landed that day and seems to be related to this regression.

Comment 7

[Julian Descottes [:jdescottes]]

This is already marked as a regression from Bug 1985997, now we added Bug 1936770. Are both the same issue?

Comment 8

[Holger Benl [:hbenl]]

For now we can only speculate what regressed this. FWIW I don't think it's related to Bug 1936770, I rather suspect Bug 2005558 (which modified the same file as Bug 1985997, which I suspect is responsible for the first mentioned Playwright test starting to permafail, but it already had one failure before that patch). Bug 2005558 landed on 2026-02-12.

Comment 9

[Julian Descottes [:jdescottes]]

Makes sense, thanks for the details Holger.

Also Bug 1936770 landed on the 14th (first Nightly), and I don't know if PW was able to run with a Nightly containing this one on the 14th or the 15th, but if the first failure was on the 13th, it's very unlikely to be related. I will remove the regressed link, overall let's wait until we have some evidence before flagging bugs as regressors.

Comment 10

[Holger Benl [:hbenl]]

The Playwright tests started passing last night and the test in the description is also passing now. I assume this was fixed with Bug 2046390.

Comment 11

[Henrik Skupin [:whimboo][⌚️UTC+2] (away July 24th - August 16th)]

Marking for m20 but not tracking the points because the patch is on the other bug.

Comment 12

[Henrik Skupin [:whimboo][⌚️UTC+2] (away July 24th - August 16th)]

The patch on bug 2046390 was uplifted to 153.