[actions] Handling of "display:none" element as origin

[jgraham]

This should probably return "move target out of bounds" or something, but at the moment the spec algorithm doesn't account for the possibility of an element without bounding client rects.

[juliandescottes]

For https://bugzilla.mozilla.org/show_bug.cgi?id=1805146 I reviewed the behavior of various browsers when trying to trigger a click on an option element of a select (which has no bounding client rect). My test code was:

def test_click_select(session, url, inline, mouse_chain):
    session.url = inline("""
            <select>
                <option value="foo">Foo</option>
                <option value="bar">Bar</option>
            </select>
        """)

    select = session.find.css("select", all=False)
    mouse_chain.click(element=select).perform()

    option = session.find.css("option[value='bar']", all=False)
    mouse_chain.click(element=option).perform()

    value = session.execute_script("return document.querySelector('select').value;")
    assert value == "bar"

And testing on various browsers, I had the following results:

• Firefox fails with webdriver.error.MoveTargetOutOfBoundsException: move target out of bounds (500): Origin element is not displayed
• Chrome fails with webdriver.error.WebDriverException: element not interactable (400): element not interactable: [object HTMLOptionElement] has no size and location
• Safari completes the test without throwing any error but fails the assert, meaning it didn't actually click on the option

(see the bug for results with older Firefox versions, as I was trying to check if there had been a regression on that front)

[whimboo]

Note that this as well affects elements with display: contents.

[css-meeting-bot]

The Browser Testing and Tools Working Group just discussed Display: None and DIsplay:contents.

The full IRC log of that discussion

<AutomatedTester> Topic: Display: None and DIsplay:contents
<AutomatedTester> github: https://github.com//issues/1642
<AutomatedTester> whimboo: we've had this issue open for a while
<AutomatedTester> ... I noticed that if you do actions we have difference of how browsers do this. E.g. Firefox errors and the others dont
<AutomatedTester> ... it would be good to define what the behaviour should be here
<AutomatedTester> ... when speaking to jgraham we should likely error
<AutomatedTester> q+
<AutomatedTester> ack next
<AutomatedTester> automatedtester: with display:none do we know the location of the element?
<AutomatedTester> jgraham: no, we don't. Display none has nothing in the layout
<AutomatedTester> q?
<AutomatedTester> automatedtester: with actions we should be going through the glass so I ma not sure the correct way here? Should we go up a level in the DOM? or just error?
<AutomatedTester> jgraham: I think we should solve this with an error as this is an edge case. Safari doesn't error but doesn't do anything. It would be to feed back to the end user
<AutomatedTester> whimboo: playwright does one thing different to use with display: contents
<AutomatedTester> q?
<AutomatedTester> action: Spec to be updated to throw an error for display:none
<AutomatedTester> action: whimboo to write a new issue for display:contents with test case and bring it back to the WG

[whimboo]

Note that this as well affects elements with display: contents.

This is actually a different issue and is not related to perform actions. In Playwright they call scrollIntoView() for the element via script evaluation. So it can be handled separately and we can ignore for now.

[whimboo]

This is actually a different issue and is not related to perform actions. In Playwright they call scrollIntoView() for the element via script evaluation. So it can be handled separately and we can ignore for now.

I had the wrong test in mind but now found the one I was originally referring to. So it's valid for Perform Actions and I filed issue #1887.