(Selenium) Chrome > 128 fails to run on lazy loaded iframes

[Zidious]

Since chrome >= 128 passing --headless will default to use --headless=new. This causes the lazy loaded iframe to have unwanted side effects when running axe and does not catch the #lazy-frame as part of the incomplete results. I have verified that this is also the case in fully headed mode. Note: this also is an issue with our @axe-core/webdriverjs integration when running fully headed or headless=new.

The test in question:

axe-core-maven-html/selenium/src/test/java/com/deque/html/axecore/selenium/Axe43xIntegrationTest.java

Lines 564 to 584 in 3d13cc7

	@Test
	public void withUnloadedIframes() {
	webDriver.get(fixture("/lazy-loaded-iframe.html"));
	String title = webDriver.getTitle();
	AxeBuilder axeBuilder = new AxeBuilder().withRules(Arrays.asList("label", "frame-tested"));
	Results axeResults = axeBuilder.analyze(webDriver);
	assertNotEquals(title, "Error");
	assertEquals(axeResults.getIncomplete().size(), 1);
	assertEquals(axeResults.getIncomplete().get(0).getId(), "frame-tested");
	assertEquals(axeResults.getIncomplete().get(0).getNodes().size(), 1);
	assertTargetEquals(
	axeResults.getIncomplete().get(0).getNodes().get(0).getTarget(),
	new String[] {"#ifr-lazy", "#lazy-iframe"});
	assertEquals(axeResults.getViolations().size(), 2);
	assertEquals(axeResults.getViolations().get(1).getId(), "label");
	assertEquals(axeResults.getViolations().get(1).getNodes().size(), 1);
	assertTargetEquals(
	axeResults.getViolations().get(1).getNodes().get(0).getTarget(),
	new String[] {"#ifr-lazy", "#lazy-baz", "input"});
	}

I have verified on Chomedriver 126 this issue doesn't appear. The intermediate workaround is to pass --headless=old as part of the chrome arguments which acts like the old --headless mode.

Follow ups include:

• Identifying if this is considered an axe-core bug, running in headed or headless=new mode does not cause the #lazy-iframe to be part of the incomplete results for the frame-tested rule. Note that this doesn't appear to be an issue with Playwright and might be a limitation on Seleniums side and/or a behavioural difference about how the lazy iframes are loaded.
• Once we've figured out a fix, we should go and update the test to account for the new changes and amend the headless flag to either the default behaviour or headless=new.