Bug 127644 – onload event is never fired if script element was added from readystatechange listener

• Status: RESOLVED
• Resolution: CONFIGURATION CHANGED
• Component: UI Events

If the functions with the below ordering were called, I think 'onload' event is never fired.

1. FrameLoader::checkCompleted()
2. Document::setReadyState(Document::Complete); -- from checkCompleted()
3. HTMLScriptElement::HTMLScriptElement -- from listener of onreadystatechange
4. ScriptRunner::queueScriptForExecution()
5. Document::incrementLoadEventDelayCount() -- from queueScriptForExecution()
6. Document::checkCallImplicitClose() -- from checkCompleted()
---> But it will return immediately because of Document::isDelayingLoadEvent()==true
7. Document::decrementLoadEventDelayCount() is called and timer fired,
FrameLoader::checkCompleted() will be called. However, m_isComplete is already set to 'true' then
we cannot reach to DOMWindow::dispatchLoadEvent();

I think isDelayingLoadEvent() is necessary to be checked after onreadystatechange to be fired.
Firefox, Chrome v30 (not WebKit) on PC are OK.

[Benjamin Poulain]

For reference: https://lists.webkit.org/pipermail/webkit-help/2014-January/003723.html

Created attachment 222255 [details]
Test case for this bug to reproduce

Created attachment 222258 [details]
Patch

I'm not an expert of WebKit. Maybe this is incomplete and will make any side-effects.

[Deepak Mittal]

When isDelayingLoadEvent() is pending then m_complete should be fale.

Changes appears ok to me,

I have verified these changes also, It is working fine..

[Ahmad Saleem]

I am not clear on Expected result of test case but all browsers show following output (in Private / Incognito tab or window):

interactive
complete
Recived onload event

Browsers tested - Safari 15.6 on macOS 12.5 , Chrome Canary 105 and Firefox Nightly 104

If it is expected result, I think it can be marked as "RESOLVED CONFIGURATION CHANGED" or else test case can be updated etc. Just wanted to share updated results. Thanks!

[Ryosuke Niwa]

Yeah, this appears to be fixed now.