[Webkit-unassigned] [Bug 19418] onbeforeunload is broken for framesets

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=19418





--- Comment #16 from William J. Edney <bedney at technicalpursuit.com>  2009-11-21 16:31:23 PST ---
In hopes that someone out there on the Webkit team can address this issue soon,
here's my latest set of test files. I tried to make the messages a bit more
consistent and clear. These are marked as 'The frameset file v2' and 'The frame
content file v2'.

Unfortunately, the behavior that I noted in comment #1 hasn't improved. It is
consistently broken in the same way.


As of today, I also noticed the following behavior, which is also inconsistent
with how both IE and Mozilla work, and is yet another reason we're broken on
Webkit-based browsers.

In IE / Mozilla:

- Launch the v2 versions of these pages.
- Attempt to close the page.
- You'll see a dialog containing the 'onbeforeunload' message installed on the
frameset.
- Choose 'Ok' to 'continue closing' (if you choose 'Cancel', the system will -
properly in my opinion - not show you the second dialog)
- Then you'll see a *second* dialog containing the 'onbeforeunload' message
installed on the nested frame.

In Webkit-based browsers you *only see the first dialog*. The onbeforeunload
handler for the nested frame is not being executed.


The Chromium team has upstreamed their bug report on this behavior back to this
bug. Their report is here:

http://code.google.com/p/chromium/issues/detail?id=17157

They also mention this bug in their bug report. It's probably related:

https://bugs.webkit.org/show_bug.cgi?id=27481


To anyone on the Webkit team - this continues to badly break us. *Please* let
me know if I can help in any way insofar as more test cases, etc. I haven't
done C in many moons, so I can't help there unfortunately. Sam, I'm not sure if
the 'rdar' bug system reflects changes here, but if not please pass along these
comments. Also, feel free to email me or engage me in any fashion to help you
all fix this bug - I'll spend time explaining it over the phone, if that's
necessary - just let me know. As has been noted previously, its been a year. I
know you all are amazingly busy but for those of us writing stateful
applications, this is a deal breaker for Safari/Chrome/etc. support.

For my part, I've tested this on the following browsers. It is (thankfully)
consistent between all of them:

Safari 4.0.4 / Mac
Webkit 51280 / Mac
Google Chrome 4.0.249.4 / Mac
Google Chrome 4.0.249.4 / Windows

Safari 4.0.4 / Windows is a strange case indeed. Launching the frameset and
then using the link to go to Google doesn't light up the 'Back' button, such
that you can't traverse back. This very well may be another bug. I don't have a
Webkit nightly for Windows, but if there's interest, I can get one and retry on
that.

Thanks to all in advance. Let me know if beer and/or wine bribery will be
involved :-) (I know for a fact that the Mozillians can be bribed ;-) ).

Cheers,

- Bill

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.