1299087 - Allow onbeforeunload dialog after interacting with flash

Status: RESOLVED INCOMPLETE

(Core :: DOM: Core & HTML, defect, P5)

Description

[Dennis Hannwacker]

User Agent: Mozilla/5.0 (Windows NT 10.0; WOW64; rv:47.0) Gecko/20100101 Firefox/47.0
Build ID: 20160623154057

Steps to reproduce:

Listening for onbeforeunload event to show dialog to check back with user, if he really wants to leave site.

Such showing of a dialog before leaving the site was restricted in Bug 636905 to be only allowed after user interaced with the page (mouse-/keyboard-events in the browser window).

The problem is that interactions with a flash application are not registered towards this check and do not count as such interactions. So in case of a fullscreen flash application clicks on ui-elements (buttons, lists, trees) inside this flash application are not recognized as interactions with the site in terms of Bug 636905.

Only mouse-wheel in/over the flash-application seems to trigger the setting of the introduced flag "userHasInteracted".


Actual results:

After user opened site with flash application, then interacted with it and then closed browser tab, the dialog opened from within the onbeforeunload-handler was not shown.


Expected results:

After user opened site with flash application, then interacted with it and then closed browser tab, the dialog opened from within the onbeforeunload-handler should be shown.

Comment 1

[Hsin-Yi Tsai (she/her) [:hsinyi]]

Hi Gijs, what do you think? Thanks!

Comment 2

[:Gijs (he/him)]

(In reply to Hsin-Yi Tsai [:hsinyi] from comment #1)
> Hi Gijs, what do you think? Thanks!

It's not clear from the question "what do you think?" what you expect from me. If it helps: I don't have time to investigate this further. I'm not part of the DOM team so I don't know how it should be prioritized otherwise. In future, please ask a more specific question.

I mean, this might be a real bug. I don't know much about flash. Does this happen to just any flash app, or only this particular one? A testcase would really help! Does this also happen when wmode=transparent? Etc.

In case you're asking me purely because I wrote the patch in 636905: I don't think it's severe enough to back out bug 636905 over. The direction our integration of flash and beforeunload has gone in mean that I think it doesn't need to be prioritized over other things I'm (and/or other people are) working on. I have no idea if that answers your question or not.

Comment 3

[Dennis Hannwacker]

> Does this also happen when wmode=transparent?
That is an excellent point. I made some tests and the problem can only be observed with wmode=direct.
So wmode=transparent and wmode=opaque seem to work fine.

Comment 4

[Dennis Hannwacker]

Attachment: Test.html

Html file for test case. Uses following test.swf and test.js

Comment 5

[Dennis Hannwacker]

Attachment: Test.swf

Swf (flash-file) for test case. Used by test.html

Comment 6

[Dennis Hannwacker]

Attachment: test.js

Javascript file (registering onberforeclose handler) for test case. Used by test.html.

Comment 7

[Hsin-Yi Tsai (she/her) [:hsinyi]]

(In reply to :Gijs Kruitbosch from comment #2)
> (In reply to Hsin-Yi Tsai [:hsinyi] from comment #1)
> > Hi Gijs, what do you think? Thanks!
> 
> It's not clear from the question "what do you think?" what you expect from
> me. If it helps: I don't have time to investigate this further. I'm not part
> of the DOM team so I don't know how it should be prioritized otherwise. In
> future, please ask a more specific question.
> 

My bad, I should have asked a more specific question. Thanks for the reply. :)

> I mean, this might be a real bug. 
> I don't know much about flash. Does this happen to just any flash app, or only this particular one?
> A testcase would really help! Does this also happen when wmode=transparent? Etc.

They are great points.

> In case you're asking me purely because I wrote the patch in 636905: I don't
> think it's severe enough to back out bug 636905 over. 

I see, thanks.

> The direction our integration of flash and beforeunload has gone in mean that I think it
> doesn't need to be prioritized over other things I'm (and/or other people
> are) working on. I have no idea if that answers your question or not.

Yes, you answered my question well. Thank you so much for the comment. :)

Comment 8

[Andrew Overholt [:overholt]]

Benjamin, do you think this needs to be prioritized above P5?

Comment 9

[Benjamin Smedberg]

Nope. It might already work with windowless Flash (wmode=opaque), which would be a fine workaround and will soon be the default. But Flash isn't the web, and I'm fine with fullpage Flash apps not being able to use beforeunload.

Comment 10

[Dennis Hannwacker]

(In reply to Benjamin Smedberg [:bsmedberg] from comment #9)
> Nope. It might already work with windowless Flash (wmode=opaque), which
> would be a fine workaround and will soon be the default.
The problem is for any application/site using Stage 3D (like for example 3D-Games or other embedded 3D-Content), wmode "opaque" is not an option. One has to use mode "direct" to have access to Stage 3D (Context3D). And in wmode "direct" the described problem can be observed.

Comment 11

[u597032]

Flash is no longer supported