[IE7] Manual test for #10165 fails when editor built from preset

[Olek Nowodziński]

Test for #10165 is: ​http://ckeditor4.t/mt/4.1.2/10165.html

If the editor is based on the full preset (./build.sh full all) with or without --leave-js-unminified, IE7 raises Permission Denied errors when accessing link dialog or clipboard paste dialog.

It passes on dev code though.

[Frederico Caldeira Knabben]

I've just played with the IE console in a test page with custom domain. You gonna love this:

>> CKEDITOR.document.getWindow().$ == window 
true 

>> CKEDITOR.document.getWindow().$.document 
"Access is denied."

>> window.document 
[object] { … }

It turned out that we save a reference to window inside CKEDITOR.document, which is actually a object reference different that "window" at a later stage. In fact, we have this as well:

>> CKEDITOR.document.getWindow().$ === window 
false

So basically, in IE, an old reference to "window" may be different than "window".

Welcome to the world of IE!

[Frederico Caldeira Knabben]

I've just pushed a fix for this in t/10505. I'm simply avoiding caching "window", as it is mutable. I doubt we would have performance issues because of this.

[Piotrek Koszuliński]

You gonna love this

Yes, I am :D

[Olek Nowodziński]

Replying to fredck:

I've just played with the IE console in a test page with custom domain. You gonna love this:

>> CKEDITOR.document.getWindow().$ == window 
true 

>> CKEDITOR.document.getWindow().$.document 
"Access is denied."

>> window.document 
[object] { … }

It turned out that we save a reference to window inside CKEDITOR.document, which is actually a object reference different that "window" at a later stage. In fact, we have this as well:

>> CKEDITOR.document.getWindow().$ === window 
false

So basically, in IE, an old reference to "window" may be different than "window".

Welcome to the world of IE!

I assume that the error is because there are few files to be loaded in the package and hence things go too fast?

[Olek Nowodziński]

Tests don't confirm the expected performance loss. This fix is sufficient.

[Olek Nowodziński]

Merged fix into master, dev (​​git:8f4b94b).