Bug 17847 – CSS caching corruption

• Status: RESOLVED
• Resolution: INVALID
• Component: CSS

[Derk-Jan Hartman]

This is a very complicated bug. I encountered this issue when using the javascript tool twinkle on Wikipedia. The tool gives you extra tabs, that open tools in a CSS window. After finishing the "job", the tool redirects you to the original page you visited.

Website of the tool:
http://en.wikipedia.org/wiki/Wikipedia:WikiProject_User_scripts/Scripts/Twinkle

In Safari 3 (and the NB), I was experiencing a problem where after the redirect, my page was only showing the background image. 
Screenshot: http://deneb.student.utwente.nl/~hartman/weird%20issue.png

It appeared as if all my CSS styling was down the drain somehow. If I opened other Wikipedia pages on other tabs or windows the problem persisted. I had to force a reload to make the issue disappear. I believe it was reproducible with this webarchive: http://deneb.student.utwente.nl/~hartman/wikipedia/saferversion.webarchive

I have now looked further into this issue. It was very hard to reduce what was causing this. However I think i have now traced the issue.
Making the following changes in the javascript tool fixed the problem for me:
http://en.wikipedia.org/w/index.php?title=User%3ATheDJ%2Fmorebits.js&diff=198208031&oldid=197808959

Basically I changed

var cssNode = document.createElement('style');
cssNode.type = 'text/css';
cssNode.rel = 'stylesheet';
cssNode.appendChild( document.createTextNode("")); // Safari bugfix 
document.getElementsByTagName("head")[0].appendChild(cssNode);
var styles = cssNode.sheet ? cssNode.sheet : cssNode.stylesSheet;
styles.insertRule("my line of CSS", 0);
styles.insertRule("another line of CSS", 0);

to 

var cssNode = document.createElement('style');
cssNode.type = 'text/css';
cssNode.rel = 'stylesheet';
cssNode.appendChild( document.createTextNode("myCSS" ));
document.getElementsByTagName("head")[0].appendChild(cssNode);


So it seems that DOM editing and then using insertRule() breaks something that, as far as I know it should not break. As you may note in the original code there is a reference to a Safari bugfix. This would be bug #14462. It might be related in some way, but I'm not sure.

[Mark Rowe (bdash)]

*** Bug 17848 has been marked as a duplicate of this bug. ***

[Shane Stephens]

Is this still happening in recent versions of WebKit? Any chance of a reduction?