Created attachment 621309 [details] ffclonescript.html User Agent: Mozilla/5.0 (X11; Linux i686; rv:12.0) Gecko/20100101 Firefox/12.0 Build ID: 20120420145725 Steps to reproduce: I am writing a program which lets the user pick SVG files with the file API and them to the current page. To do that, I use DOMParser.parseFromString, then Document.importNode. The attach sample reproduces my use case with a circle and a click handler. Actual results: If the SVG file contains <script> elements, these scripts fail to work in the imported SVG elements. Under Firebug, I get 'click is not defined' Expected results: The <script> elements should continue to work. I also tested with Chromium and Opera and they seem no to have this problem.
It's not importNode that does this. It's inserting a <script> element into a document, which is done by parseFromString in your case. Henri, what's the right behavior here per spec?
(In reply to Boris Zbarsky (:bz) from comment #1) > Henri, what's the right behavior here per spec? The right behavior is that parseFromString marks scripts unexecutable. The spec even has a note that recounts the consequences of the normative text: "script elements get marked unexecutable and the contents of noscript get parsed as markup." The right behavior for cloning is to clone the "already started" flag: http://www.whatwg.org/specs/web-apps/current-work/#already-started Firefox is working as designed and per spec. Resolving INVALID in the sense that the allegation of actual behavior is true but the actual behavior is not a defect. I guess bugs need to be filed on WebKit and Opera if they revive scripts in this case.
For what it's worth, it looks like Opera and WebKit fail this for HTML <script> too. And for innerHTML as well, not just for DOMParser. I filed https://bugs.webkit.org/show_bug.cgi?id=86376 Filing a bug on Opera is too painful; I'm not going to worry about it.
I copied your bug report in the Opera bug tracking system. I am not an Opera developer or partner, so I used their "wizard", which generated a bug ID (DSK-363698) in an automated email reply. Hopefully someone will look at it but since the bug tracking system is not public, I as an Opera outsider cannot follow the progress. Thanks anyway for the thorough and prompt answer (as usual)
I poked jgraham about the Opera bug, and he said he could get it fixed.
I tested IE9/10 and it fails the test too. I have not filed the bug with MS though, I do not have the required Windows Live ID (whatever that is)
I downloaded the latest Windows8 preview and did more tests (my first tests were done on IE9 and an early IE10 platform preview). Good news: it turns out MS has fixed the problem and IE10 now executes Boris's testcase correctly.