[Webkit-unassigned] [Bug 10313] xsl:import and document() don't work in stylesheets loaded via XMLHttpRequest

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=10313





--- Comment #59 from Andreas Wictor <andreas.wictor at xcerion.com>  2010-06-28 07:40:25 PST ---
(In reply to comment #51)
> I really dislike it that both Document and DOMImplementation get an m_originDocument reference. It doesn't seem that these should ever be different, but it raises all sorts of unpleasant questions about when exactly they are different in practice.

I'm not sure I understand the problem you are describing here.

> I'm going to say r- for the above comments, but there is also something I don't understand in how subresources are loaded in general. There is that old ugly code in XSLStyleSheet::loadChildSheets() that iterates over stylesheets to find imports and includes - it it obsoleted by your code? How do they interact?
> 
> I'll try to figure this out myself later, but an explanation from you would be appreciated, too.

No, my code does not obsolete any existing code in XSLStyleSheet. My understanding of this code is that when we find imports and includes in the loadChildSheets function we create XSLImportRule objects to represent them. These XSLImportRule objects are saved to a list by the append function inherited from StyleList, and then loaded by a call to loadSheet. Later when libxslt needs to get these imported stylesheets we call locateStylesheetSubResource which will iterate through the list of saved XSLImportRule objects and try to find the correct one. My contribution to this process is that I made loads synchronous for transformations not origination from a pi, because otherwise locateStylesheetSubResource will find the correct XSLImportRule object but it will not be fully loaded.

> 
> +    void finishLoadingSheetAsynchronously(DocLoader*, const KURL&);
> +    void finishLoadingSheetSynchronously(DocLoader*, const KURL&);
> 
> A synchronous load can be started while an asynchronous one is in progress. But we're using global variables for loading, see globalProcessor and globalDocLoader. Is there a race condition here?

I have found a way to remove these global variables by using the _private field that exists on most libxml structs. My original plan was to submit a patch after my work on this bug was done, but maybe it should be included here?

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