Bug 32445 – REGRESSION: http://www.mcafee.com/japan/ displays unstyled

• Component: CSS
• Status: RESOLVED FIXED

[Alexey Proskuryakov]

http://www.mcafee.com/japan/ stylesheets are served with two Content-Type headers, which confuses WebKit:

$ curl -I http://www.mcafee.com/japan/css2009/masthead.css
HTTP/1.1 200 OK
Server: Microsoft-IIS/5.0
Content-Type: text/html; charset=Shift_jis
X-Powered-By: ASP.NET
Date: Mon, 20 Jul 2009 00:54:00 GMT
Content-Type: text/css
Accept-Ranges: bytes
Last-Modified: Tue, 03 Mar 2009 05:54:10 GMT
ETag: "1bb93179c49bc91:e79"
Content-Length: 1079

WebKit sees Content-Type as "text/html; charset=Shift_jis, text/css", which is not a proper CSS MIME type, so the stylesheet gets ignored in strict mode.

Handling of HTTP headers is performed by closed source Apple libraries, but we can work around the issue for now.

[Alexey Proskuryakov]

Created attachment 44706 [details]
proposed fix

[Alexey Proskuryakov]

<rdar://problem/7068303>

[WebKit Review Bot]

style-queue ran check-webkit-style on attachment 44706 [details] without any errors.

[David Levin]

The brace on line 274 in WebCore/html/HTMLLinkElement.cpp should be inside the ifdef.

[Darin Adler]

Comment on attachment 44706 [details]
proposed fix

> +    bool needsSiteSpecificQuirks = parent && parent->doc() && parent->doc()->settings() && parent->doc()->settings()->needsSiteSpecificQuirks();

The old code had the nice property that it would evaluate this only when strict mode was on.

> +    bool needsSiteSpecificQuirks = document()->page() && document()->page()->settings()->needsSiteSpecificQuirks();

Same here.

Could we structure things so that is still true?

r=me

[Darin Adler]

Comment on attachment 44706 [details]
proposed fix

> +        if (url.contains("mcafee.com/japan/", false))

It's a little strange to do this as a substring check. But I can't think of anything better.

[Alexey Proskuryakov]

> The old code had the nice property that it would evaluate this only when strict
> mode was on.

I noticed that it was lost, but couldn't come up with a better idea. One constraint was that I wanted the new code to be easily removable once it's no longer needed. Also, I doubt that this is such hot code.

Not landing this right away, maybe a fresh look will help.


(In reply to comment #4)
> The brace on line 274 in WebCore/html/HTMLLinkElement.cpp should be inside the
> ifdef.

Oops, thanks!

[Alexey Proskuryakov]

Committed <http://trac.webkit.org/changeset/52032> without re-arranging the code. I added a comment about the reason we use such a strange URL check to both places where we do now.