Bug 66219 – CSS style code incorrectly assumes WebCore::Document::documentElement() does not return NULL

• Component: CSS
• Status: NEW

[Berend-Jan Wever]

Created attachment 103903 [details]
Repro (unzip and load repro.html)

WebCore::Document::documentElement() can return NULL. There are a few places in the code where this is not taken into account.
http://codesearch.google.com/codesearch#search/&q=documentElement%5C(%5C)%5C-%5C%3ErenderStyle&exact_package=chromium&type=cs

Repro.html:
<iframe src="repro.svg" onload="go(this)"></iframe>
<script>
  function go(oIframe) {
    setTimeout(function() {
      document.adoptNode(oIframe.contentDocument.documentElement);
    }, 1);
  }
</script>

Repro.svg:
<?xml version="1.0" standalone="no"?>
<?xml-stylesheet href="repro.css" ?>

Repro.css:
@media (width:1) {
  @page  {}
}

(I bet there is a simpler way to trigger this).

id:             webkit.dll!WebCore::Node::renderStyle ReadAV@NULL (c335e7a3c5b21e67401e64aac7846349)
description:    Attempt to read from unallocated NULL pointer+0x28 in webkit.dll!WebCore::Node::renderStyle
application:    Chromium 14.0.828.0
stack:          webkit.dll!WebCore::Node::renderStyle
                webkit.dll!WebCore::widthMediaFeatureEval
                webkit.dll!WebCore::min_widthMediaFeatureEval
                webkit.dll!WebCore::MediaQueryEvaluator::eval
                webkit.dll!WebCore::CSSStyleSelector::affectedByViewportChange
                webkit.dll!WebCore::FrameView::layout
                webkit.dll!WebCore::FrameView::updateLayoutAndStyleIfNeededRecursive
                webkit.dll!WebKit::WebFrameImpl::layout
                webkit.dll!WebKit::WebViewImpl::layout
                chrome.dll!RenderWidget::DoDeferredUpdate
                chrome.dll!RenderWidget::DoDeferredUpdateAndSendInputAck
                chrome.dll!RenderWidget::InvalidationCallback
                ...

@hyatt: in r45919 you added code to handle "rem" css units, which I think introduced this particular instance of this issue. The w3c docs do not specify how to handle "rem" units when there is no document element, so I don't know how to fix this. If you have a fix, could you see if the other places in the CSS code where the same problem seems to exist might be fixed in the same way?

[Berend-Jan Wever]

Chromium: https://code.google.com/p/chromium/issues/detail?id=92880

[Berend-Jan Wever]

*** Bug 67474 has been marked as a duplicate of this bug. ***

[Luke Macpherson]

Created attachment 112137 [details]
Patch

[Luke Macpherson]

Patch added fixes the crash by treating rems as ems when the root style is unknown.
Patch does not add a layout test.

[Darin Adler]

Comment on attachment 112137 [details]
Patch

View in context: https://bugs.webkit.org/attachment.cgi?id=112137&action=review

> Source/WebCore/ChangeLog:8
> +        No new tests. (OOPS!)

Why no tests?

[Luke Macpherson]

Still working on it.