Attachment #141826 for bug #86315

View Details Raw Unified Return to Bug 86315



2012-05-14  Eric Seidel  <eric@webkit.org>

        Styles are not recalculated when the seamless attribute is dynamically added/removed
        https://bugs.webkit.org/show_bug.cgi?id=86315

        Reviewed by NOBODY (OOPS!).

        Covered by fast/frames/seamless/seamless-css-cascade.html.

        * html/HTMLIFrameElement.cpp:
        (WebCore::HTMLIFrameElement::isPresentationAttribute):
         - Make seamless a presentational attribute, which means style on the <iframe> will
           be forced to recalculate when it changes.  This is correct, but not observable
           until the layout changes are landed (as then the iframe should correctly revert to not
           being sized to fit its content if seamless is removed).
        (WebCore::HTMLIFrameElement::parseAttribute):
         - When the seamless attribute is added or remove, force the content document to recalc
           its style resolver, which will refresh the list of inherited stylesheets from the
           parent.  This doesn't need to happen synchronously.  When the layout changes land
           the content document will actually cause that recalc to redirect to the parent document
           in the seamless case anyway, but it's more correct to ask the content document directly.

2012-05-14  Takashi Sakamoto  <tasak@google.com>

        Crash in WebCore::RenderObject::repaint

Lines 51-57 PassRefPtr<HTMLIFrameElement> HTMLIFrameElement::create(const QualifiedName& tag a/Source/WebCore/html/HTMLIFrameElement.cpp_sec1
- a/Source/WebCore/html/HTMLIFrameElement.cpp -2 / +6 lines
51		51
52	bool HTMLIFrameElement::isPresentationAttribute(const QualifiedName& name) const	52	bool HTMLIFrameElement::isPresentationAttribute(const QualifiedName& name) const
53	{	53	{
54	if (name == widthAttr \|\| name == heightAttr \|\| name == alignAttr \|\| name == frameborderAttr)	54	if (name == widthAttr \|\| name == heightAttr \|\| name == alignAttr \|\| name == frameborderAttr \|\| name == seamlessAttr)
55	return true;	55	return true;
56	return HTMLFrameElementBase::isPresentationAttribute(name);	56	return HTMLFrameElementBase::isPresentationAttribute(name);
57	}	57	}
Lines 87-93 void HTMLIFrameElement::parseAttribute(Attribute* attr) a/Source/WebCore/html/HTMLIFrameElement.cpp_sec2
87	m_name = newName;	87	m_name = newName;
88	} else if (attr->name() == sandboxAttr)	88	} else if (attr->name() == sandboxAttr)
89	setSandboxFlags(attr->isNull() ? SandboxNone : SecurityContext::parseSandboxPolicy(attr->value()));	89	setSandboxFlags(attr->isNull() ? SandboxNone : SecurityContext::parseSandboxPolicy(attr->value()));
90	else	90	else if (attr->name() == seamlessAttr) {
		91	// If we're adding or removing the seamless attribute, we need to force the content document to recalculate its StyleResolver.
		92	if (contentDocument())
		93	contentDocument()->styleResolverChanged(DeferRecalcStyle);
		94	} else
91	HTMLFrameElementBase::parseAttribute(attr);	95	HTMLFrameElementBase::parseAttribute(attr);
92	}	96	}
93		97



2012-05-14  Eric Seidel  <eric@webkit.org>

        Styles are not recalculated when the seamless attribute is dynamically added/removed
        https://bugs.webkit.org/show_bug.cgi?id=86315

        Reviewed by NOBODY (OOPS!).

        Add a subtest to cover this case.

        * fast/frames/seamless/seamless-css-cascade-expected.txt:
        * fast/frames/seamless/seamless-css-cascade.html:

2012-05-14  Takashi Sakamoto  <tasak@google.com>

        Crash in WebCore::RenderObject::repaint

Lines 5-8 PASS window.getComputedStyle(three).color is "rgb(255, 255, 255)" a/LayoutTests/fast/frames/seamless/seamless-css-cascade-expected.txt_sec1
- a/LayoutTests/fast/frames/seamless/seamless-css-cascade-expected.txt +3 lines
5	PASS window.getComputedStyle(rootElement).color is "rgb(255, 165, 0)"	5	PASS window.getComputedStyle(rootElement).color is "rgb(255, 165, 0)"
6	PASS window.getComputedStyle(rootElement).color is "rgb(1, 2, 3)"	6	PASS window.getComputedStyle(rootElement).color is "rgb(1, 2, 3)"
7	PASS window.getComputedStyle(one).color is "rgb(3, 2, 1)"	7	PASS window.getComputedStyle(one).color is "rgb(3, 2, 1)"
		8	PASS window.getComputedStyle(one).color is "rgb(0, 0, 0)"
		9	PASS window.getComputedStyle(two).color is "rgb(128, 0, 128)"
		10	PASS window.getComputedStyle(three).color is "rgb(0, 0, 0)"
8		11

Lines 42-46 window.onload = function () { a/LayoutTests/fast/frames/seamless/seamless-css-cascade.html_sec1
- a/LayoutTests/fast/frames/seamless/seamless-css-cascade.html +6 lines
42	document.head.appendChild(styleSheet);	42	document.head.appendChild(styleSheet);
43	// #one's style is only specified by this parent, so adding a later sheet should override the color and update the child frame.	43	// #one's style is only specified by this parent, so adding a later sheet should override the color and update the child frame.
44	shouldBeEqualToString("window.getComputedStyle(one).color", "rgb(3, 2, 1)");	44	shouldBeEqualToString("window.getComputedStyle(one).color", "rgb(3, 2, 1)");
		45
		46	// Test that removing the seamless attribute recalculates the child's style.
		47	window.iframe.removeAttribute("seamless");
		48	shouldBeEqualToString("window.getComputedStyle(one).color", "rgb(0, 0, 0)"); // black, default.
		49	shouldBeEqualToString("window.getComputedStyle(two).color", "rgb(128, 0, 128)"); // purple, selector in child.
		50	shouldBeEqualToString("window.getComputedStyle(three).color", "rgb(0, 0, 0)"); // black, default.
45	}	51	}
46	</script>	52	</script>

Return to Bug 86315

Lines 1-3 a/Source/WebCore/ChangeLog_sec1
- a/Source/WebCore/ChangeLog +22 lines
		1	2012-05-14 Eric Seidel <eric@webkit.org>
		2
		3	Styles are not recalculated when the seamless attribute is dynamically added/removed
		4	https://bugs.webkit.org/show_bug.cgi?id=86315
		5
		6	Reviewed by NOBODY (OOPS!).
		7
		8	Covered by fast/frames/seamless/seamless-css-cascade.html.
		9
		10	* html/HTMLIFrameElement.cpp:
		11	(WebCore::HTMLIFrameElement::isPresentationAttribute):
		12	- Make seamless a presentational attribute, which means style on the <iframe> will
		13	be forced to recalculate when it changes. This is correct, but not observable
		14	until the layout changes are landed (as then the iframe should correctly revert to not
		15	being sized to fit its content if seamless is removed).
		16	(WebCore::HTMLIFrameElement::parseAttribute):
		17	- When the seamless attribute is added or remove, force the content document to recalc
		18	its style resolver, which will refresh the list of inherited stylesheets from the
		19	parent. This doesn't need to happen synchronously. When the layout changes land
		20	the content document will actually cause that recalc to redirect to the parent document
		21	in the seamless case anyway, but it's more correct to ask the content document directly.
		22
1	2012-05-14 Takashi Sakamoto <tasak@google.com>	23	2012-05-14 Takashi Sakamoto <tasak@google.com>
2		24
3	Crash in WebCore::RenderObject::repaint	25	Crash in WebCore::RenderObject::repaint

Lines 1-3 a/LayoutTests/ChangeLog_sec1
- a/LayoutTests/ChangeLog +12 lines
		1	2012-05-14 Eric Seidel <eric@webkit.org>
		2
		3	Styles are not recalculated when the seamless attribute is dynamically added/removed
		4	https://bugs.webkit.org/show_bug.cgi?id=86315
		5
		6	Reviewed by NOBODY (OOPS!).
		7
		8	Add a subtest to cover this case.
		9
		10	* fast/frames/seamless/seamless-css-cascade-expected.txt:
		11	* fast/frames/seamless/seamless-css-cascade.html:
		12
1	2012-05-14 Takashi Sakamoto <tasak@google.com>	13	2012-05-14 Takashi Sakamoto <tasak@google.com>
2		14
3	Crash in WebCore::RenderObject::repaint	15	Crash in WebCore::RenderObject::repaint