593286 - White rectangle visible for a short time over acid3 test when reloading

Status: RESOLVED FIXED

(Core :: Web Painting, defect)

Description

[Kai de Leeuw]

User-Agent:       Mozilla/5.0 (Windows NT 5.1; rv:2.0b6pre) Gecko/20100902 Firefox/4.0b6pre
Build Identifier: Mozilla/5.0 (Windows NT 5.1; rv:2.0b6pre) Gecko/20100902 Firefox/4.0b6pre

Summary is enough.

Reproducible: Always

Steps to Reproduce:
1. Open http://acid3.acidtests.org/
2. When finished counting up to 97 press the reload button
3. Behold the white rectangle that is visible for a short time and then disappears
Actual Results:  
White rectangle over acid3 test.

Expected Results:  
No white rectangle over acid3 test.

Comment 1

[Kai de Leeuw]

The rectangle also shows when you navigate to another page.

Comment 2

[José Jeria]

Kai, do you have any regression date for this?

Comment 3

[Timothy Nikkel (:tnikkel)]

Probably caused by bug 130078.

Comment 4

[Timothy Nikkel (:tnikkel)]

Attachment: reduced testcase

Comment 5

[Timothy Nikkel (:tnikkel)]

There is a (transparent) iframe (well actually an object with data attribute that points to an html document) in the acid3 test, which is basically all you need to reproduce this. When we create the content viewer for the new document the mParent pointer on the docshell for the iframe gets nulled out (via the path nsDocShell::CreateContentViewer -> Embed -> SetupNewViewer -> DestroyChildren). But we still draw the old page for a while yet.

When we call nsContentUtils::IsChildOfSameType in PresShell::UpdateCanvasBackground (for an opaque bg color on the root content document) and nsSubDocumentFrame::BuildDisplayList (for a container layer on the root content document) it returns null because the parent pointer is null, not because it is the root content document.

So I think for drawing purposes, we need to look at the frames/views for this purpose, not the docshell.

Comment 6

[Timothy Nikkel (:tnikkel)]

Attachment: patch

Comment 7

[Timothy Nikkel (:tnikkel)]

Attachment: patch v2

We need to return a correct value even when we don't have a root frame. The only thing different in this patch is the definition of IsRootContentDocument.

Comment 8

[Robert O'Callahan (:roc) (email my personal email if necessary)]

Comment on attachment 472786 [details] [diff] [review]
patch v2

Prefer vm over VM

Comment 9

[Timothy Nikkel (:tnikkel)]

http://hg.mozilla.org/mozilla-central/rev/588f3801b693