Bug 13444 – REGRESSION: "No parent frame" error messages during layout tests

• Component: Page Loading
• Status: RESOLVED FIXED

[Alexey Proskuryakov]

I haven't investigated these.

fast/dom/Window ..........ERROR: No parent frame in transitionToCommitted:, FrameLoadTypeInternal
(/Users/ap/WebKit/WebCore/loader/FrameLoader.cpp:4054 updateHistoryForInternalLoad)
.....ERROR: No parent frame in transitionToCommitted:, FrameLoadTypeInternal
(/Users/ap/WebKit/WebCore/loader/FrameLoader.cpp:4054 updateHistoryForInternalLoad)
.ERROR: No parent frame in transitionToCommitted:, FrameLoadTypeInternal
(/Users/ap/WebKit/WebCore/loader/FrameLoader.cpp:4054 updateHistoryForInternalLoad)
..ERROR: No parent frame in transitionToCommitted:, FrameLoadTypeInternal
(/Users/ap/WebKit/WebCore/loader/FrameLoader.cpp:4054 updateHistoryForInternalLoad)
.....

[Darin Adler]

<rdar://problem/5153031>

[Alexey Proskuryakov]

*** Bug 13330 has been marked as a duplicate of this bug. ***

[Maxime BRITTO]

The layout test affected by theses logs is setting-properties-on-closed-window.html which :
    1 -  creates a variable (w) and give it a new page : w = open('about:blank');
    2 -  closes this page : w.close();
    3 -  try to access the variable to test if it crashes : w.status = i;

At step 1 : the webcore handle the loading as a FrameLoadTypeStandard an there is no problem.
At step 3 : the webcore loads the formely closed page as a FrameLoadTypeInternal an when it's trying to update history for internal loads the error is logged.

I don't know why it is forbidden to load an empty page internaly. There is a comment in the code (just before the log expression) which is mentionning the bug 3556159 (rdar?) as a justification.

I tried to replace the about:blank by http://www.google.fr in the setting-properties-on-closed-window.html and it works fine.

[Maxime BRITTO]

(In reply to comment #3)
> I don't know why it is forbidden to load an empty page internaly. There is a
> comment in the code (just before the log expression) which is mentionning the
> bug 3556159 (rdar?) as a justification.
> 
> I tried to replace the about:blank by http://www.google.fr in the
> setting-properties-on-closed-window.html and it works fine.
> 
I came back on this and I discovered that the frame loaded Internaly with no parent is not necessarily empty. 
The comment in the code was relative to frame based web pages as sub-frame loading was the only user of FrameLoadTypeInternal. But JS uses it when it reloads a page to access a variable (like in the test)

There is still one problem : this rdar bug number in the code comments.

Fixed in <http://trac.webkit.org/projects/webkit/changeset/24299>.