QWebFrame::render() clips content right of 640 pixels wide when rendering to a QImage

• Component: WebKit Qt
• Status: RESOLVED INVALID

[Will Stokes]

If you try to render content in a QWebFrame to an image that is wider than 640 pixels, the content in the QWebFrame gets clipped right of the 640 pixel column. Attached I've provided a small test case that produces the problem.

[Will Stokes]

Created attachment 54152 [details]
screenshot of bug

[Will Stokes]

Created attachment 54153 [details]
example program that produces bug (.cpp file)

[Will Stokes]

Created attachment 54154 [details]
example program that produces bug (.h file)

[Sriram]

Seems like a bad bug for use cases where Webkit is embedded.

[Robert Hogan]

(In reply to comment #4)
> Seems like a bad bug for use cases where Webkit is embedded.

QWebFrame::render() clips to the frameRect() of the frame. In the absence of content to update the size of the rect you will need to update it first with the expected size of the image. So in your example you need to do:

  webView.page()->setViewportSize(QSize(maxWidth, 480));

QWebFrame::render() could remove the need for this step and check the viewport() of the painter and adjust the frame->view() as follows:

        WebCore::FrameView* view = frame->view();
        view->setFrameRect(QRect(QPoint(0, 0), painter.viewport()));
        view->adjustViewSize();

Simon/Kenneth, what do you think?

[Kenneth Rohde Christiansen]

(In reply to comment #5)
> (In reply to comment #4)
> > Seems like a bad bug for use cases where Webkit is embedded.
> 
> QWebFrame::render() clips to the frameRect() of the frame. In the absence of content to update the size of the rect you will need to update it first with the expected size of the image. So in your example you need to do:
> 
>   webView.page()->setViewportSize(QSize(maxWidth, 480));
> 
> QWebFrame::render() could remove the need for this step and check the viewport() of the painter and adjust the frame->view() as follows:
> 
>         WebCore::FrameView* view = frame->view();
>         view->setFrameRect(QRect(QPoint(0, 0), painter.viewport()));
>         view->adjustViewSize();
> 
> Simon/Kenneth, what do you think?

you might be getting the QWebElement from an existing view, so in that case you do not want to change the viewport, unless you would somewhat clone the element.

[Robert Hogan]

(In reply to comment #6)
> 
> you might be getting the QWebElement from an existing view, so in that case you do not want to change the viewport, unless you would somewhat clone the element.

So do you think the bug is invalid, or just hard to fix? ;-)

[Antonio Gomes]

(In reply to comment #7)
> (In reply to comment #6)
> > 
> > you might be getting the QWebElement from an existing view, so in that case you do not want to change the viewport, unless you would somewhat clone the element.
> 
> So do you think the bug is invalid, or just hard to fix? ;-)

invalid to me.

[Kenneth Rohde Christiansen]

(In reply to comment #8)
> (In reply to comment #7)
> > (In reply to comment #6)
> > > 
> > > you might be getting the QWebElement from an existing view, so in that case you do not want to change the viewport, unless you would somewhat clone the element.
> > 
> > So do you think the bug is invalid, or just hard to fix? ;-)
> 
> invalid to me.

I'm not sure, he is called setPreferredContentSize(), so I'm not sure it should be clipped to the frameRect, especially not if the fixedLayoutSize is larger.

Adam Treat, can you give some input here?

[Viatcheslav Ostapenko]

(In reply to comment #8)
> (In reply to comment #7)
> > (In reply to comment #6)
> > > 
> > > you might be getting the QWebElement from an existing view, so in that case you do not want to change the viewport, unless you would somewhat clone the element.
> > 
> > So do you think the bug is invalid, or just hard to fix? ;-)
> 
> invalid to me.

No, I don't think it's invalid.
Generally, there is now no _right_ way to render contents without clipping to viewport. Suggested QWebElement::render is not very usable in this case, because it doesn't have any clipping paramter and will traverse the whole webkit render tree if only part of the page (different from viewport) been rendered.

I think QWebFramePrivate::renderRelativeCoords unnecessarily clips rendered content to viewport (framerect). It used in 3 places (variants of QWebFrame::render method) and those places all default to framerect, if clipping was not specified or empty. IMHO, if clipping _was_ specified, than it should be used as it is (well, may be clipped to contents rect). 

So far clipping in QWebFramePrivate::renderRelativeCoords contradicts to the code written in QWebFrame::render(QPainter* painter, RenderLayer layer, const QRegion& clip) .

[Benjamin Poulain]

As said in comments, invalid bug.

[Viatcheslav Ostapenko]

(In reply to comment #11)
> As said in comments, invalid bug.

Ok!
1. Here is use case:
I need to create full page snapshot, but because of limited RAM I have to create it in chunks.
2. qtwebkit doc doesn't say, that QWebFrame::render clips to viewport, so current behavior doesn't match documentation.

[Benjamin Poulain]

(In reply to comment #12)

What you suggest is basically extending the API of QWebElement::render(). If you plan to do that, you should create a new bug report.

If you want to change the API of QWebFrame::render(), this is behavior breakage of public APIs, and you need to start a thread on the mailing list about that.

In both cases, this bug report is invalid. The bug is "QWebFrame::render() clips content right of 640 pixels wide when rendering to a QImage" which is simply how the API is defined.

[Antonio Gomes]

Maybe patching the docs?

[Viatcheslav Ostapenko]

(In reply to comment #13)
> (In reply to comment #12)
> 
> If you want to change the API of QWebFrame::render(), this is behavior breakage of public APIs, and you need to start a thread on the mailing list about that.

Ok.
Return QWebFrame::paintContents ?

(In reply to comment #14)
> Maybe patching the docs?

Yes.

Also, renderRelativeCoords clips every rect from the clip region in a cycle. I think it should clip the clip region outside of cycle - this will reduce possible number of regions and eliminate situation, when view->paintContents is called with an empty rect (it happens rarely).

[Viatcheslav Ostapenko]

"reduce possible number of regions" = "reduce possible number of rects"