Bug 32807 – FrameView::adjustSize does not size to true content height

• Component: Layout and Rendering
• Status: RESOLVED WONTFIX

[Aaron Boodman]

Currently, FrameView::adjustSize sets the content size to the max() of the content's height and the container height. This means that if the content does not fill the container, the content size will still be set to the contianer's height.

This makes it impossible to implement containers that shrinkwrap their contents height using APIs like FrameView::contentSize() and ChromeClient::contentsSizeChanged().

It is possible to get the content's true height using frame->document()->documentElement()->scrollHeight(), but there is no corresponding event for when this changes.

Instead of adding more API for when the scroll height changes, it seems to me that it would be better to just change ::adjustSize() to set the content size to the true height. Applications that want the old behavior will still be able to get it, since they can get the container height.

[Aaron Boodman]

Created attachment 45305 [details]
Incomplete patch -- for discussion only

This patch is missing changes to ChangeLogs and layout tests.

I don't have access to my WebKit development computer to create a proper patch until Tuesday, but I wanted to get feedback on this change before then.

[Adam Treat]

Why only for height?  Why not do the same for width as well?  I can't totally speak for the larger implications here, but I think scrollHeight does take into account overflow::hidden while what is currently there does not.  That is a behavior change that I don't think is wanted.  I guess dhyatt can speak more to this.

[Aaron Boodman]

Width already does the right thing for whatever reason.

[Dave Hyatt]

I'd look into frame flattening here.  I think it kind of does what you want (makes an iframe or frame size around its content).

[David Levin]

(In reply to comment #4)
> I'd look into frame flattening here.  I think it kind of does what you want (makes an iframe or frame size around its content).

I looked at frame flattening (and did an experiment to verify). Frame flattening doesn't quite work for the desired use case because frame flattening applies recursively to all enclosed frames.

[David Levin]

This has been handled in other ways in Chromium code.