[Webkit-unassigned] [Bug 116689] Regression(r145959) Convert <select> to new flex box causes strange positioning problems

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=116689





--- Comment #9 from Roger Fong <roger_fong at apple.com>  2013-05-29 13:57:37 PST ---

> > 2. Why doesn't the RenderFlexibleBox code update the layer's scroll info? What about it makes it different from RenderDeprecatedFlexibleBox in that regard? Or is it actually a typo?
> 
> !! I'm becoming more convinced that RenderFlexibleBox just needs to update the RenderLayer's scroll info like RenderDeprecatedFlexibleBox does.

Simon pointed out that the code to do this is there, just not getting called.
It's not getting called because of some delayUpdateScrollInfo stuff.
A few more of the dots are starting to connect now.

Upon entering RenderFlexibleBox::layoutBlock, gDelayUpdateScrollInfo is already 1 from a previous call to RenderDeprecatedFlexibleBox::layoutHorizontalBox (this is the select element's containing parent)

It remains at 1 by the time we leave call updatescrollinfoafterlayout, and thus we ignore updating the render layer's scroll info.

We then eventually call clearLayoutOverflow(), (still while gDelayUpdateScrollInfo is 1) and then when we finally return to RenderDeprecatedFlexibleBox::layoutHorizontalBox and call finishDelayUpdateScrollInfo, we set gDelayUpdateScrollInfo back down to 0 and perform all the delayed scroll updates, including on the layer that was ignored earlier.
However, since clearLayoutOverflow has already been called at this point and layoutOverFlowRect has been 0'ed out, the scroll origin gets improperly set.

I will look into these delayscrollinfo changes but perhaps what needs to happen is that the clearLayoutOverflow must also be delayed until finishDelayUpdateScrollInfo, that way the scollinfo will be updated properly.

This is fun.

-- 
Configure bugmail: https://bugs.webkit.org/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the assignee for the bug.