253977 - More block bugs flushed out by columns work

Status: RESOLVED FIXED

(Core :: Layout, defect)

Description

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

Here are some bugs that I found that I have fixes for.

-- nsBlockFrame::Reflow should always call DrainOverflowLines no matter what the
type of reflow. It is never correct to leave a prev-in-flow with an overflow
line list.

-- I need a way to determine if incrementally reflowing a block frame did
something that requires its next in flow to be reflowed. According to
nsIFrame.h, NS_FRAME_REFLOW_NEXTINFLOW should be added to the outgoing reflow
status to indicate this. However, that isn't currently implemented. So I've
implemented it, at least for blocks.

-- nsBlockFrame::DoRemoveFrame did not handle the case where the frame to be
removed, or some of its continuations, are on the overflow line list. So I have
it search both the normal and overflow lines. I also simplified it a bit by
removing the outer while loop and turning it into a recursive tail call.

I've also included a little bit of frame dump code that ought to migrate from my
tree to the trunk.

Comment 1

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

Attachment: fixes

As described.

Comment 2

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

Attachment: better fix

That last patch did not correctly reset prevSibling when we start deleting
continuations in the overflow lines.

Comment 3

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

Checked in.

Comment 4

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

Backed out because it caused ~10ms Tp regression on btek. Maybe it's noise but I
have to go to bed.

It's probably the extra calls to GetOverflowLines on every block reflow, and
possible in DoRemoveFrame (although in the latter case we should always find the
frame to delete before we try the overflow lines, since Tp doesn't test blocks
with next-in-flows, I think). I should probably try adding state bits to
indicate the presence of overflow lines, overflow out-of-flows, and overflow
placeholders, it would be easy. But not right now...

Comment 5

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

Relanded. Let's see how btek behaves this time.

Comment 6

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

btek looks happy. Yay!

Comment 7

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

Looks like this did in fact regress Tp by maybe 5ms.