266971 - {ib} splits almost screw up float containing blocks (BuildFloatList removal)

Status: RESOLVED FIXED

(Core :: Layout, defect)

Description

[Boris Zbarsky [:bzbarsky]]

Consider the following situation:

<inline>
  Text
  <block />
  More test <float />
  <block />
  Text
</inline>

When we construct frames we put the float in the float list of the nearest block
ancestor of the inline....  Then we go through AdjustOutOfFlowPtrs and change
the relevant parent pointers to point to the anonymous block that the {ib} split
created.  Finally, the BuildFloatList() for that anonymous block pulls the
floats into the right framelist too.

So the end result is OK, but it rather depends on BuildFloatList()...  Perhaps
AdjustOutOfFlowPtrs should simply remove the relevant floats from the frame
constructor state?  Or perhaps we want to change the way we do ib splits to
create the anonymous block when we first hit a block kid and push it as the
float containing block?

Comment 1

[Hixie (not reading bugmail)]

We're filing bugs on things that we _almost_ get wrong now? ;-)

Comment 2

[Boris Zbarsky [:bzbarsky]]

Yes, because I want to move in the direction of removing BuildFloatList(), since
it makes pageload O(N^2) in some cases.  I've fixed a bunch of
CSSFrameConstructor issues recently such that we're getting closer to being able
to get rid of it, I think.  Fixing this bug would be a step on the way there.

Comment 3

[Mats Palmgren (inactive)]

Attachment: Testcase

Comment 4

[Mats Palmgren (inactive)]

Attachment: assert

This assert disappears with the patch for bug 307277, does that answer your
question in 266971 comment 14? (I will check that in now...)

Comment 5

[Mats Palmgren (inactive)]

I was referring to Boris' question in bug 307277 comment 14. Sorry for the spam.

Comment 6

[Boris Zbarsky [:bzbarsky]]

Yeah, it does.  Most excellent.

Comment 7

[Mats Palmgren (inactive)]

It fixed the attached testcase in this bug, but I'm seeing the assertion with
the testcase in bug 307277 though...

Comment 8

[Mats Palmgren (inactive)]

Nevermind, I must have looked in the wrong window or something because it
seems like it's actually gone now... Oh well, maybe we should put in that
assertion and see what falls out?

Comment 9

[Boris Zbarsky [:bzbarsky]]

Whatever falls out should block bug 282173.  ;)

Comment 10

[Boris Zbarsky [:bzbarsky]]

Fixed by checkin for bug 307277