*** DO NOT LAND *** Implement column-span:all without reparenting renderers. Work in progress.

*** DO NOT LAND *** Implement column-span:all without reparenting renderers.

This is a slightly different approach to implementing column-span:all,
and was sparked by
https://codereview.chromium.org/296413007/diff/200001/Source/core/rendering/RenderMultiColumnFlowThread.cpp#newcode176

Reordering and reparenting renderers is risky business, but it can be
avoided by adding more spanner awareness here and there in central
layout code. This is work in progress, and I'm pretty sure that I'll
have to add more.

[mstensho (USE GERRIT), 2014-09-10 21:39:58]

The first commit is what's already in https://codereview.chromium.org/296413007
, i.e. column-span:all implemented by shuffling the spanner renderer around.

The second patch is the patch I want to show you, where I keep the renderer
where it's initially inserted, and use a special column set for it instead.
Quite a few extra hooks are required in central layout classes with this
solution, but it might be acceptable.

(Just in case you wanted to diff the new approach against it, rather than
master)

[mstensho (USE GERRIT), 2014-09-17 15:09:23]

Just updated the patch. If we copy margin properties from the spanner to the
spanner set, more stuff will just work automatically, and we won't need that
many hooks.

The block direction margins should be ignored on the spanner inside the flow
thread now. The RenderMultiColumnSpannerSet will take care of them instead, in
regular block layout ways.

I haven't worried about inline direction margins yet. They are also copied
(after all, what's inline direction and what's block direction may theoretically
change dynamically), but I haven't added code to ignore them anywhere; i.e. they
are probably applied to both the spanner and the spanner set. Still, we don't
get double inline direction margins, for some reason. So it works. :)

Some code is still required to make sure that margins in columns don't interact
with spanner margins. They are not to collapse with each other.

[mstensho (USE GERRIT), 2014-09-19 14:49:53]

New CL: https://codereview.chromium.org/584033002/