[Reland] Correct fixed-position recording for Slimming Paint

[Reland] Correct fixed-position recording for Slimming Paint

Before this CL, the location of display items of the descendants of a
fixed-position container is specified in absolute coordinate. Scrolling the
viewport with non-composited fixed position container will need a full subtree
invalidation.

This CL paints fixed position layers by adding a full-layer transform then
change the painting root to the current layer for painting. Identical to
how we paint non-composited transformed layer. This way scrolling won't need
to invalidate any display items in the subtree.

As a drive-by change, LayerFixedPositionRecorder, FixedPositionDisplayItem,
and FixedPositionContainerDisplayItem are added for layerization.
Currently they are no-op.

BUG=482170
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=194784

[trchen, 2015-04-30 00:53:32]

The CQ bit was checked by trchen@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2015-04-30 00:53:56]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1110653003/1

[commit-bot: I haz the power, 2015-04-30 00:56:51]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-04-30 00:56:51]

Dry run: Try jobs failed on following builders:
  mac_blink_rel on tryserver.blink (JOB_FAILED,
http://build.chromium.org/p/tryserver.blink/builders/mac_blink_rel/builds/53892)

[trchen, 2015-04-30 01:01:13]

trchen@chromium.org changed reviewers:
+ chrishtr@chromium.org

[trchen, 2015-04-30 01:01:13]

This CL fixes the CrOS browser_tests assertion failures.

The only difference from previous CL is the change to
isFixedPositionedContainer() in DeprecatedLayerPainter.cpp. A layer is only a
fixed-pos container if it has transform, not any transform-related property.

The bug mechanism is that we have 3 layers like this:

layer(transformed)
  layer(not transformed, but has transform-related property)
    layer(fixed-pos)

When painting the fixed-pos layer, it tries to convert local layer coords to the
transformed layer. Normally it is supposed be valid, but the assertion inside of
accumulateOffsetTowardsAncestor mistook the intermediate layer being
transformed.

[trchen, 2015-04-30 01:05:31]

The CQ bit was checked by trchen@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2015-04-30 01:05:43]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1110653003/20001

[commit-bot: I haz the power, 2015-04-30 02:38:10]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-04-30 02:38:11]

Dry run: This issue passed the CQ dry run.

[ajuma, 2015-04-30 14:22:32]

On 2015/04/30 01:01:13, trchen wrote:
> This CL fixes the CrOS browser_tests assertion failures.
> 
> The only difference from previous CL is the change to
> isFixedPositionedContainer() in DeprecatedLayerPainter.cpp. A layer is only a
> fixed-pos container if it has transform, not any transform-related property.

This is indeed what the spec says, but note that Firefox (and, not surprisingly,
Safari) creates fixed-pos containers for any transform-related property. An
alternative would be to fix up accumulateOffsetTowardsAncestor so that it
doesn't assume that non-root fixed position containers are transformed (e.g.,
see https://codereview.chromium.org/1111423002/).

[chrishtr, 2015-04-30 17:34:59]

https://codereview.chromium.org/1110653003/diff/20001/Source/core/paint/Depre...
File Source/core/paint/DeprecatedPaintLayer.cpp (right):

https://codereview.chromium.org/1110653003/diff/20001/Source/core/paint/Depre...
Source/core/paint/DeprecatedPaintLayer.cpp:939: return layer->isRootLayer() ||
layer->transform();
This should be landed first I think, probably with the approach from ajuma.

[Xianzhu, 2015-04-30 17:54:44]

wangxianzhu@chromium.org changed reviewers:
+ wangxianzhu@chromium.org

[Xianzhu, 2015-04-30 17:54:44]

https://codereview.chromium.org/1110653003/diff/20001/Source/core/paint/Depre...
File Source/core/paint/DeprecatedPaintLayer.cpp (right):

https://codereview.chromium.org/1110653003/diff/20001/Source/core/paint/Depre...
Source/core/paint/DeprecatedPaintLayer.cpp:939: return layer->isRootLayer() ||
layer->transform();
If we want change this, we should also check other hasTransformRelatedProperty()
callers. For example, perhaps we also change
LayoutObject::canContainFixedPositionObjects().

I think this should be a separate change because it changes behavior other than
the bug this CL is supposed to fix.

[trchen, 2015-04-30 22:27:58]

The CQ bit was checked by trchen@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2015-04-30 22:28:43]

Dry run: CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1110653003/40001

[trchen, 2015-04-30 22:28:54]

Got it. Reverted the isFixedPositionedContainer change since ajuma made a fix
for the assertion failure. Now the CL is identical to the previous one.

[chrishtr, 2015-04-30 22:29:48]

lgtm

[trchen, 2015-05-01 00:16:52]

The CQ bit was unchecked by trchen@chromium.org

[trchen, 2015-05-01 01:12:20]

The CQ bit was checked by trchen@chromium.org

[commit-bot: I haz the power, 2015-05-01 01:12:37]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1110653003/40001

[commit-bot: I haz the power, 2015-05-01 01:27:17]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-05-01 01:27:20]

Failed to apply patch for
LayoutTests/virtual/slimmingpaint/fast/repaint/absolute-position-change-containing-block-expected.txt:
While running patch -p1 --forward --force --no-backup-if-mismatch;
  patching file
LayoutTests/virtual/slimmingpaint/fast/repaint/absolute-position-change-containing-block-expected.txt
  Hunk #1 FAILED at 17.
  1 out of 1 hunk FAILED -- saving rejects to file
LayoutTests/virtual/slimmingpaint/fast/repaint/absolute-position-change-containing-block-expected.txt.rej

Patch:      
LayoutTests/virtual/slimmingpaint/fast/repaint/absolute-position-change-containing-block-expected.txt
Index:
LayoutTests/virtual/slimmingpaint/fast/repaint/absolute-position-change-containing-block-expected.txt
diff --git
a/LayoutTests/virtual/slimmingpaint/fast/repaint/absolute-position-change-containing-block-expected.txt
b/LayoutTests/virtual/slimmingpaint/fast/repaint/absolute-position-change-containing-block-expected.txt
index
0f6254ca247653d1146f3bb50e9bb76922b6d217..19df4810f2d2f21d86b7338ceec1803bbc74f2fb
100644
---
a/LayoutTests/virtual/slimmingpaint/fast/repaint/absolute-position-change-containing-block-expected.txt
+++
b/LayoutTests/virtual/slimmingpaint/fast/repaint/absolute-position-change-containing-block-expected.txt
@@ -17,6 +17,8 @@
       "paintInvalidationClients": [
         "LayoutBlockFlow (positioned) DIV id='container' class='fixed blue'",
         "LayoutBlockFlow (positioned) DIV class='absolute green'",
+        "LayoutBlockFlow DIV id='container' class='fixed blue'",
+        "LayoutBlockFlow (positioned) DIV class='absolute green'",
         "LayoutBlockFlow BODY",
         "LayoutBlockFlow DIV id='container' class='fixed blue'",
         "LayoutBlockFlow (positioned) DIV class='absolute green'"

[trchen, 2015-05-01 02:35:36]

The CQ bit was checked by trchen@chromium.org

[trchen, 2015-05-01 02:35:36]

The patchset sent to the CQ was uploaded after l-g-t-m from
chrishtr@chromium.org
Link to the patchset: https://codereview.chromium.org/1110653003/#ps60001
(title: "rebase")

[commit-bot: I haz the power, 2015-05-01 02:35:44]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1110653003/60001

[commit-bot: I haz the power, 2015-05-01 03:56:50]

Committed patchset #4 (id:60001) as
https://src.chromium.org/viewvc/blink?view=rev&revision=194784