szager/vmpstr: this might be related to the other painting bugs (e.g.  issue 813835 )?

I can't immediately reproduce this. I will try and see if this reproduces before some of our fixes (maybe this was fixed by something else)

Update: I think it's something timing dependent. On a debug ToT build, I can't reproduce no matter how much I try. However, if I do a bisect-builds tool, then every version I try reproduces pretty quickly.

I'm currently building mac release to try and repro on ToT.

Update: Bisected this down to https://codereview.chromium.org/2728273002 which was a year or so ago.

I have a probably-too-aggressive fix here: https://chromium-review.googlesource.com/c/chromium/src/+/942347

We can probably rethink what needs to happen more, but for now it's not a bad fix for a rare case and it fixes the printing dialog.

 Issue 814662  has been merged into this issue.

Another approach to fix this could be to analyze every change in 
https://codereview.chromium.org/2728273002. There are various attach/detach
methods that were updated. Perhaps we need to force the ...rebuild_ state on
one of them?

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/93b15c2c0cfeea861d3f64af9ab9dab670060925

commit 93b15c2c0cfeea861d3f64af9ab9dab670060925
Author: Vladimir Levin <vmpstr@chromium.org>
Date: Sat Mar 03 00:07:45 2018

[RLS] Ensure that iframes graphics layers are attached.

This patch ensures that if we remove a child frame's graphics layer from
its parent, that we reattach its (possibly new) root graphics layer back
to the parent it had before.

The parent does not change during this, because we're in a stack
processing the child frame, and the parent would reference the parent
frame that is currently paused.

This mimics pre-RLS code path where the child frame always had a parent
accessible directly from the compositor (ie root_content_layer_), except
here we save off whatever we were attached to already.

R=chrishtr@chromium.org, skobes@chromium.org, pdr@chromium.org

Bug:  815121 
Cq-Include-Trybots: master.tryserver.blink:linux_trusty_blink_rel;master.tryserver.chromium.linux:linux_layout_tests_slimming_paint_v2
Change-Id: I9e63d2d513df24eaa13a30ee36582b5c807f4aff
Reviewed-on: https://chromium-review.googlesource.com/946566
Commit-Queue: vmpstr <vmpstr@chromium.org>
Reviewed-by: Steve Kobes <skobes@chromium.org>
Reviewed-by: Chris Harrelson <chrishtr@chromium.org>
Cr-Commit-Position: refs/heads/master@{#540692}
[add] https://crrev.com/93b15c2c0cfeea861d3f64af9ab9dab670060925/third_party/WebKit/LayoutTests/compositing/iframe-graphics-tree-changes-parents-does-not-expected.html
[add] https://crrev.com/93b15c2c0cfeea861d3f64af9ab9dab670060925/third_party/WebKit/LayoutTests/compositing/iframe-graphics-tree-changes-parents-does-not.html
[add] https://crrev.com/93b15c2c0cfeea861d3f64af9ab9dab670060925/third_party/WebKit/LayoutTests/compositing/resources/iframe-graphics-tree-changes-parents-does-not-frame.html
[modify] https://crrev.com/93b15c2c0cfeea861d3f64af9ab9dab670060925/third_party/WebKit/Source/core/paint/compositing/PaintLayerCompositor.cpp
[modify] https://crrev.com/93b15c2c0cfeea861d3f64af9ab9dab670060925/third_party/WebKit/Source/core/paint/compositing/PaintLayerCompositor.h

Merge request for #12

Your change meets the bar and is auto-approved for M66. Please go ahead and merge the CL to branch 3359 manually. Please contact milestone owner if you have questions.
Owners: cmasso@(Android), cmasso@(iOS), josafat@(ChromeOS), abdulsyed@(Desktop)

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/aa2ef48e42025130cb0006b2dd3d1851cbac4268

commit aa2ef48e42025130cb0006b2dd3d1851cbac4268
Author: Vladimir Levin <vmpstr@chromium.org>
Date: Tue Mar 06 19:47:55 2018

[RLS] Ensure that iframes graphics layers are attached.

This patch ensures that if we remove a child frame's graphics layer from
its parent, that we reattach its (possibly new) root graphics layer back
to the parent it had before.

The parent does not change during this, because we're in a stack
processing the child frame, and the parent would reference the parent
frame that is currently paused.

This mimics pre-RLS code path where the child frame always had a parent
accessible directly from the compositor (ie root_content_layer_), except
here we save off whatever we were attached to already.

R=chrishtr@chromium.org, pdr@chromium.org, skobes@chromium.org
TBR=vmpstr@chromium.org

(cherry picked from commit 93b15c2c0cfeea861d3f64af9ab9dab670060925)

Bug:  815121 
Cq-Include-Trybots: master.tryserver.blink:linux_trusty_blink_rel;master.tryserver.chromium.linux:linux_layout_tests_slimming_paint_v2
Change-Id: I9e63d2d513df24eaa13a30ee36582b5c807f4aff
Reviewed-on: https://chromium-review.googlesource.com/946566
Commit-Queue: vmpstr <vmpstr@chromium.org>
Reviewed-by: Steve Kobes <skobes@chromium.org>
Reviewed-by: Chris Harrelson <chrishtr@chromium.org>
Cr-Original-Commit-Position: refs/heads/master@{#540692}
Reviewed-on: https://chromium-review.googlesource.com/951967
Reviewed-by: vmpstr <vmpstr@chromium.org>
Cr-Commit-Position: refs/branch-heads/3359@{#36}
Cr-Branched-From: 66afc5e5d10127546cc4b98b9117aff588b5e66b-refs/heads/master@{#540276}
[add] https://crrev.com/aa2ef48e42025130cb0006b2dd3d1851cbac4268/third_party/WebKit/LayoutTests/compositing/iframe-graphics-tree-changes-parents-does-not-expected.html
[add] https://crrev.com/aa2ef48e42025130cb0006b2dd3d1851cbac4268/third_party/WebKit/LayoutTests/compositing/iframe-graphics-tree-changes-parents-does-not.html
[add] https://crrev.com/aa2ef48e42025130cb0006b2dd3d1851cbac4268/third_party/WebKit/LayoutTests/compositing/resources/iframe-graphics-tree-changes-parents-does-not-frame.html
[modify] https://crrev.com/aa2ef48e42025130cb0006b2dd3d1851cbac4268/third_party/WebKit/Source/core/paint/compositing/PaintLayerCompositor.cpp
[modify] https://crrev.com/aa2ef48e42025130cb0006b2dd3d1851cbac4268/third_party/WebKit/Source/core/paint/compositing/PaintLayerCompositor.h

 Issue 819221  has been merged into this issue.

This bug is verified in 66.0.3359.31 /10452.11.0 Jaq, Squawks, Chell, Clapper