diff options
| author | nick <nick@chromium.org> | 2015-12-17 17:44:12 -0800 |
|---|---|---|
| committer | Commit bot <commit-bot@chromium.org> | 2015-12-18 01:44:55 +0000 |
| commit | 8814e652c72b6e5dd62a42772590c15f7dc8372b (patch) | |
| tree | 139f32ef371f3e41bef4dff39d04896a6a9a521b /native_client_sdk | |
| parent | a75be0ec62646aebed4cfd9d05c4690997bdb959 (diff) | |
| download | chromium_src-8814e652c72b6e5dd62a42772590c15f7dc8372b.zip chromium_src-8814e652c72b6e5dd62a42772590c15f7dc8372b.tar.gz chromium_src-8814e652c72b6e5dd62a42772590c15f7dc8372b.tar.bz2 | |
Enhance the WebContentsObserverSanityChecker to validate tree structure
In particular, assert the following:
- Parents are created/committed before children
- Children are deleted/detached before parents
- The value returned by GetParent() does not change over an RFH's lifetime.
This exposed two problematic and surprising behaviors:
- Calling GetParent() during FrameDeleted/RenderFrameDeleted would sometimes
return a different node from GetParent() during creation (!!!). This is
because we deleted the children of the old frame after setting the new frame
as the current.
- Parents were getting detached before their children, due to our reliance
on ScopedVector's cleanup of children in FrameTreeNode.
Review URL: https://codereview.chromium.org/1410373008
Cr-Commit-Position: refs/heads/master@{#365976}
Diffstat (limited to 'native_client_sdk')
0 files changed, 0 insertions, 0 deletions