NEW 50342
Unnecessary invalidation of copyright <span> on IE9 Flying Images demo 
https://bugs.webkit.org/show_bug.cgi?id=50342
Summary Unnecessary invalidation of copyright <span> on IE9 Flying Images demo
Darin Fisher (:fishd, Google)
Reported 2010-12-01 12:01:48 PST
Unnecessary invalidation of copyright <span> on IE9 Flying Images demo Watching invalidations that occur on the IE9 Flying Images demo (http://ie.microsoft.com/testdrive/Performance/FlyingImages/Default.html), I observe that the copyright span at the bottom of the page is continually invalidated. I reduced this down somewhat. See the attached test case. In the test case, if I give #board an explicit "left" and "top" coordinate, then the problem goes away. (If you are running Chromium, you can observe invalidations by running with the --show-paint-rects command line switch.)
Attachments
testcase (1.05 KB, text/html)
2010-12-01 12:04 PST, Darin Fisher (:fishd, Google) 		no flags
Details
Patch (9.17 KB, patch)
2010-12-01 17:31 PST, James Robinson 		no flags
DetailsFormatted DiffDiff
Show Obsolete (1) View All Add attachment proposed patch, testcase, etc.
Darin Fisher (:fishd, Google)
Comment 1 2010-12-01 12:04:00 PST
Created attachment 75300 [details] testcase
Sam Weinig
Comment 2 2010-12-01 12:49:43 PST
You can also observe this using QuartzDebug on OS X.
Simon Fraser (smfr)
Comment 3 2010-12-01 12:56:26 PST
To debug, see who's calling repaint() on the span
James Robinson
Comment 4 2010-12-01 13:21:06 PST
render tree: RenderView 0x7ffff7ef1a98 #document 0x7ffff7ef8000 RenderBlock 0x7ffff7f30478 HTML 0x7fffeb51c800 * RenderBody 0x7ffff7f30ef8 BODY 0x7fffeb064240 RenderBlock (positioned) 0x7fffe6e78718 DIV 0x7fffeaa4cb00 RenderBlock (positioned) 0x7fffe6e78558 DIV 0x7fffeaa4ca80 STYLE=top: -300px; RenderBlock (positioned) 0x7fffe6e78478 SPAN 0x7fffeaa4c980 RenderText 0x7fffeaa4c918 #text 0x7fffeaa48540 "Blah, blah" RenderBlock::layoutInlineChildren() on the <body>'s RenderBlock is marking the <span> as m_normalChildNeedsLayout because the <span> has a percentage width. This means it later gets marked for repainting even though none of its children actually move. http://trac.webkit.org/browser/trunk/WebCore/rendering/RenderBlockLineLayout.cpp#L552 seems a little pessimistic.
James Robinson
Comment 5 2010-12-01 17:31:10 PST
Created attachment 75337 [details] Patch
Eric Seidel (no email)
Comment 6 2010-12-02 17:38:02 PST
Comment on attachment 75337 [details] Patch This looks like a mitz/hyatt patch to review. :)
James Robinson
Comment 7 2010-12-02 17:44:34 PST
Actually hyatt typed out the meat of this patch verbatim in IRC. I just added the comment, testcase, and verified that this doesn't cause all the other layout tests to explode.
WebKit Commit Bot
Comment 8 2010-12-02 21:58:23 PST
Comment on attachment 75337 [details] Patch Rejecting patch 75337 from commit-queue. Failed to run "['./WebKitTools/Scripts/webkit-patch', '--status-host=queues.webkit.org', '--bot-id=abarth-cq-sl', 'build-and-test', '--no-clean', '--no-update', '--test', '--non-interactive']" exit_code: 2 Last 500 characters of output: ning/auto ....... fast/blockflow ................................................ fast/body-propagation/background-color ..................... fast/body-propagation/background-image ......................... fast/body-propagation/overflow ................... fast/borders ... fast/borders/border-fit.html -> failed Exiting early after 1 failures. 5969 tests run. 105.41s total testing time 5968 test cases (99%) succeeded 1 test case (<1%) had incorrect layout 3 test cases (<1%) had stderr output Full output: http://queues.webkit.org/results/6751028
Eric Seidel (no email)
Comment 9 2010-12-14 01:28:47 PST
Seems this patch will need an update since it breaks a test.
Eric Seidel (no email)
Comment 10 2010-12-14 15:14:06 PST
Attachment 75337 [details] was posted by a committer and has review+, assigning to James Robinson for commit.
James Robinson
Comment 11 2010-12-14 17:40:08 PST
Comment on attachment 75337 [details] Patch Clearing flags, patch was reverted.
Note You need to log in before you can comment on or make changes to this bug.