REGRESSION (r128678): Several tests fail on WK2 bots (compositing/rtl/rtl-fixed-overflow.html, compositing/rtl/rtl-fixed.html, fast/regions/float-pushed-width-change.html, fast/repaint/fixed-move-after-keyboard-scroll.html)

• Component: Tools / Tests
• Status: RESOLVED FIXED

[Alexey Proskuryakov]

http://build.webkit.org/results/Apple%20MountainLion%20Debug%20WK2%20(Tests)/r128782%20(1139)/results.html

There is one more layer than before:

--- /Volumes/Data/slave/mountainlion-debug-tests-wk2/build/layout-test-results/compositing/rtl/rtl-fixed-overflow-expected.txt
+++ /Volumes/Data/slave/mountainlion-debug-tests-wk2/build/layout-test-results/compositing/rtl/rtl-fixed-overflow-actual.txt
@@ -4,7 +4,12 @@
   (children 1
     (GraphicsLayer
       (bounds 1000.00 1000.00)
-      (children 1
+      (children 2
+        (GraphicsLayer
+          (position 265.00 50.00)
+          (bounds 100.00 100.00)
+          (drawsContent 1)
+        )
         (GraphicsLayer
           (position 265.00 50.00)
           (bounds 100.00 100.00)

[Alexey Proskuryakov]

Confusingly, r128678 added expected results for platform/mac, but those are the same as cross platform ones.

[Alexey Proskuryakov]

Two more tests affected by r128678 on WebKit2 only:
fast/repaint/fixed-move-after-keyboard-scroll.html
fast/regions/float-pushed-width-change.html

[Simon Fraser (smfr)]

Probably because of Beth's fixed position stuff?

[Beth Dakin]

(In reply to comment #3)
> Probably because of Beth's fixed position stuff?

Yes, I'm working on it. We need some wk2-specific test results because of the fact that we're only doing setAcceleratedCompositingForFixedPositionEnabled(true) in WK2, not WK1.

[Beth Dakin]

Created attachment 164453 [details]
Patch

[Tim Horton]

Comment on attachment 164453 [details]
Patch

r=me, assuming you've checked with the person who wrote the test that you're changing (fast/regions/float-pushed-width-change). Does seem safe, though.

[Beth Dakin]

Thanks, Tim! http://trac.webkit.org/changeset/128811