REGRESSION(r109016): TransparencyWin unit tests failing on Chromium

[Adrienne Walker]

During http://trac.webkit.org/log/?action=stop_on_copy&mode=stop_on_copy&rev=109020&stop_rev=109013&limit=100, a number of TransparencyWinTest tests started failing.  This seems very likely to be due to http://trac.webkit.org/changeset/109016.

It's not clear to me how serious these failures are.  Brett--are these something that we should not roll into Chromium? Or, is this just a case where the test needs to be updated?

From http://build.chromium.org/p/chromium.webkit/builders/Webkit%20Vista/builds/12187/ steps:

--snip--
TransparencyWin.WhiteLayerPixelTest: 
Did not complete.
TransparencyWin.WhiteLayer: 
.\tests\TransparencyWinTest.cpp(154): error: Value of: IntSize(14, 12) == helper.m_layerSize
  Actual: false
Expected: true
.\tests\TransparencyWinTest.cpp(155): error: Value of: IntRect(1, 1, 14, 12) == helper.drawRect()
  Actual: false
Expected: true
.\tests\TransparencyWinTest.cpp(186): error: Value of: IntSize(12, 3) == helper.m_layerSize
  Actual: false
Expected: true
.\tests\TransparencyWinTest.cpp(187): error: Value of: IntRect(0, 0, 12, 3) == helper.drawRect()
  Actual: false
Expected: true
TransparencyWin.TextComposite: 
.\tests\TransparencyWinTest.cpp(206): error: Value of: IntSize(14, 12) == helper.m_layerSize
  Actual: false
Expected: true
.\tests\TransparencyWinTest.cpp(207): error: Value of: IntRect(1, 1, 14, 12) == helper.drawRect()
  Actual: false
Expected: true
TransparencyWin.OpaqueCompositeLayer: 
.\tests\TransparencyWinTest.cpp(225): error: Value of: IntSize(14, 12) == helper.m_layerSize
  Actual: false
Expected: true
.\tests\TransparencyWinTest.cpp(226): error: Value of: IntRect(1, 1, 14, 12) == helper.drawRect()
  Actual: false
Expected: true
.\tests\TransparencyWinTest.cpp(241): error: Value of: IntSize(14, 14) == helper.m_layerSize
  Actual: false
Expected: true
.\tests\TransparencyWinTest.cpp(242): error: Value of: IntRect(1, 1, 14, 14) == helper.drawRect()
  Actual: false
Expected: true
.\tests\TransparencyWinTest.cpp(274): error: Value of: IntSize(12, 3) == helper.m_layerSize
  Actual: false
Expected: true
.\tests\TransparencyWinTest.cpp(275): error: Value of: IntRect(0, 0, 12, 3) == helper.drawRect()
  Actual: false
Expected: true
TransparencyWin.NoLayer: 
.\tests\TransparencyWinTest.cpp(114): error: Value of: IntSize(14, 12) == helper.m_layerSize
  Actual: false
Expected: true
.\tests\TransparencyWinTest.cpp(134): error: Value of: IntSize(12, 3) == helper.m_layerSize
  Actual: false
Expected: true
.\tests\TransparencyWinTest.cpp(135): error: Value of: IntRect(4, 1, 12, 3) == helper.drawRect()
  Actual: false
Expected: true
--snip--

[Adrienne Walker]

Committed r109043: <http://trac.webkit.org/changeset/109043>

[Kenichi Ishibashi]

http://build.chromium.org/p/chromium.webkit/builders/Win%20%28dbg%29/builds/5186

I committed r109071 to fix compile errors.
http://trac.webkit.org/changeset/109071

[Kenichi Ishibashi]

http://build.chromium.org/p/chromium.webkit/builders/Webkit%20Win%20%28dbg%29%281%29/builds/10511/steps/webkit_unit_tests/logs/stdio

Further suppression added:
http://trac.webkit.org/changeset/109077

Looks like DRT crashed or hanged. I think it's not safe to roll r109016 and later. I'm not familiar with these code so I couldn't handle the failures.

[Adam Klein]

Aha, I think this is probably a simple problem in the tests. They construct ImageBuffers like so:

ImageBuffer::create(IntSize(16, 16), ColorSpaceDeviceRGB)

But now the second argument is the scale.  I'm going to try a speculative fix of adding the scale argument and re-enabling.

[Adam Klein]

(In reply to comment #4)
> Aha, I think this is probably a simple problem in the tests. They construct ImageBuffers like so:
> 
> ImageBuffer::create(IntSize(16, 16), ColorSpaceDeviceRGB)
> 
> But now the second argument is the scale.  I'm going to try a speculative fix of adding the scale argument and re-enabling.

Speculatively fixed in http://trac.webkit.org/changeset/109119, waiting for bots to cycle before declaring victory.