Fix a regression when painting CMYK over ARGB32_Premultiplied

With the introduction of CMYK32 the old logic of assuming depth
meant compatible alpha version no longer works. So change the logic
to more explicitly return compatible opaque or alpha versions and
remove the now invalid qt_maybeAlphaVersionWithSameDepth.

Change-Id: Ib1f7b76b0ce0eae7d49a0dfe369918a746bbe2b4
Reviewed-by: Giuseppe D'Angelo <giuseppe.dangelo@kdab.com>

1 files changed, 1 insertions, 1 deletions

diff --git a/src/plugins/platforms/windows/qwindowsbackingstore.cpp b/src/plugins/platforms/windows/qwindowsbackingstore.cpp
index 0f9d0172d9..07918f6094 100644
--- a/src/plugins/platforms/windows/qwindowsbackingstore.cpp
+++ b/src/plugins/platforms/windows/qwindowsbackingstore.cpp
@@ -117,7 +117,7 @@ void QWindowsBackingStore::resize(const QSize &size, const QRegion &region)
         if (QImage::toPixelFormat(format).alphaUsage() == QPixelFormat::UsesAlpha)
             m_alphaNeedsFill = true;
         else // upgrade but here we know app painting does not rely on alpha hence no need to fill
-            format = qt_maybeAlphaVersionWithSameDepth(format);
+            format = qt_maybeDataCompatibleAlphaVersion(format);
 
         QWindowsNativeImage *oldwni = m_image.data();
         auto *newwni = new QWindowsNativeImage(size.width(), size.height(), format);