Merge branch '4.7' of scm.dev.nokia.troll.no:qt/oslo-staging-2 into 4.7-integration

* '4.7' of scm.dev.nokia.troll.no:qt/oslo-staging-2:
  Don't rely on mapFromGlobal in QGraphicsScenePrivate::itemsAtPosition.
  Refactor blend_transformed_bilinear to simplify the blend type checking
  Note RTL behavior changes in docs and changelog
  Implement qt_memfill32 with Neon.
  Implement the composition mode Plus with Neon.
  Fix the broken unicode detection of ODBC driver.

8 files changed, 196 insertions, 96 deletions

diff --git a/src/gui/graphicsview/qgraphicsscene.cpp b/src/gui/graphicsview/qgraphicsscene.cpp
index 3c23884..4fb754c 100644
--- a/src/gui/graphicsview/qgraphicsscene.cpp
+++ b/src/gui/graphicsview/qgraphicsscene.cpp
@@ -1074,7 +1074,7 @@ void QGraphicsScenePrivate::enableMouseTrackingOnViews()
 /*!
     Returns all items for the screen position in \a event.
 */
-QList<QGraphicsItem *> QGraphicsScenePrivate::itemsAtPosition(const QPoint &screenPos,
+QList<QGraphicsItem *> QGraphicsScenePrivate::itemsAtPosition(const QPoint &/*screenPos*/,
                                                               const QPointF &scenePos,
                                                               QWidget *widget) const
 {
@@ -1083,16 +1083,12 @@ QList<QGraphicsItem *> QGraphicsScenePrivate::itemsAtPosition(const QPoint &scre
     if (!view)
         return q->items(scenePos, Qt::IntersectsItemShape, Qt::DescendingOrder, QTransform());
 
-    const QRectF pointRect(QPointF(widget->mapFromGlobal(screenPos)), QSizeF(1, 1));
+    const QRectF pointRect(scenePos, QSizeF(1, 1));
     if (!view->isTransformed())
         return q->items(pointRect, Qt::IntersectsItemShape, Qt::DescendingOrder);
 
     const QTransform viewTransform = view->viewportTransform();
-    if (viewTransform.type() <= QTransform::TxScale) {
-        return q->items(viewTransform.inverted().mapRect(pointRect), Qt::IntersectsItemShape,
-                        Qt::DescendingOrder, viewTransform);
-    }
-    return q->items(viewTransform.inverted().map(pointRect), Qt::IntersectsItemShape,
+    return q->items(pointRect, Qt::IntersectsItemShape,
                     Qt::DescendingOrder, viewTransform);
 }
 
diff --git a/src/gui/kernel/qwidget.cpp b/src/gui/kernel/qwidget.cpp
index 3d2bfe2..ea3dcab 100644
--- a/src/gui/kernel/qwidget.cpp
+++ b/src/gui/kernel/qwidget.cpp
@@ -4864,6 +4864,8 @@ void QWidgetPrivate::resolveLayoutDirection()
     has been called for the parent do not inherit the parent's layout
     direction.
 
+    This method no longer affects text layout direction since Qt 4.7.
+
     \sa QApplication::layoutDirection
 */
 void QWidget::setLayoutDirection(Qt::LayoutDirection direction)
diff --git a/src/gui/painting/qdrawhelper.cpp b/src/gui/painting/qdrawhelper.cpp
index 276da93..59bef66 100644
--- a/src/gui/painting/qdrawhelper.cpp
+++ b/src/gui/painting/qdrawhelper.cpp
@@ -1757,9 +1757,7 @@ Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_solid_Plus_impl(uint *dest, int
     for (int i = 0; i < length; ++i) {
         PRELOAD_COND(dest)
         uint d = dest[i];
-#define MIX(mask) (qMin(((qint64(s)&mask) + (qint64(d)&mask)), qint64(mask)))
-        d = (MIX(AMASK) | MIX(RMASK) | MIX(GMASK) | MIX(BMASK));
-#undef MIX
+        d = comp_func_Plus_one_pixel(d, s);
         coverage.store(&dest[i], d);
     }
 }
@@ -1781,9 +1779,7 @@ Q_STATIC_TEMPLATE_FUNCTION inline void comp_func_Plus_impl(uint *dest, const uin
         uint d = dest[i];
         uint s = src[i];
 
-#define MIX(mask) (qMin(((qint64(s)&mask) + (qint64(d)&mask)), qint64(mask)))
-        d = (MIX(AMASK) | MIX(RMASK) | MIX(GMASK) | MIX(BMASK));
-#undef MIX
+        d = comp_func_Plus_one_pixel(d, s);
 
         coverage.store(&dest[i], d);
     }
@@ -5163,6 +5159,61 @@ static void blend_tiled_rgb444(int count, const QSpan *spans, void *userData)
         blend_tiled_generic<RegularSpans>(count, spans, userData);
 }
 
+template<int>
+Q_STATIC_INLINE_FUNCTION void blend_transformed_bilinear_argb_clamp_coordinates  (int &x1, int &x2, int &y1, int &y2,
+                                                                                  const int image_width, const int image_height,
+                                                                                  const int image_x1, const int image_x2,
+                                                                                  const int image_y1, const int image_y2);
+
+template<>
+Q_STATIC_INLINE_FUNCTION void blend_transformed_bilinear_argb_clamp_coordinates <BlendTransformedBilinearTiled> (int &x1, int &x2, int &y1, int &y2,
+                                                                                                                 const int image_width, const int image_height,
+                                                                                                                 const int image_x1, const int image_x2,
+                                                                                                                 const int image_y1, const int image_y2)
+{
+    Q_UNUSED(image_x1)
+    Q_UNUSED(image_x2)
+    Q_UNUSED(image_y1)
+    Q_UNUSED(image_y2)
+    x1 %= image_width;
+    if (x1 < 0) x1 += image_width;
+    x2 = x1 + 1;
+    x2 %= image_width;
+
+    y1 %= image_height;
+    if (y1 < 0) y1 += image_height;
+    y2 = y1 + 1;
+    y2 %= image_height;
+
+    Q_ASSERT(x1 >= 0 && x1 < image_width);
+    Q_ASSERT(x2 >= 0 && x2 < image_width);
+    Q_ASSERT(y1 >= 0 && y1 < image_height);
+    Q_ASSERT(y2 >= 0 && y2 < image_height);
+}
+
+template <>
+Q_STATIC_INLINE_FUNCTION void blend_transformed_bilinear_argb_clamp_coordinates<BlendTransformedBilinear>(int &x1, int &x2, int &y1, int &y2,
+                                                                                               const int image_width, const int image_height,
+                                                                                               const int image_x1, const int image_x2,
+                                                                                               const int image_y1, const int image_y2)
+{
+    Q_UNUSED(image_width)
+    Q_UNUSED(image_height)
+    if (x1 < image_x1) {
+        x2 = x1 = image_x1;
+    } else if (x1 >= image_x2) {
+        x2 = x1 = image_x2;
+    } else {
+        x2 = x1 + 1;
+    }
+    if (y1 < image_y1) {
+        y2 = y1 = image_y1;
+    } else if (y1 >= image_y2) {
+        y2 = y1 = image_y2;
+    } else {
+        y2 = y1 + 1;
+    }
+}
 
 template <SpanMethod spanMethod, TextureBlendType blendType>  /* blendType must be either BlendTransformedBilinear or BlendTransformedBilinearTiled */
 Q_STATIC_TEMPLATE_FUNCTION void blend_transformed_bilinear_argb(int count, const QSpan *spans, void *userData)
@@ -5179,8 +5230,8 @@ Q_STATIC_TEMPLATE_FUNCTION void blend_transformed_bilinear_argb(int count, const
 
     const int image_x1 = data->texture.x1;
     const int image_y1 = data->texture.y1;
-    const int image_x2 = data->texture.x2;
-    const int image_y2 = data->texture.y2;
+    const int image_x2 = data->texture.x2 - 1;
+    const int image_y2 = data->texture.y2 - 1;
     const int image_width = data->texture.width;
     const int image_height = data->texture.height;
     const int scanline_offset = data->texture.bytesPerLine / 4;
@@ -5210,43 +5261,16 @@ Q_STATIC_TEMPLATE_FUNCTION void blend_transformed_bilinear_argb(int count, const
                 int l = qMin(length, buffer_size);
                 const uint *end = buffer + l;
                 uint *b = buffer;
-                while (b < end) {
+                while (b != end) {
                     int x1 = (x >> 16);
                     int x2;
                     int y1 = (y >> 16);
                     int y2;
 
-                    if (blendType == BlendTransformedBilinearTiled) {
-                        x1 %= image_width;
-                        if (x1 < 0) x1 += image_width;
-                        x2 = x1 + 1;
-                        x2 %= image_width;
-
-                        y1 %= image_height;
-                        if (y1 < 0) y1 += image_height;
-                        y2 = y1 + 1;
-                        y2 %= image_height;
-
-                        Q_ASSERT(x1 >= 0 && x1 < image_width);
-                        Q_ASSERT(x2 >= 0 && x2 < image_width);
-                        Q_ASSERT(y1 >= 0 && y1 < image_height);
-                        Q_ASSERT(y2 >= 0 && y2 < image_height);
-                    } else {
-                        if (x1 < image_x1) {
-                            x2 = x1 = image_x1;
-                        } else if (x1 >= image_x2 - 1) {
-                            x2 = x1 = image_x2 - 1;
-                        } else {
-                            x2 = x1 + 1;
-                        }
-                        if (y1 < image_y1) {
-                            y2 = y1 = image_y1;
-                        } else if (y1 >= image_y2 - 1) {
-                            y2 = y1 = image_y2 - 1;
-                        } else {
-                            y2 = y1 + 1;
-                        }
-                    }
+                    blend_transformed_bilinear_argb_clamp_coordinates<blendType>(x1, x2, y1, y2,
+                                                                                 image_width, image_height,
+                                                                                 image_x1, image_x2,
+                                                                                 image_y1, image_y2);
 
                     int y1_offset = y1 * scanline_offset;
                     int y2_offset = y2 * scanline_offset;
@@ -5311,7 +5335,7 @@ Q_STATIC_TEMPLATE_FUNCTION void blend_transformed_bilinear_argb(int count, const
                 int l = qMin(length, buffer_size);
                 const uint *end = buffer + l;
                 uint *b = buffer;
-                while (b < end) {
+                while (b != end) {
                     const qreal iw = w == 0 ? 1 : 1 / w;
                     const qreal px = x * iw - 0.5;
                     const qreal py = y * iw - 0.5;
@@ -5326,37 +5350,10 @@ Q_STATIC_TEMPLATE_FUNCTION void blend_transformed_bilinear_argb(int count, const
                     int idistx = 256 - distx;
                     int idisty = 256 - disty;
 
-                    if (blendType == BlendTransformedBilinearTiled) {
-                        x1 %= image_width;
-                        if (x1 < 0) x1 += image_width;
-                        x2 = x1 + 1;
-                        x2 %= image_width;
-
-                        y1 %= image_height;
-                        if (y1 < 0) y1 += image_height;
-                        y2 = y1 + 1;
-                        y2 %= image_height;
-
-                        Q_ASSERT(x1 >= 0 && x1 < image_width);
-                        Q_ASSERT(x2 >= 0 && x2 < image_width);
-                        Q_ASSERT(y1 >= 0 && y1 < image_height);
-                        Q_ASSERT(y2 >= 0 && y2 < image_height);
-                    } else {
-                        if (x1 < image_x1) {
-                            x2 = x1 = image_x1;
-                        } else if (x1 >= image_x2 - 1) {
-                            x2 = x1 = image_x2 - 1;
-                        } else {
-                            x2 = x1 + 1;
-                        }
-                        if (y1 < image_y1) {
-                            y2 = y1 = image_y1;
-                        } else if (y1 >= image_y2 - 1) {
-                            y2 = y1 = image_y2 - 1;
-                        } else {
-                            y2 = y1 + 1;
-                        }
-                    }
+                    blend_transformed_bilinear_argb_clamp_coordinates<blendType>(x1, x2, y1, y2,
+                                                                                 image_width, image_height,
+                                                                                 image_x1, image_x2,
+                                                                                 image_y1, image_y2);
 
                     int y1_offset = y1 * scanline_offset;
                     int y2_offset = y2 * scanline_offset;
@@ -7911,11 +7908,13 @@ void qInitDrawhelperAsm()
 
             functionForMode_C[QPainter::CompositionMode_SourceOver] = qt_blend_argb32_on_argb32_scanline_neon;
             functionForModeSolid_C[QPainter::CompositionMode_SourceOver] = comp_func_solid_SourceOver_neon;
+            functionForMode_C[QPainter::CompositionMode_Plus] = comp_func_Plus_neon;
             destFetchProc[QImage::Format_RGB16] = qt_destFetchRGB16_neon;
             destStoreProc[QImage::Format_RGB16] = qt_destStoreRGB16_neon;
 
             qMemRotateFunctions[QImage::Format_RGB16][0] = qt_memrotate90_16_neon;
             qMemRotateFunctions[QImage::Format_RGB16][2] = qt_memrotate270_16_neon;
+            qt_memfill32 = qt_memfill32_neon;
         }
 #endif
 
diff --git a/src/gui/painting/qdrawhelper_neon.cpp b/src/gui/painting/qdrawhelper_neon.cpp
index 03fe075..ed15c5c 100644
--- a/src/gui/painting/qdrawhelper_neon.cpp
+++ b/src/gui/painting/qdrawhelper_neon.cpp
@@ -51,6 +51,44 @@
 
 QT_BEGIN_NAMESPACE
 
+void qt_memfill32_neon(quint32 *dest, quint32 value, int count)
+{
+    const int epilogueSize = count % 16;
+    if (count >= 16) {
+        quint32 *const neonEnd = dest + count - epilogueSize;
+        register uint32x4_t valueVector1 asm ("q0") = vdupq_n_u32(value);
+        register uint32x4_t valueVector2 asm ("q1") = valueVector1;
+        while (dest != neonEnd) {
+            asm volatile (
+                "vst2.32     { d0, d1, d2, d3 }, [%[DST]] !\n\t"
+                "vst2.32     { d0, d1, d2, d3 }, [%[DST]] !\n\t"
+                : [DST]"+r" (dest)
+                : [VALUE1]"w"(valueVector1), [VALUE2]"w"(valueVector2)
+                : "memory"
+            );
+        }
+    }
+
+    switch (epilogueSize)
+    {
+    case 15:     *dest++ = value;
+    case 14:     *dest++ = value;
+    case 13:     *dest++ = value;
+    case 12:     *dest++ = value;
+    case 11:     *dest++ = value;
+    case 10:     *dest++ = value;
+    case 9:      *dest++ = value;
+    case 8:      *dest++ = value;
+    case 7:      *dest++ = value;
+    case 6:      *dest++ = value;
+    case 5:      *dest++ = value;
+    case 4:      *dest++ = value;
+    case 3:      *dest++ = value;
+    case 2:      *dest++ = value;
+    case 1:      *dest++ = value;
+    }
+}
+
 static inline uint16x8_t qvdiv_255_u16(uint16x8_t x, uint16x8_t half)
 {
     // result = (x + (x >> 8) + 0x80) >> 8
@@ -622,6 +660,61 @@ void QT_FASTCALL comp_func_solid_SourceOver_neon(uint *destPixels, int length, u
     }
 }
 
+void QT_FASTCALL comp_func_Plus_neon(uint *dst, const uint *src, int length, uint const_alpha)
+{
+    if (const_alpha == 255) {
+        uint *const end = dst + length;
+        uint *const neonEnd = end - 3;
+
+        while (dst < neonEnd) {
+            asm volatile (
+                "vld2.8     { d0, d1 }, [%[SRC]] !\n\t"
+                "vld2.8     { d2, d3 }, [%[DST]]\n\t"
+                "vqadd.u8 q0, q0, q1\n\t"
+                "vst2.8     { d0, d1 }, [%[DST]] !\n\t"
+                : [DST]"+r" (dst), [SRC]"+r" (src)
+                :
+                : "memory", "d0", "d1", "d2", "d3", "q0", "q1"
+            );
+        }
+
+        while (dst != end) {
+            *dst = comp_func_Plus_one_pixel(*dst, *src);
+            ++dst;
+            ++src;
+        }
+    } else {
+        int x = 0;
+        const int one_minus_const_alpha = 255 - const_alpha;
+        const uint16x8_t constAlphaVector = vdupq_n_u16(const_alpha);
+        const uint16x8_t oneMinusconstAlphaVector = vdupq_n_u16(one_minus_const_alpha);
+
+        const uint16x8_t half = vdupq_n_u16(0x80);
+        for (; x < length - 3; x += 4) {
+            const uint32x4_t src32 = vld1q_u32((uint32_t *)&src[x]);
+            const uint8x16_t src8 = vreinterpretq_u8_u32(src32);
+            uint8x16_t dst8 = vld1q_u8((uint8_t *)&dst[x]);
+            uint8x16_t result = vqaddq_u8(dst8, src8);
+
+            uint16x8_t result_low = vmovl_u8(vget_low_u8(result));
+            uint16x8_t result_high = vmovl_u8(vget_high_u8(result));
+
+            uint16x8_t dst_low = vmovl_u8(vget_low_u8(dst8));
+            uint16x8_t dst_high = vmovl_u8(vget_high_u8(dst8));
+
+            result_low = qvinterpolate_pixel_255(result_low, constAlphaVector, dst_low, oneMinusconstAlphaVector, half);
+            result_high = qvinterpolate_pixel_255(result_high, constAlphaVector, dst_high, oneMinusconstAlphaVector, half);
+
+            const uint32x2_t result32_low = vreinterpret_u32_u8(vmovn_u16(result_low));
+            const uint32x2_t result32_high = vreinterpret_u32_u8(vmovn_u16(result_high));
+            vst1q_u32((uint32_t *)&dst[x], vcombine_u32(result32_low, result32_high));
+        }
+
+        for (; x < length; ++x)
+            dst[x] = comp_func_Plus_one_pixel_const_alpha(dst[x], src[x], const_alpha, one_minus_const_alpha);
+    }
+}
+
 static const int tileSize = 32;
 
 extern "C" void qt_rotate90_16_neon(quint16 *dst, const quint16 *src, int sstride, int dstride, int count);
diff --git a/src/gui/painting/qdrawhelper_neon_p.h b/src/gui/painting/qdrawhelper_neon_p.h
index cd2dbfc..451edbc 100644
--- a/src/gui/painting/qdrawhelper_neon_p.h
+++ b/src/gui/painting/qdrawhelper_neon_p.h
@@ -120,6 +120,7 @@ void qt_transform_image_rgb16_on_rgb16_neon(uchar *destPixels, int dbpl,
                                             const QTransform &targetRectTransform,
                                             int const_alpha);
 
+void qt_memfill32_neon(quint32 *dest, quint32 value, int count);
 void qt_memrotate90_16_neon(const uchar *srcPixels, int w, int h, int sbpl, uchar *destPixels, int dbpl);
 void qt_memrotate270_16_neon(const uchar *srcPixels, int w, int h, int sbpl, uchar *destPixels, int dbpl);
 
@@ -131,6 +132,7 @@ void QT_FASTCALL qt_destStoreRGB16_neon(QRasterBuffer *rasterBuffer,
                                         int x, int y, const uint *buffer, int length);
 
 void QT_FASTCALL comp_func_solid_SourceOver_neon(uint *destPixels, int length, uint color, uint const_alpha);
+void QT_FASTCALL comp_func_Plus_neon(uint *dst, const uint *src, int length, uint const_alpha);
 
 #endif // QT_HAVE_NEON
 
diff --git a/src/gui/painting/qdrawhelper_p.h b/src/gui/painting/qdrawhelper_p.h
index d04c70d..75f42a0 100644
--- a/src/gui/painting/qdrawhelper_p.h
+++ b/src/gui/painting/qdrawhelper_p.h
@@ -1944,6 +1944,30 @@ const uint qt_bayer_matrix[16][16] = {
     ((((argb >> 24) * alpha) >> 8) << 24) | (argb & 0x00ffffff)
 
 
+#if QT_POINTER_SIZE == 8 // 64-bit versions
+#define AMIX(mask) (qMin(((qint64(s)&mask) + (qint64(d)&mask)), qint64(mask)))
+#define MIX(mask) (qMin(((qint64(s)&mask) + (qint64(d)&mask)), qint64(mask)))
+#else // 32 bits
+// The mask for alpha can overflow over 32 bits
+#define AMIX(mask) quint32(qMin(((qint64(s)&mask) + (qint64(d)&mask)), qint64(mask)))
+#define MIX(mask) (qMin(((quint32(s)&mask) + (quint32(d)&mask)), quint32(mask)))
+#endif
+
+inline int comp_func_Plus_one_pixel_const_alpha(uint d, const uint s, const uint const_alpha, const uint one_minus_const_alpha)
+{
+    const int result = (AMIX(AMASK) | MIX(RMASK) | MIX(GMASK) | MIX(BMASK));
+    return INTERPOLATE_PIXEL_255(result, const_alpha, d, one_minus_const_alpha);
+}
+
+inline int comp_func_Plus_one_pixel(uint d, const uint s)
+{
+    const int result = (AMIX(AMASK) | MIX(RMASK) | MIX(GMASK) | MIX(BMASK));
+    return result;
+}
+
+#undef MIX
+#undef AMIX
+
 // prototypes of all the composition functions
 void QT_FASTCALL comp_func_SourceOver(uint *dest, const uint *src, int length, uint const_alpha);
 void QT_FASTCALL comp_func_DestinationOver(uint *dest, const uint *src, int length, uint const_alpha);
diff --git a/src/gui/painting/qdrawhelper_sse2.cpp b/src/gui/painting/qdrawhelper_sse2.cpp
index 22c0384..30454af 100644
--- a/src/gui/painting/qdrawhelper_sse2.cpp
+++ b/src/gui/painting/qdrawhelper_sse2.cpp
@@ -161,22 +161,6 @@ void QT_FASTCALL comp_func_SourceOver_sse2(uint *destPixels, const uint *srcPixe
     }
 }
 
-inline int comp_func_Plus_one_pixel_const_alpha(uint d, const uint s, const uint const_alpha, const uint one_minus_const_alpha)
-{
-#define MIX(mask) (qMin(((qint64(s)&mask) + (qint64(d)&mask)), qint64(mask)))
-    const int result = (MIX(AMASK) | MIX(RMASK) | MIX(GMASK) | MIX(BMASK));
-#undef MIX
-    return INTERPOLATE_PIXEL_255(result, const_alpha, d, one_minus_const_alpha);
-}
-
-inline int comp_func_Plus_one_pixel(uint d, const uint s)
-{
-#define MIX(mask) (qMin(((qint64(s)&mask) + (qint64(d)&mask)), qint64(mask)))
-    const int result = (MIX(AMASK) | MIX(RMASK) | MIX(GMASK) | MIX(BMASK));
-#undef MIX
-    return result;
-}
-
 void QT_FASTCALL comp_func_Plus_sse2(uint *dst, const uint *src, int length, uint const_alpha)
 {
     int x = 0;
diff --git a/src/sql/drivers/odbc/qsql_odbc.cpp b/src/sql/drivers/odbc/qsql_odbc.cpp
index 9a35ac5..66dabfa 100644
--- a/src/sql/drivers/odbc/qsql_odbc.cpp
+++ b/src/sql/drivers/odbc/qsql_odbc.cpp
@@ -2106,7 +2106,7 @@ void QODBCDriverPrivate::checkSqlServer()
         serverType = QString::fromUtf8((const char *)serverString.constData(), t);
 #endif
         isFreeTDSDriver = serverType.contains(QLatin1String("tdsodbc"), Qt::CaseInsensitive);
-        unicode = isFreeTDSDriver == false;
+        unicode = unicode && !isFreeTDSDriver;
     }
 }