Fix bug in cubic-clipper (SkEdgeClipper). When we chop the cubic on Top/Bottom

of the cliprect, we (correctly) clamp the Y coordinate of the control-point
right next to the on-curve point that was chopped (this ensures we don't go
slightly outside of the clip-rect due to imperfect T value calculation). However,
the code was also clamping the other control-point as well, resulting in warping
the cubic, which could sometimes force it outside of the clip.

The fix is to just remove the line of code that clampped the 2nd control-point.

unittest added to reproduce a test cubic that triggered an assert, due to the
cubic being outside of the cliprect. The test (w/o the fix) will assert in 
a SK_DEBUG build.
Review URL: https://codereview.appspot.com/7100056

git-svn-id: http://skia.googlecode.com/svn/trunk@7184 2bbb7eff-a529-9590-31e7-b0007b416f81

1 files changed, 40 insertions, 0 deletions

diff --git a/tests/PathTest.cpp b/tests/PathTest.cpp
index a83a0a4fc6..11829d4d49 100644
--- a/tests/PathTest.cpp
+++ b/tests/PathTest.cpp
@@ -52,6 +52,45 @@ static void test_addrect_isfinite(skiatest::Reporter* reporter) {
     REPORTER_ASSERT(reporter, path.isFinite());
 }
 
+static void build_big_path(SkPath* path, bool reducedCase) {
+    if (reducedCase) {
+        path->moveTo(577330, 1971.72f);
+        path->cubicTo(10.7082f, -116.596f, 262.057f, 45.6468f, 294.694f, 1.96237f);
+    } else {
+        path->moveTo(60.1631f, 7.70567f);
+        path->quadTo(60.1631f, 7.70567f, 0.99474f, 0.901199f);
+        path->lineTo(577379, 1977.77f);
+        path->quadTo(577364, 1979.57f, 577325, 1980.26f);
+        path->quadTo(577286, 1980.95f, 577245, 1980.13f);
+        path->quadTo(577205, 1979.3f, 577187, 1977.45f);
+        path->quadTo(577168, 1975.6f, 577183, 1973.8f);
+        path->quadTo(577198, 1972, 577238, 1971.31f);
+        path->quadTo(577277, 1970.62f, 577317, 1971.45f);
+        path->quadTo(577330, 1971.72f, 577341, 1972.11f);
+        path->cubicTo(10.7082f, -116.596f, 262.057f, 45.6468f, 294.694f, 1.96237f);
+        path->moveTo(306.718f, -32.912f);
+        path->cubicTo(30.531f, 10.0005f, 1502.47f, 13.2804f, 84.3088f, 9.99601f);
+    }
+}
+
+static void test_clipped_cubic(skiatest::Reporter* reporter) {
+    SkAutoTUnref<SkSurface> surface(new_surface(640, 480));
+
+    // This path used to assert, because our cubic-chopping code incorrectly
+    // moved control points after the chop. This test should be run in SK_DEBUG
+    // mode to ensure that we no long assert.
+    SkPath path;
+    for (int doReducedCase = 0; doReducedCase <= 1; ++doReducedCase) {
+        build_big_path(&path, SkToBool(doReducedCase));
+        
+        SkPaint paint;
+        for (int doAA = 0; doAA <= 1; ++doAA) {
+            paint.setAntiAlias(SkToBool(doAA));
+            surface->getCanvas()->drawPath(path, paint);
+        }
+    }
+}
+
 // Inspired by http://ie.microsoft.com/testdrive/Performance/Chalkboard/
 // which triggered an assert, from a tricky cubic. This test replicates that
 // example, so we can ensure that we handle it (in SkEdge.cpp), and don't
@@ -2214,6 +2253,7 @@ static void TestPath(skiatest::Reporter* reporter) {
     test_arb_round_rect_is_convex(reporter);
     test_arb_zero_rad_round_rect_is_rect(reporter);
     test_addrect_isfinite(reporter);
+    test_clipped_cubic(reporter);
 }
 
 #include "TestClassDef.h"