In certain cases, the coordinates used for pattern rendering can go

negative, eg., when a negative translation is applied in the shader
matrix.  This causes the rasterizer to blow up, since it accesses
memory outside the pattern bitmap.

Since the integer modulus operator for C++ has unspecified behaviour with
negative arguments, its value may go negative.  In this case, we must offset
by the modulus to make it positive again.  I decided to do this at a low level
in skia, since it seems better to make skia robust than to pray that you get
strictly non-negative translations.

(There is a fix for this in src/skia/tile_patch.diff, but it's wrong.)

This patch should be upstreamed to skia.

Review URL: http://codereview.chromium.org/55044

git-svn-id: svn://svn.chromium.org/chrome/trunk/src@12789 0039d316-1c4b-4281-b951-d872f2087c98

1 files changed, 3 insertions, 2 deletions

diff --git a/skia/sgl/SkBitmapProcState_matrixProcs.cpp b/skia/sgl/SkBitmapProcState_matrixProcs.cpp
index a16c8b3..c3787a0 100644
--- a/skia/sgl/SkBitmapProcState_matrixProcs.cpp
+++ b/skia/sgl/SkBitmapProcState_matrixProcs.cpp
@@ -37,8 +37,9 @@ void decal_filter_scale(uint32_t dst[], SkFixed fx, SkFixed dx, int count);
 #define TILEY_PROCF(fy, max)    (((fy) & 0xFFFF) * ((max) + 1) >> 16)
 #define TILEX_LOW_BITS(fx, max) ((((fx) & 0xFFFF) * ((max) + 1) >> 12) & 0xF)
 #define TILEY_LOW_BITS(fy, max) ((((fy) & 0xFFFF) * ((max) + 1) >> 12) & 0xF)
-#define TILEX_TRANS(x, max)     ((x) % ((max) + 1))
-#define TILEY_TRANS(y, max)     ((y) % ((max) + 1))
+#define SK_MOD(a, b)            (((a)%(b)) < 0 ? ((a)%(b) + (b)) : (a)%(b))
+#define TILEX_TRANS(x, max)     (SK_MOD((x), ((max) + 1)))
+#define TILEY_TRANS(y, max)     (SK_MOD((y), ((max) + 1)))
 #include "SkBitmapProcState_matrix.h"
 
 #define MAKENAME(suffix)        GeneralXY ## suffix