Fix PNG decoding divide by zero error due to zero row count.

core/src/fxcodec/codec/fx_codec_flate.cpp

Index: core/src/fxcodec/codec/fx_codec_flate.cpp
diff --git a/core/src/fxcodec/codec/fx_codec_flate.cpp b/core/src/fxcodec/codec/fx_codec_flate.cpp
index e8878e453524d58e28c991315a9189c3582d8cdb..37aecf106d2e8bde56c874ab25abde5c6a0316b4 100644
--- a/core/src/fxcodec/codec/fx_codec_flate.cpp
+++ b/core/src/fxcodec/codec/fx_codec_flate.cpp
@@ -400,6 +400,8 @@ static FX_BOOL PNG_Predictor(uint8_t*& data_buf,
   if (row_size <= 0)
     return FALSE;
   const int row_count = (data_size + row_size) / (row_size + 1);
+  if (row_count <= 0)

[Tom Sepez, 2015/10/02 16:44:25]

Yeah, ok, but this gets better using only unsigned types and safe types for
overflow handling. I'm sure there's some lurking grief if you can call this with
arbitrary arguments.

[Lei Zhang, 2015/10/02 17:29:31]

Acknowledged.

+    return FALSE;
   const int last_row_size = data_size % (row_size + 1);
   uint8_t* dest_buf = FX_Alloc2D(uint8_t, row_size, row_count);
   int byte_cnt = 0;