Why does the CONVENC function from Communications Toolbox 3.0.1 (R14SP1) fail when the message input is a non-double type?

When I execute the following code:

trellis = poly2trellis(7,[ 171 133 ]);

msg = uint16 ([1 0 1]);

convenc (msg, trellis)

I receive the following segmentation violation:

 ------------------------------------------------------------------------

        Segmentation violation detected at Tue Jan 25 13:56:13 2005

 ------------------------------------------------------------------------

 Configuration:

   MATLAB Version:   7.0.1.24704 (R14) Service Pack 1

   Operating System: Microsoft Windows XP

   Window System:    Version 5.1 (Build 2600: Service Pack 1)

   Processor ID:     x86 Family 15 Model 2 Stepping 4, GenuineIntel

   Virtual Machine:  Java 1.4.2_04 with Sun Microsystems Inc. Java HotSpot(TM) Client VM

     (mixed mode)

   Default Charset:  ibm-5348_P100-1997

 Register State:

   EAX = 00000010  EBX = 00000002

   ECX = 0e08b010  EDX = 00000000

   ESI = 000b0080  EDI = 00000001

   EBP = 00cdd7a4  ESP = 00cdd770

   EIP = 095e1123  FLG = 00010202

 Stack Trace:

   [0] convcore.dll:_ConvEncode(0x0d276e60 "àD'\r", 0, 0x00cdd7e0 "8ÜÍ", 0x787d4f97) + 291 bytes

   [1] convcore.dll:_mexFunction(2, 0x00cddee0, 7, 0x00cddf40) + 67 bytes

   [2] libmex.dll:_mexRunMexFile(2, 0x00cddee0, 7, 0x00cddf40) + 103 bytes

   [3] libmex.dll:public: virtual void __thiscall Mfh_mex::dispatch_file(int,struct mxArray_tag * *,int,struct mxArray_tag * *)(2, 0x00cddee0, 7, 0x00cddf40) + 157 bytes

   [4] m_dispatcher.dll:public: virtual void __thiscall Mfh_file::dispatch_fh(int,struct mxArray_tag * *,int,struct mxArray_tag * *)(2, 0x00cddee0, 7, 0x00cddf40) + 273 bytes

   [5] m_interpreter.dll:int __cdecl mdDispatch(int,char const *,int,struct mxArray_tag * *,int,struct mxArray_tag * *,class Mfh_MATLAB_fn * *)(1758, 0x0d9e39ac "convcore", 2, 0x00cddee0) + 88 bytes

 (snip)

However, when I do not typecast the msg input to uint16, the crash does not occur. Why?

Answer 1

[MathWorks Support Team]

We have verified that there is a bug in Communications Toolbox 3.0.1 (R14SP1) in the way that the CONVENC function handles non-double precision inputs. To work around this issue, typecast the input to double precision. For example,

trellis = poly2trellis(7,[ 171 133 ]);

msg = uint16 ([1 0 1]);

convenc (double(msg), trellis)

[Tasos Giannoulis]

It's been several releases that CONVENC works with non-double data types. So it is no longer needed to cast to double:

>> trellis = poly2trellis(7,[ 171 133 ]);
>> msg = uint16 ([1 0 1]);
>> convenc (msg, trellis)

ans =

    1×6 uint16 row vector

     1   1   1   0   0   0