Possible filter2D bug (Bug #2576)

The function cv::filter2D has a bug related to its memory management. I am working on OS X 10.6.8, using opencv2.4.0.

The execution of this code:

for(i=0; i < n_iter; i++) {
        std::cout << "Normalize_a, iteration " << i << std::endl;

        Utils::print_info(tmp, "TMP");

        cv::Mat tmp2(tmp.rows, tmp.cols, tmp.type());
        Utils::print_info(tmp2, "TMP2");

        Utils::print_info(flipped_dog, "Flipped dog");     

        std::cout << "Applying filter 2D" << std::endl;
        std::cout << anchor << std::endl;

        cv::filter2D(tmp, tmp2, -1, flipped_dog, anchor, 0, border_mode);
        std::cout << "Filter 2D applied" << std::endl;

        tmp = tmp + tmp2 - 0.01;
        cv::threshold(tmp, tmp, 0, 0, cv::THRESH_TOZERO);
    } // for i, iterations

according to the image (tmp, tmp2) size, results in this output

Normalize attn, iteration 0
Printing info on cv::Mat TMP
     rows: 157, cols: 250
     min - max: 0 - 1
Printing info on cv::Mat TMP2
     rows: 157, cols: 250
     min - max: 0.0217404 - 1
Printing info on cv::Mat Flipped dog
     rows: 79, cols: 79
     min - max: -5.70341e-05 - 0.00397742
Applying filter 2D
[39, 39]

followed by

va_main(97201) malloc: *** error for object 0x1010ca200: incorrect checksum for freed object - object was probably modified after being freed.
*** set a breakpoint in malloc_error_break to debug
Abort trap

I will try and modify the code (it is not the best you might desire), but it is perfectly legal (isn't it?).

Regards