https://stackoverflow.com/questions/13850678
italiano
english
français
española
中国
日本の
العربية
Deutsch
한국어
Português
RussianYou have index overrun on filt_lines vector in innermost loop. I suppose it should be
for (size_t j = i + 1; j < filt_lines.size(); j++)
instead of
for (size_t j=i+1;j<p_lines.size();j++)
free(): invalid pointer for vector<vec4i> c++ [closed]
Question
my problem is that a std:vector<vec4i> ( p_lines and filt_lines) are created in a function which can not be deallocate at the end of the function. So i get an free(): invalid pointer error while running the code. In some cases i go segmentation fault error, but i couldt trace the source with gdb.
The aim of the function is to find the edges on a picture and there intersection point, in the region where the most lines are intersecting it is drown a rectangle. The pictures are coming from a ROS video feed.
void Probabilistic_Hough( Mat src, Mat &result )
{
result=src;
cvtColor( src, src, CV_RGB2GRAY );
Canny( src, src, 50, 200, 3 );
HoughLinesP( src, p_lines, 1, CV_PI/180, 85, 40, 10 );
vector<Vec4i> filt_lines;
int fls=filt_lines.size();
int pls=p_lines.size();
if (!p_lines.empty()){
for(size_t i = 0; i < p_lines.size();i++)
{
Vec4i l = p_lines[i];
if ((l[0]<l[2]-10 || l[0]>l[2]+10) && (l[1]<l[3]-10 || l[1]>l[3]+10))
{
line( result/*probabilistic_hough*/, Point(l[0], l[1]), Point(l[2], l[3]), Scalar(255,0,0), 2);
filt_lines.push_back(l);
}
}
fls=filt_lines.size();
pls=p_lines.size();
printf("f=%d p=%d",fls,pls);
if (filt_lines.size()>1 && !filt_lines.empty())
{
int sizeX=src.rows/11, sizeY=src.cols/11;
int imgGrid[11][11];
for (int i=0;i<11;i++)
{
for (int j=0;j<11;j++) imgGrid[i][j]=0;
}
for (size_t i = 0;i<filt_lines.size()-1;i++)
{
Point2f o1 (filt_lines[i][0], filt_lines[i][1]) ;
Point2f p1 (filt_lines[i][2], filt_lines[i][3]) ;
for (size_t j=i+1;j<filt_lines.size();j++)
{
Point2f o2 (filt_lines[j][0], filt_lines[j][1]) ;
Point2f p2 (filt_lines[j][2], filt_lines[j][3]) ;
Point2f x = o2 - o1;
Point2f d1 = p1 - o1;
Point2f d2 = p2 - o2;
float cross = d1.x*d2.y - d1.y*d2.x;
if (abs(cross) > 1e-8)
{
double t1 = (x.x * d2.y - x.y * d2.x)/cross;
Point2f crosP = o1 + d1 * t1;
if (crosP.x>0.0 && crosP.y>0.0 && crosP.y< src.rows && crosP.x<src.cols)
imgGrid[int(crosP.x/sizeX)][int(crosP.y/sizeY)]++;
}
}
}
int countP=0,maxi=0, maxj=0;
for (int i=0;i<11;i++)
{
for (int j=0;j<11;j++)
{
if (countP<imgGrid[i][j])
{
countP=imgGrid[i][j];
maxi=i;
maxj=j;
}
}
}
printf("c=%d \n",countP);
if (countP>0)
{
Point vanishP1 = cvPoint(maxi*sizeX,maxj*sizeY), vanishP2 = cvPoint((maxi+1)*sizeX,(maxj+1)*sizeY);
CvScalar red= CV_RGB(220,0,0);
if (vanishP1.x>=0 && vanishP1.y>=0 && vanishP2.x>0 && vanishP2.y>0)
rectangle(src,vanishP1,vanishP2,red,2,8,0);
result=src;
} else {printf("only 1 intersection, out of img"); }
} else
{
printf("No filtered lines");
}
} else
{
printf("no lines on image");
}
}
I tired to debug with gdb and the invalid pointer address is the beginning address of the p_lines vector. The gdb frame points to the las } of the function.
How can I solve the free(): invalid pointer error and what could be the problem that cause segmentation fault.
The error msg from gdb:
glibc detected *** /home/xyz: free(): invalid pointer: 0x080cc482 ***
======= Backtrace: =========
/lib/tls/i686/cmov/libc.so.6(+0x6b591)[0x12fe591]
/lib/tls/i686/cmov/libc.so.6(+0x6cde8)[0x12ffde8]
/lib/tls/i686/cmov/libc.so.6(cfree+0x6d)[0x1302ecd]
/usr/lib/libstdc++.so.6(_ZdlPv+0x21)[0x11fa741]
/home/xyz(_Z19Probabilistic_HoughN2cv3MatERS0_+0x8c5)[0x804b3b5]
/home/xyz(_Z13imageCallbackRKN5boost10shared_ptrIKN11sensor_msgs6Image_ISaIvEEEEE+0x5c8)[0x804b9a0][0x804b3b5]
/home/xyz(_Z13imageCallbackRKN5boost10shared_ptrIKN11sensor_msgs6Image_ISaIvEEEEE+0x5c8)[0x804b9a0]
and the print of p_lines is(_M_start = 0x80cc482 is equal with free(): i. p. :0x080cc482):
(gdb) p p_lines
$1 = {<std::_Vector_base<cv::Vec<int, 4>, std::allocator<cv::Vec<int, 4> > >> = {
_M_impl = {<std::allocator<cv::Vec<int, 4> >> = {<__gnu_cxx::new_allocator<cv::Vec<int, 4> >> = {<No data fields>}, <No data fields>}, _M_start = 0x80cc482, _M_finish = 0x80cc540,
_M_end_of_storage = 0x80cc540}}, <No data fields>}
I remade some test after correcting my typo and added an if (countP>0) at the end. Found the frame where the invalid pointer operation appears, img has 12 p_lines and 2 filt_lines, however countP=0 (so the two line intersects out of the img size).
The gdb showed that after calling HoughLinesP() the starting addres of p_lines is _M_start = 0x80d4e70, after the crash the the free(): i.p. appeared at 0x080d4e71 and in my functions gdb frame the staring address of p_lines is _M_start = 0x80d4e71.
What the frack? How is it possible that the starting addres of the vector changes with 1 bit?
Finally I found the exact moment when the bit change in the p_lines happens:
(gdb) next
98 imgGrid[int(crosP.x/sizeX)][int(crosP.y/sizeY)]++;
(gdb) p p_lines
$68 = {<std::_Vector_base<cv::Vec<int, 4>, std::allocator<cv::Vec<int, 4> > >> = {
_M_impl = {<std::allocator<cv::Vec<int, 4> >> = {<__gnu_cxx::new_allocator<cv::Vec<int, 4> >> = {<No data fields>}, <No data fields>},
_M_start = 0x80d4c78, _M_finish = 0x80d4d38, _M_end_of_storage = 0x80d4d38}}, <No data fields>}
(gdb) next
84 for (size_t j=i+1;j<filt_lines.size();j++)
(gdb) p p_lines
$69 = {<std::_Vector_base<cv::Vec<int, 4>, std::allocator<cv::Vec<int, 4> > >> = {
_M_impl = {<std::allocator<cv::Vec<int, 4> >> = {<__gnu_cxx::new_allocator<cv::Vec<int, 4> >> = {<No data fields>}, <No data fields>},
_M_start = 0x80d4c79, _M_finish = 0x80d4d38, _M_end_of_storage = 0x80d4d38}}, <No data fields>}
(gdb) p j
$71 = 1
So in this case the size of filt_lines is 2. and and the bit shift appears after the 1st line and 2nd are checked and the jumps back to the for when j=1 and j<2. Why? :/
Pleas help me out, solving this error. thx
Solution
Licensed under: CC-BY-SA with attribution
Not affiliated with StackOverflow
