Unable to access/modify desired pointer value after calling a library function

**xy127** · 07-09-2016

I'm using an external library https://people.sc.fsu.edu/~jburkardt/c_src/jacobi_eigenvalue/jacobi_eigenvalue.c to calculate eigenvalue of a matrix. However, after calling the library function, I received Segmentation fault when I tried to access desired pointer values.

Code:

#define M 3
double Gyration[M*M] = { 
          gyration[0][0],gyration[0][1],gyration[0][2], 
          gyration[1][0],gyration[1][1],gyration[1][2],
          gyration[2][0],gyration[2][1],gyration[2][2] };

jacobi_eigenvalue(M, Gyration, it_max, v, d, &it_num, &rot_num);
rog = sqrt(d[0]+d[1]+d[2]);  
*Denominator= 2*pow(rog,4); 
*Nominator= pow((d[2]-d[1]),2)+pow((d[2]-d[0]),2)+pow((d[1]-d[0]),2);
* Asp = (*Nominator)/(*Denominator);
# undef M

Denominator, Nominator, and Asp were defined as double variables in main function and passed as pointers. However I received segmentation fault when I was trying to access them after library function call.
I guess the issue is dereferencing a pointer that has been freed since in the library function, malloc is used. But, I am not sure where exactly the issue is and how to fix it.
Thank you in advance.

**laserlight** · 07-09-2016

It looks like Denominator, Nominator, and Asp have nothing to do with the library function call, so perhaps the mistake lies elsewhere. I suggest that you post the smallest and simplest compilable program that demonstrates the error, e.g., a main function to setup the values and call your function, and then your function that calls the library function.

**xy127** · 07-09-2016

That's correct that these three variables have no relation with the library function. But I am able to access *Denominator, *Nominator, and *Asp (which are 0) before the function call.

**WoodSTokk** · 07-09-2016

Originally Posted by xy127

I'm using an external library https://people.sc.fsu.edu/~jburkardt/c_src/jacobi_eigenvalue/jacobi_eigenvalue.c to calculate eigenvalue of a matrix. However, after calling the library function, I received Segmentation fault when I tried to access desired pointer values.

Code:

#define M 3
double Gyration[M*M] = { 
          gyration[0][0],gyration[0][1],gyration[0][2], 
          gyration[1][0],gyration[1][1],gyration[1][2],
          gyration[2][0],gyration[2][1],gyration[2][2] };

jacobi_eigenvalue(M, Gyration, it_max, v, d, &it_num, &rot_num);
rog = sqrt(d[0]+d[1]+d[2]);  
*Denominator= 2*pow(rog,4); 
*Nominator= pow((d[2]-d[1]),2)+pow((d[2]-d[0]),2)+pow((d[1]-d[0]),2);
* Asp = (*Nominator)/(*Denominator);
# undef M

Denominator, Nominator, and Asp were defined as double variables in main function and passed as pointers. However I received segmentation fault when I was trying to access them after library function call.

What i have found in the internet is this declaration:

Code:

void jacobi_eigenvalue (int n, double a[], int it_max, double v[], double d[], int &it_num, int &rot_num);

Edit: Ooops, i see you have post the link to the library. I think I need more coffee.
This give me a few questions.
From where come your variable 'it_max' and is it of type int?
From where come your variable 'v' and is it an array of type double (and has the array the right size)?
From where come your variable 'd' and is it an array of type double (and has the array the right size)?
From where come your variable 'it_num' and is it of type int?
From where come your variable 'rot_num' and is it of type int?

You can output all values with printf before you call the function.

Originally Posted by xy127

I guess the issue is dereferencing a pointer that has been freed since in the library function, malloc is used. But, I am not sure where exactly the issue is and how to fix it.
Thank you in advance.

No! A function will never free a pointer. If you write functions, also never free pointers. The function doesn't know if the pointer point to a malloc'ed memory or to a local variable or struct.

**xy127** · 07-09-2016

1. I defined It_max(max number of iterations) as int it_max=100 in my own function(not the library function) which is not shown above.
2. 'v' is a double type vector that stores eigenvectors. I defined 'v' as double v[M] where M=3 in my function.
3. 'd' is a double type vector that stores eigenvalues. I defined d as double d[M]
4. it_num and rot_num are both defined as integer type in my function
5. There shows no error when gdb passes the library function. And I was able to access v[] and d[] ( The output of library function)

**Salem** · 07-10-2016

> 2. 'v' is a double type vector that stores eigenvectors. I defined 'v' as double v[M] where M=3 in my function.
Whereas the comment in the library code reads

> Output, double V[N*N], the matrix of eigenvectors.

It seems to me that your supplied v array is far too small, and the library scribbles over memory it shouldn't.
The result is typically a segfault, when the real owner of the memory finds it's been trashed.

**laserlight** · 07-10-2016

Originally Posted by xy127

That's correct that these three variables have no relation with the library function. But I am able to access *Denominator, *Nominator, and *Asp (which are 0) before the function call.

That does not mean that the problem lies with the library function call. The problem could lie with your code, e.g., code that you did not show that results in undefined behaviour, but only surfaced after the library function call was made. This is why I suggested that you come up with the smallest and simplest compilable program that demonstrates the error: in the process of doing so, you may discover the problem, and if you don't, you would have ruled out code that is irrelevant to the problem.

Even if you don't want to do what I recommended for whatever reason, at the very least you could try:

Code:

#define M 3
double Gyration[M*M] = { 
          gyration[0][0],gyration[0][1],gyration[0][2], 
          gyration[1][0],gyration[1][1],gyration[1][2],
          gyration[2][0],gyration[2][1],gyration[2][2] };
 
/*jacobi_eigenvalue(M, Gyration, it_max, v, d, &it_num, &rot_num);*/
printf("%f %f %f\n", *Denominator, *Nominator, *Asp);
# undef M

Does the above result in zeroes being printed to standard output? If so, uncomment the previous line. What happens? If you then get a segmentation fault, then you need to do what I recommended, because then all I can tell you is that either the mistake lies in your code (that I cannot see because you did not provide it) or the mistake lies in jacobi_eigenvalue, but even if I were interested in doing so, I am not about to spend time to try and debug jacobi_eigenvalue when the mistake could reasonably lie in your code.

**xy127** · 07-10-2016

Originally Posted by Salem

> 2. 'v' is a double type vector that stores eigenvectors. I defined 'v' as double v[M] where M=3 in my function.
Whereas the comment in the library code reads

> Output, double V[N*N], the matrix of eigenvectors.

It seems to me that your supplied v array is far too small, and the library scribbles over memory it shouldn't.
The result is typically a segfault, when the real owner of the memory finds it's been trashed.

Sorry, I had a typo in previous reply. 'v' is defined as double v[M*M]. 'v' has same size as gyration matrix

**Salem** · 07-10-2016

Originally Posted by xy127

Sorry, I had a typo in previous reply. 'v' is defined as double v[M*M]. 'v' has same size as gyration matrix

So how do we know there aren't any other "typos" masquerading as facts in your posts?

Short, Self Contained, Correct Example

**xy127** · 07-10-2016

Originally Posted by Salem

So how do we know there aren't any other "typos" masquerading as facts in your posts?

Short, Self Contained, Correct Example

I apologize for that mistake. But if it was v[M], then error will occur inside library function since the input is not in correct form.

**laserlight** · 07-10-2016

No worries. So, where's the code?

**xy127** · 07-10-2016

Thanks !

I'm posting the entire function. The code itself contains at least 20 functions. This function is independent with others. So I am not going to post all the code.

Code:

void asphericity(particle *p, int N, float aveLTail[MAX2][MAX2], double &Nominator, double &Denominator, double *Asp ,double *cmx, double *cmy, double *cmz){
    int i, j, k,m,n,M;
    double x[10000];
    double y[10000];
    double z[10000];
    double *Q;
    double gyration[3][3];
    double v[M*M];
    double d[M];
    float ut, up;
    float bs;
    double hdistance;
    double tdistance;
    double rog; 
    int hcount=0;
    int tcount=0;
    int it_max;
    int it_num;
    int rot_num;
    double hx=0;
    double hy=0;
    double hz=0;
    double tx=0;
    double ty=0;
    double tz=0;
    m=0;
    Nominator=NULL;
    Denominator=NULL;
   
    it_max=100;
    for(i = 1; i<=N; i++) {
       if (p[i].mol == p[i-1].mol){
           if (p[i-1].type==1 ||p[i-1].type==2 || p[i-1].type==3 || p[i-1].type==6){
                hcount++;
                hx += p[i-1].x;
                hy += p[i-1].y;
                hz += p[i-1].z;
           }
           else if (p[i-1].type==4 && p[i-2].type==9){
                tcount++;
                tx += p[i-1].x;
                ty += p[i-1].y;
                tz += p[i-1].z;
                hcount += 2;
                hx += p[i-2].x+p[i-3].x;
                hy += p[i-2].y+p[i-3].y;
                hz += p[i-2].z+p[i-3].z;
            }  
           else if (p[i-1].type==4 && p[i-2].type!=9){
                tcount++;
                tx += p[i-1].x;
                ty += p[i-1].y;
                tz += p[i-1].z;
           }
                
       }
       else if (p[i].mol != p[i-1].mol && p[i-1].mol!=0) {
                tcount++;
                tx += p[i-1].x;
                ty += p[i-1].y;
                tz += p[i-1].z;
               (hx) /= hcount; /*geometric center of head beads*/
               (hy) /= hcount;
               (hz) /= hcount;
               (tx) /= tcount;
               (ty) /= tcount;
               (tz) /= tcount;
               hcount=0;
               tcount=0;
               hdistance=sqrt(pow((hx-*cmx),2)+pow((hy-*cmy),2)+pow((hz-*cmz),2));
               tdistance=sqrt(pow((tx-*cmx),2)+pow((ty-*cmy),2)+pow((tz-*cmz),2));
               if(hdistance>tdistance){
                  x[m]=hx;
                  y[m]=hy;
                  z[m]=hz;
                  m++;
                 }
               hx=0;
               hy=0;
               hz=0;
               tx=0;
               ty=0;
               tz=0;
             }              
       } 
    

    for(j = 0; j<m; j++) {
        gyration[0][0]+= pow((x[j]-*cmx),2);
        gyration[0][1]+=(x[j]-*cmx)*(y[j]-*cmy);
        gyration[0][2]+=(x[j]-*cmx)*(z[j]-*cmz); 
        gyration[1][0]+=(y[j]-*cmy)*(x[j]-*cmx);
        gyration[1][1]+=pow((y[j]-*cmy),2);
        gyration[1][2]+=(y[j]-*cmy)*(z[j]-*cmz);
        gyration[2][0]+=(z[j]-*cmz)*(x[j]-*cmx);
        gyration[2][1]+=(z[j]-*cmz)*(y[j]-*cmy);
        gyration[2][2]+=pow((z[j]-*cmz),2);
     }
    for(i =0; i<3; i++){
       for(j=0; j<3; j++){
           gyration[i][j]/=m;
          }
    }

 #define M 3
    double Gyration[M*M] = { 
              gyration[0][0],gyration[0][1],gyration[0][2], 
              gyration[1][0],gyration[1][1],gyration[1][2],
              gyration[2][0],gyration[2][1],gyration[2][2] };

    jacobi_eigenvalue(M, Gyration, it_max, v, d, &it_num, &rot_num);
     double *Nominator;
     double *Denominator;
    rog = sqrt(d[0]+d[1]+d[2]);  
    *Denominator= 2*pow(rog,4); 
    *Nominator= pow((d[2]-d[1]),2)+pow((d[2]-d[0]),2)+pow((d[1]-d[0]),2);
    * Asp = (*Nominator)/(*Denominator);
# undef M
 }

**xy127** · 07-10-2016

Originally Posted by laserlight

No worries. So, where's the code?

Please ignore the previous code. I made some modification based on your suggestion. The first printf prints all three variables 0. But Segmentation fault appears at send printf. So, something is wrong with the library function.

Code:

void asphericity(particle *p, int N, float aveLTail[MAX2][MAX2], double *Nominator, double *Denominator, double *Asp ,double *cmx, double *cmy, double *cmz){
    int i, j, k,m,n,M;
    double x[10000];
    double y[10000];
    double z[10000];
    double *Q;
    double gyration[3][3];
    double v[M*M];
    double d[M];
    float ut, up;
    float bs;
    double hdistance;
    double tdistance;
    double rog; 
    int hcount=0;
    int tcount=0;
    int it_max;
    int it_num;
    int rot_num;
    double hx=0;
    double hy=0;
    double hz=0;
    double tx=0;
    double ty=0;
    double tz=0;
    m=0;

   
    it_max=100;
    for(i = 1; i<=N; i++) {
       if (p[i].mol == p[i-1].mol){
           if (p[i-1].type==1 ||p[i-1].type==2 || p[i-1].type==3 || p[i-1].type==6){
                hcount++;
                hx += p[i-1].x;
                hy += p[i-1].y;
                hz += p[i-1].z;
           }
           else if (p[i-1].type==4 && p[i-2].type==9){
                tcount++;
                tx += p[i-1].x;
                ty += p[i-1].y;
                tz += p[i-1].z;
                hcount += 2;
                hx += p[i-2].x+p[i-3].x;
                hy += p[i-2].y+p[i-3].y;
                hz += p[i-2].z+p[i-3].z;
            }  
           else if (p[i-1].type==4 && p[i-2].type!=9){
                tcount++;
                tx += p[i-1].x;
                ty += p[i-1].y;
                tz += p[i-1].z;
           }
                
       }
       else if (p[i].mol != p[i-1].mol && p[i-1].mol!=0) {
                tcount++;
                tx += p[i-1].x;
                ty += p[i-1].y;
                tz += p[i-1].z;
               (hx) /= hcount; /*geometric center of head beads*/
               (hy) /= hcount;
               (hz) /= hcount;
               (tx) /= tcount;
               (ty) /= tcount;
               (tz) /= tcount;
               hcount=0;
               tcount=0;
               hdistance=sqrt(pow((hx-*cmx),2)+pow((hy-*cmy),2)+pow((hz-*cmz),2));
               tdistance=sqrt(pow((tx-*cmx),2)+pow((ty-*cmy),2)+pow((tz-*cmz),2));
               if(hdistance>tdistance){
                  x[m]=hx;
                  y[m]=hy;
                  z[m]=hz;
                  m++;
                 }
               hx=0;
               hy=0;
               hz=0;
               tx=0;
               ty=0;
               tz=0;
             }              
       } 
    

    for(j = 0; j<m; j++) {
        gyration[0][0]+= pow((x[j]-*cmx),2);
        gyration[0][1]+=(x[j]-*cmx)*(y[j]-*cmy);
        gyration[0][2]+=(x[j]-*cmx)*(z[j]-*cmz); 
        gyration[1][0]+=(y[j]-*cmy)*(x[j]-*cmx);
        gyration[1][1]+=pow((y[j]-*cmy),2);
        gyration[1][2]+=(y[j]-*cmy)*(z[j]-*cmz);
        gyration[2][0]+=(z[j]-*cmz)*(x[j]-*cmx);
        gyration[2][1]+=(z[j]-*cmz)*(y[j]-*cmy);
        gyration[2][2]+=pow((z[j]-*cmz),2);
     }
    for(i =0; i<3; i++){
       for(j=0; j<3; j++){
           gyration[i][j]/=m;
          }
    }

 #define M 3
    double Gyration[M*M] = { 
              gyration[0][0],gyration[0][1],gyration[0][2], 
              gyration[1][0],gyration[1][1],gyration[1][2],
              gyration[2][0],gyration[2][1],gyration[2][2] };
    printf("%.6f\t%.6f\t%.6f\n", *Denominator, *Nominator, *Asp);
    jacobi_eigenvalue(M, Gyration, it_max, v, d, &it_num, &rot_num);
    printf("%.6f\t%.6f\t%.6f\n", *Denominator, *Nominator, *Asp);
    rog = sqrt(d[0]+d[1]+d[2]);  
    *Denominator= 2*pow(rog,4); 
    *Nominator= pow((d[2]-d[1]),2)+pow((d[2]-d[0]),2)+pow((d[1]-d[0]),2);
    * Asp = (*Nominator)/(*Denominator);
# undef M
 }

**WoodSTokk** · 07-10-2016

Thats a good example for using desciptive names for variables and defines.
Line 2:

Code:

    int i, j, k,m,n,M;

M is defined as an integer, but not inizialized.

Line 8+9:

Code:

    double v[M*M];
    double d[M];

v and d are defined as double vectors with a size of M, but M isn't initialized.
Nobody knows how big or small this vectors are.

Line 104+105:

Code:

#define M 3
    double Gyration[M*M] = {
…

M is now a define to represent 3.
And Gyration is a double vector in size of 9.
But have v and d the right size?

To fix this, set line 104 as the first line (in front of the function).
Delete the integer M on line 2.
To avoid confusion, rename M to SIZE_M. It makes it clear that this is a defined name.

Thread: Unable to access/modify desired pointer value after calling a library function

Thread Tools

Search Thread

Display

Unable to access/modify desired pointer value after calling a library function

Similar Threads

Trying to modify array in a shared library function

Calling Function via function pointer inside a structure pointer

sine function without use of math library using expansion and calling other functions

unable to access structure within a function

Calling C function from dynamically loaded library

Tags for this Thread