C++: Allocating memory in OpenGL drawing loop causes memory corruption

Question

I have an Open GL ES main loop, which calls 2 functions; draw and update, once each per loop.

The following is the code from the draw loop:

float* sample_data = userdata->sample_data;
int fft_length = userdata->fft_length;

cout << 'a' << endl;
cout << fft_length << endl;

GLfloat* points = new GLfloat[3 * fft_length];

cout << 'b' << endl;

// Draw tingz
for(int ix = 0; ix < fft_length; ++ ix)
{
    float ratio = float(ix) / float(fft_length - 1);
    float x = -1.0 + 2.0 * ratio;
    float y = sample_data[ix];

    points[3 * ix + 0] = x;
    points[3 * ix + 1] = y;
    points[3 * ix + 2] = 0.0;
}

glVertexAttribPointer(0, 3, GL_FLOAT, GL_FALSE, 0, points);
glEnableVertexAttribArray(0);

glDrawArrays(GL_LINE_STRIP, 0, fft_length);

std::cout << 'd' << std::endl;
delete [] points;
std::cout << 'e' << std::endl;

// End of function

The output prints everything once, but then only prints a and 256, which is the length of the fft.

Hence I assume it is the call to 'new' where the problem occurs. The next line of text on the screen which appears is:

*** glibc detected *** ./main.out: malloc(): memory corruption: 0x01319648 ***

This puzzles me, as I can't immediately see anything wrong. I allocate memory and delete it again. My guess is I made a stupid mistake which I just can't spot?

do you typecast with `float(ix)`? why not with `(float) ix`? — mch, Aug 21 '14 at 15:33
@mch I wasn't aware that there was a difference? Except that the second works with C compilers. — FreelanceConsultant, Aug 21 '14 at 15:34
@ratchetfreak That sounds like it might be the reason for the error. Which buffer are you talking about and what do you think the missing code is? — FreelanceConsultant, Aug 21 '14 at 15:41
add a `glBindBuffer(GL_ARRAY_BUFFER, 0);` before the `glVertexAttribPointer` call — ratchet freak, Aug 21 '14 at 15:48
@ratchetfreak Hmm, tried that but no difference I'm afraid... Unfortunately I don't fully understand the process through with OpenGL allocates memory, copies data and uses it for drawing before releasing again, so I can't know for sure which calls are required and where. — FreelanceConsultant, Aug 21 '14 at 15:53
"... the second works with C compilers." Inarguable, but certainly `new GLfloat[3 * fft_length]` does *not*. Further, there is no reason I can see not to use `std::vector points(3*fft_length);` and belay managing the memory to RAII instead, sending `points.data()` where needed.. — WhozCraig, Aug 21 '14 at 17:02
@WhozCraig Okay but I didn't do it like that, so although it would arguably be better that way it isn't. — FreelanceConsultant, Aug 21 '14 at 17:48
i would use a `thread_local static std::vector` The opengles implementation might not be standards compliant in that it doesn't read the memory right away. more probable is that you have a bug outside of this code. — programmerjake, Aug 21 '14 at 20:43
@programmerjake Could be, I'll report anything I find tomorrow — FreelanceConsultant, Aug 21 '14 at 22:03
@user3728501 Christian's [**answer here**](http://stackoverflow.com/questions/17149728/when-should-glvertexattribpointer-be-called) has a good sequence of how/when to invoke that, and under what conditions. Worth a look. — WhozCraig, Aug 23 '14 at 01:27

C++: Allocating memory in OpenGL drawing loop causes memory corruption

0 Answers0