Very hard to debug segfault, help (symbols) wanted

Gaming and Visualization Technologies APIs OpenGL
1

I have a bug which only shows when not running through valgrind.

Is it possible to sign an NDA or whatever and get access to a driver with symbols? So I have some kind of clue what’s going on…

Here’s the backtrace (from gdb):
#0 0x000000004009ea74 in ?? ()
#1 0x0000000040000000 in ?? ()
#2 0x00007fffe4f783b8 in ?? ()
#3 0x00007ffff229ffc8 in ?? ()
#4 0x00007ffff229ffc8 in ?? ()
#5 0x00007ffff4d8891e in ?? () from /usr/lib/libnvidia-glcore.so.310.40
#6 0x00007ffff4d8d948 in ?? () from /usr/lib/libnvidia-glcore.so.310.40
#7 0x00007ffff4a6dd04 in ?? () from /usr/lib/libnvidia-glcore.so.310.40
#8 … my code … calling glDrawElements(GL_TRIANGLES…) using VBOs

The code crashing runs when only drawing that mesh, so it’s not something local to that mesh.

Usually I catch 99% of bugs with valgrind, but this one is very very hard. Spent 60 hours on it already. So, some help from nvidia would be much appreciated.

2

If a VBO rendering command fails with other rendering calls around it, but not when rendering the elements alone is in my experience most often caused by incorrect vertex attribute enable state.
Please check all your vertex attribute enables when rendering the mesh alone and when rendering it in the failing case.

3

Also, you can try to reduce the draw count. Like just pushing 3 vertices. That also allows you to more or less test if you’re overflowing reading of the supplied arrrays.