pythonopenglpyopengl

glGetBufferSubData pyopengl random result and segfault


To extend some stuff in my library I need to read data from a VBO back to CPU Memory. No Transformation-Feedback involved here.

When I read data from the Buffer I either get "random data" or a "segmentation fault" or "illegal hardware instruction" or "malloc".

Here are the different errors I get:

malloc

python(678,0x7fff746f7000) malloc: *** error for object 0x7fa532e1d6b8: 
incorrect checksum for freed object - object was probably modified after being freed.
*** set a breakpoint in malloc_error_break to debug
[1]    678 abort      python -m glib.examples.transformation_domain

segfault

[1]    2448 segmentation fault  python -m demos.read_vbo

illegal hardware instruction

objc[2789]: Method cache corrupted. This may be a message to an invalid object, or a memory error somewhere else.
objc[2789]: receiver 0x7fcaee1efaf0, SEL 0x7fff8314a468, isa 0x7fff72e0cf18, cache 0x7fff72e0cf28, buckets 0x7fcaee1d0a10, mask 0x1f, occupied 0x10
objc[2789]: receiver 64 bytes, buckets 528 bytes
objc[2789]: selector 'key'
objc[2789]: isa '_CFXNotificationNameWildcardObjectRegistration'
objc[2789]: Method cache corrupted.
[1]    2789 illegal hardware instruction  python -m glib.examples.transformation_domain

Since I was wondering if maybe there is something in my application which could lead to this problem (maybe a bad state, ...) I isolated the problem:

"""
this code demosntrates problems with glGetBufferSubData
@author Nicolas 'keksnicoh' Heimann
"""
from gllib.glfw import * 
from OpenGL.GL import * 
import numpy as np 

# spawn glfw stuff
if not glfwInit(): raise RuntimeError('glfw.Init() error')
glfwWindowHint(GLFW_CONTEXT_VERSION_MAJOR, 4);
glfwWindowHint(GLFW_CONTEXT_VERSION_MINOR, 1);
glfwWindowHint(GLFW_OPENGL_FORWARD_COMPAT, GL_TRUE);
glfwWindowHint(GLFW_OPENGL_PROFILE, GLFW_OPENGL_CORE_PROFILE);
window = glfwCreateWindow(500, 500)
glfwMakeContextCurrent(window)

# push data to vbo
data = np.array([1,2,3,4,5,6], dtype=np.float32)
vbo = glGenBuffers(1)
glBindBuffer(GL_ARRAY_BUFFER, vbo)
glBufferData(GL_ARRAY_BUFFER, 6*4, data, GL_STATIC_READ)
glBindBuffer(GL_ARRAY_BUFFER, 0)

# pullback
recv_data = np.empty_like(data)
glBindBuffer(GL_ARRAY_BUFFER, vbo)
glGetBufferSubData(GL_ARRAY_BUFFER, 0, 6*4, recv_data)
glBindBuffer(GL_ARRAY_BUFFER, 0)

print(recv_data)

This code yields different results each run or ends in one of the above described errors.

    1 keksnicoh@dhcp-172-21-66-45 ~/unih/bsc/opengl_plot_prototype (git)-[master] % python -m demos.read_vbo                                 :(
[  0.00000000e+00   1.58456325e+29   0.00000000e+00   1.58456325e+29
   4.02037986e-33   1.40129846e-45]
keksnicoh@dhcp-172-21-66-45 ~/unih/bsc/opengl_plot_prototype (git)-[master] % python -m demos.read_vbo
[  0.00000000e+00  -4.65661287e-10   0.00000000e+00  -4.65661287e-10
   5.88545355e-44   0.00000000e+00]
keksnicoh@dhcp-172-21-66-45 ~/unih/bsc/opengl_plot_prototype (git)-[master] % python -m demos.read_vbo
[  0.00000000e+00   3.68934881e+19   0.00000000e+00   3.68934881e+19
   1.89559592e+28   1.40129846e-43]
keksnicoh@dhcp-172-21-66-45 ~/unih/bsc/opengl_plot_prototype (git)-[master] % python -m demos.read_vbo
[  0.00000000e+00   2.52435490e-29   0.00000000e+00   2.52435490e-29
   1.89559592e+28   1.40129846e-43]
keksnicoh@dhcp-172-21-66-45 ~/unih/bsc/opengl_plot_prototype (git)-[master] % python -m demos.read_vbo
[  0.00000000e+00   3.68934881e+19   0.00000000e+00   3.68934881e+19
   2.92698291e-36   1.40129846e-45]
keksnicoh@dhcp-172-21-66-45 ~/unih/bsc/opengl_plot_prototype (git)-[master] % python -m demos.read_vbo
[  0.00000000e+00  -3.68934881e+19   0.00000000e+00  -3.68934881e+19
   7.42639198e-31   1.40129846e-45]
keksnicoh@dhcp-172-21-66-45 ~/unih/bsc/opengl_plot_prototype (git)-[master] % python -m demos.read_vbo
[  0.00000000e+00  -4.65661287e-10   0.00000000e+00  -4.65661287e-10
   7.22223950e-33   1.40129846e-45]
keksnicoh@dhcp-172-21-66-45 ~/unih/bsc/opengl_plot_prototype (git)-[master] % python -m demos.read_vbo
[1]    4263 segmentation fault  python -m demos.read_vbo
139 keksnicoh@dhcp-172-21-66-45 ~/unih/bsc/opengl_plot_prototype (git)-[master] % python -m demos.read_vbo                               :(
[1]    4335 segmentation fault  python -m demos.read_vbo

It seems like that this glGetBufferSubData may read from the wrong memory space? Or maybe there are some problems with pointer recv_data?

This is my OpenGL Setup:

[...] init GLFW
[...] load OPENGL_CORE_PROFILE 4.10
[...] initialize glfw window
  + Opengl version     410
  + GLSL Version       410
  + GLFW3              (3, 1, 2)

I use MacBookPro Retina 2013

Date/Time:             2015-06-22 21:30:53.795 +0200
OS Version:            Mac OS X 10.10.2 (14C109)
Report Version:        11

Solution

  • After some more investigation on this problem I found a solution. First of all I realized that something might happens to the recv_data argument. Causing an invalid operation exception I could see the internal trace to the c-invocation of clGetBufferSubData

        OpenGL.error.GLError: GLError(
        err = 1281,
        description = 'invalid value',
        baseOperation = glGetBufferSubData,
        pyArgs = (
            GL_ARRAY_BUFFER,
            0,
            48,
            array([ 0.,  0.,  0.,  0.,  0.,  0.], dtype=float32),
        ),
        cArgs = (
            GL_ARRAY_BUFFER,
            0,
            48,
            array([0, 0, 0, 0, 0, 0], dtype=uint8),  *** < data is now unit8 instead of float32
        ),
        cArguments = (
            GL_ARRAY_BUFFER,
            0,
            48,
            array([0, 0, 0, 0, 0, 0], dtype=uint8),  
        )
    )
    

    One can see, that maybe the argument was handled wrong, since it should not be an unit8? The function glGetBufferSubData returns a numpy.ndarray of type unit8. Then I convert this via. numpy view back to a float32 and could finally read the data.

    # pullback
    glBindBuffer(GL_ARRAY_BUFFER, vbo)
    raw_unit8_data = glGetBufferSubData(GL_ARRAY_BUFFER, 0, 6*4)
    glBindBuffer(GL_ARRAY_BUFFER, 0)
    print(raw_unit8_data.view('<f4'))
    

    Thanks to another stackoverflow article for the short conversion

    Since this behavior looks odd to me I created a bug ticket on pyopengl ticket system .