1

I have created a class to initialize and clear Xlib and glx properly. 

OpenGLContext::OpenGLContext()
    :m_display(nullptr)
    ,m_context(nullptr)
    ,m_vi(nullptr)
{
    memset(&m_cmap, 0, sizeof(Colormap));
    memset(&m_swa, 0, sizeof(XSetWindowAttributes));
    memset(&m_win, 0, sizeof(Window));
    m_display = XOpenDisplay(NULL);
    assert(m_display);

    static int dblBuf[] =  {GLX_RGBA, GLX_RED_SIZE, 1, GLX_GREEN_SIZE, 1, GLX_BLUE_SIZE, 1, GLX_DEPTH_SIZE, 12, GLX_DOUBLEBUFFER, None};
    m_vi = glXChooseVisual(m_display, DefaultScreen(m_display), dblBuf);
    m_context = glXCreateContext(m_display, m_vi, None, True);
    m_cmap = XCreateColormap(m_display, RootWindow(m_display, m_vi->screen), m_vi->visual, AllocNone);
    m_swa.colormap = m_cmap;
    m_win = XCreateWindow(
                m_display,
                RootWindow(  m_display, m_vi->screen  ),
                0, 0, /* width */ 640, /* height */ 480, 0, m_vi->depth, InputOutput, m_vi->visual,
                CWBorderPixel | CWColormap | CWEventMask, &m_swa
            );

    char* dummy[] = { "", 0 };
    XSetStandardProperties(m_display, m_win, "glxsimple", "glxsimple", None, dummy, 0, NULL);
    glXMakeCurrent(m_display, m_win, m_context);
    XMapWindow(GetDisplay(), GetWindow());
}

OpenGLContext::~OpenGLContext()
{
    XUnmapWindow(m_display, m_win);
    glXMakeCurrent(m_display, None, NULL);
    XFreeColormap(m_display, m_cmap);
    XDestroyWindow(m_display, m_win);
    glXDestroyContext(m_display, m_context);
    XFree(m_vi);
    XCloseDisplay(m_display);
}

Unfortunately, valgrind reports a memory leak.

==28742== 12,796 (584 direct, 12,212 indirect) bytes in 1 blocks are definitely lost in loss record 631 of 637
==28742==    at 0x4C29F5D: malloc (vg_replace_malloc.c:263)
==28742==    by 0xBCD7E7C: driConcatConfigs (in /usr/lib64/mesa/swrastg_dri.so)
==28742==    by 0xBCDBDFF: dri_init_screen_helper (in /usr/lib64/mesa/swrastg_dri.so)
==28742==    by 0xBCDAF0D: drisw_init_screen (in /usr/lib64/mesa/swrastg_dri.so)
==28742==    by 0xBCD8583: driCreateNewScreen (in /usr/lib64/mesa/swrastg_dri.so)
==28742==    by 0x5295604: driswCreateScreen (in /usr/lib64/opengl/xorg-x11/lib/libGL.so.1.2)
==28742==    by 0x527412B: __glXInitialize (in /usr/lib64/opengl/xorg-x11/lib/libGL.so.1.2)
==28742==    by 0x5270154: glXGetFBConfigs (in /usr/lib64/opengl/xorg-x11/lib/libGL.so.1.2)
==28742==    by 0x5270B57: glXChooseFBConfig (in /usr/lib64/opengl/xorg-x11/lib/libGL.so.1.2)
==28742==    by 0x4E9A7CE: ??? (in /usr/lib64/librrfaker.so)
==28742==    by 0x4E5B676: glXChooseVisual (in /usr/lib64/librrfaker.so)
==28742==    by 0x46D23B: Zion::Core::OpenGLContext::OpenGLContext() (OpenGLContext.cpp:23)

Note that I’m using VirtualGL (which explains librrfaker.so). Is there something I did wrong? or should I assume this is a bug on VirtualGL’s side?

4

2 回答 2

1

You shouldn't automatically assume that's it a bug. valgrind will sometimes return false positive results when dealing with optimised libraries. To be certain you need to ensure you are running against an unoptimised build of the library.

Edit

This extract from a valgrind manual suggest that this is only an issue when it checks for uninitialised variables. I remember being hit by false positives before but when I think of it now it was indeed for uninitialised values.

于 2012-04-08T19:43:18.407 回答
1

这可能是真正的内存泄漏，Valgrind 也检测到库中的泄漏。

在实践中 - 例如，如果这些函数仅在初始化时调用一次。虽然它们在技术上是泄漏 - 您可以放心地忽略它们，或者如果您愿意，您可以在上游调查泄漏（如果来源可用）。

检查文档，在glXChooseFBConfig上，返回值有望在XFree调用时被释放。

于 2015-09-10T09:40:08.780 回答