[Webkit-unassigned] [Bug 201507] [GTK] Accelerated compositing is disabled due to failure in EGL display creation

[bugzilla-daemon at webkit.org]

https://bugs.webkit.org/show_bug.cgi?id=201507

--- Comment #32 from Michael Catanzaro <mcatanzaro at gnome.org> ---
(In reply to Carlos Garcia Campos from comment #31)
> This is just a guess, we would need to confirm it. Since you seem to be the
> only one affected, we would need you to build WebKit without WPE renderer
> and check if the problem is gone.

Problem is, without a reproducer, it's impossible to ever know for sure if it's fixed. I can only guess based on whether I see pages that include video crashing (or, in the unlikely event I'm running in a terminal, if I see the error messages appearing there). I haven't noticed the issue for about two weeks now, when it happened two or three days in a row, and then there were several calm weeks before that.

Since multiple weeks without hitting the bug isn't evidence that it's fixed, might I suggest we build some debugging into WPEBackend-fdo to try to figure out what's going wrong next time this happens? Something must be going wrong in Instance::initialize in ws.cpp. Instead of returning false when initialization fails, how about we make this a fatal error and crash the web process instead? If we change each return false to g_error() then the backtrace from the crash will tell us exactly where it's failing inside this function. I don't see any other way to proceed, because the only way we have to indicate failure other than crashing is the bool return value, which doesn't convey enough information about the failure:

bool Instance::initialize(EGLDisplay eglDisplay)
{
    if (m_eglDisplay == eglDisplay)
        return true;

    if (m_eglDisplay != EGL_NO_DISPLAY) {
        g_warning("Multiple EGL displays are not supported.\n");
        return false;
    }

    const char* extensions = eglQueryString(eglDisplay, EGL_EXTENSIONS);
    if (isEGLExtensionSupported(extensions, "EGL_WL_bind_wayland_display")) {
        s_eglBindWaylandDisplayWL = reinterpret_cast<PFNEGLBINDWAYLANDDISPLAYWL>(eglGetProcAddress("eglBindWaylandDisplayWL"));
        assert(s_eglBindWaylandDisplayWL);
        s_eglQueryWaylandBufferWL = reinterpret_cast<PFNEGLQUERYWAYLANDBUFFERWL>(eglGetProcAddress("eglQueryWaylandBufferWL"));
        assert(s_eglQueryWaylandBufferWL);
    }
    if (!s_eglBindWaylandDisplayWL || !s_eglQueryWaylandBufferWL)
        return false;

    if (isEGLExtensionSupported(extensions, "EGL_KHR_image_base")) {
        s_eglCreateImageKHR = reinterpret_cast<PFNEGLCREATEIMAGEKHRPROC>(eglGetProcAddress("eglCreateImageKHR"));
        assert(s_eglCreateImageKHR);
        s_eglDestroyImageKHR = reinterpret_cast<PFNEGLDESTROYIMAGEKHRPROC>(eglGetProcAddress("eglDestroyImageKHR"));
        assert(s_eglDestroyImageKHR);
    }
    if (!s_eglCreateImageKHR || !s_eglDestroyImageKHR)
        return false;

    if (!s_eglBindWaylandDisplayWL(eglDisplay, m_display))
        return false;

    m_eglDisplay = eglDisplay;

    /* Initialize Linux dmabuf subsystem. */
    if (isEGLExtensionSupported(extensions, "EGL_EXT_image_dma_buf_import")
        && isEGLExtensionSupported(extensions, "EGL_EXT_image_dma_buf_import_modifiers")) {
        s_eglQueryDmaBufFormatsEXT = reinterpret_cast<PFNEGLQUERYDMABUFFORMATSEXTPROC>(eglGetProcAddress("eglQueryDmaBufFormatsEXT"));
        assert(s_eglQueryDmaBufFormatsEXT);
        s_eglQueryDmaBufModifiersEXT = reinterpret_cast<PFNEGLQUERYDMABUFMODIFIERSEXTPROC>(eglGetProcAddress("eglQueryDmaBufModifiersEXT"));
        assert(s_eglQueryDmaBufModifiersEXT);
    }

    if (s_eglQueryDmaBufFormatsEXT && s_eglQueryDmaBufModifiersEXT) {
        if (m_linuxDmabuf)
            assert(!"Linux-dmabuf has already been initialized");
        m_linuxDmabuf = linux_dmabuf_setup(m_display);
    }

    return true;
}

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.webkit.org/pipermail/webkit-unassigned/attachments/20191217/04c7b862/attachment.htm>