Memory leak in VR_InitInternal2

[risa2000]

@mann1x here (cmbruns/pyopenvr#71) pointed out that he observed a memory leak when repeatedly calling OpenVR init function.

I was able to reproduce it in the native code:

#include <iostream>

#define OPENVR_BUILD_STATIC
#include <openvr/openvr.h>

//  Initialize OpenVR subsystem and return IVRSystem interace.
vr::IVRSystem* init_vrsys(vr::EVRApplicationType app_type)
{
    vr::EVRInitError eError = vr::VRInitError_None;
    vr::IVRSystem* vrsys = vr::VR_Init(&eError, app_type);

    if (eError != vr::VRInitError_None) {
        std::cerr << vr::VR_GetVRInitErrorAsEnglishDescription(eError) << '\n';
        return nullptr;
    }
    return vrsys;
}

int main(int argc, char* argv[])
{

    while(true) {
        // Initializing OpenVR runtime (IVRSystem)
        auto vrsys = init_vrsys(vr::VRApplication_Background);
        if (nullptr != vrsys) {
            break;
        }
    }

    vr::VR_Shutdown();
    return 0;
}

When running this code without starting SteamVR, the VR_Init call fails with an error Not starting vrserver for background app (121). Each failure seem to leak ~ 100KB on the stack in function VR_InitInternal2 as illustrated on the picture below.

This apparently becomes a problem when someone repeatedly calls VR_Init in an attempt to detect whether SteamVR has started.

The issue is observed on the stable 1.7.15 and the latest beta as well.

[kisak-valve]

Hello @risa2000, this issue is already being tracked at #195. Closing as a duplicate.

[risa2000]

@kisak-valve Thanks for letting me know, it is funny how the reports look similar even though I was not aware of the former.
Anyway, do you plan to fix? Why has it been open for so long?