Forgot to mention that the SharedBufferStack code was found at the
following links:
http://www.netmite.com/android/mydroid/2.0/frameworks/base/libs/ui/SharedBufferStack.cpp
http://www.netmite.com/android/mydroid/2.0/frameworks/base/include/private/ui/SharedBufferStack.h



On Sep 14, 11:22 am, timedilation <[email protected]> wrote:
> Any resolutions or workarounds so far on the freezing issue? I am
> seeing the same with my opengl app - mainly on HTC phones. I have
> never seen this happen on my G1 though and also never on the Motorola
> Droid. I am using Google recommended GLSurfaceView and all the
> standard stuff but the game itself is graphically fairly intensive.
>
> In my case too I get the same messages in logcat:
> WARN/SharedBufferStack(3878): waitForCondition(LockCondition) timed
> out ...... CPU may be pegged. trying again.
>
> This is what I have found from my own analysis thus far:
> Apparently the following code from SharedBufferStack.h must be causing
> these messages to be logged:
> //----------------------------------------
> template <typename T>
> status_t SharedBufferBase::waitForCondition(T condition)
> {
>     const SharedBufferStack& stack( *mSharedStack );
>     SharedClient& client( *mSharedClient );
>     const nsecs_t TIMEOUT = s2ns(1);
>     Mutex::Autolock _l(client.lock);
>     while ((condition()==false) &&
>             (stack.identity == mIdentity) &&
>             (stack.status == NO_ERROR))
>     {
>         status_t err = client.cv.waitRelative(client.lock, TIMEOUT);
>
>         // handle errors and timeouts
>         if (CC_UNLIKELY(err != NO_ERROR)) {
>             if (err == TIMED_OUT) {
>                 if (condition()) {
>                     LOGE("waitForCondition(%s) timed out (identity=
> %d), "
>                         "but condition is true! We recovered but it "
>                         "shouldn't happen." , T::name(),
>                         stack.identity);
>                     break;
>                 } else {
>                     LOGW("waitForCondition(%s) timed out "
>                         "(identity=%d, status=%d). "
>                         "CPU may be pegged. trying again.", T::name(),
>                         stack.identity, stack.status);
>                 }
>             } else {
>                 LOGE("waitForCondition(%s) error (%s) ",
>                         T::name(), strerror(-err));
>                 return err;
>             }
>         }
>     }
>     return (stack.identity != mIdentity) ? status_t(BAD_INDEX) :
> stack.status;}
>
> //----------------------------------------
>
> The GL thread hangs indefinitely while doing the eglSwapBuffer() call.
> This java call maps to a JNI C++ call via the function below:
> (taken from this 
> link:http://www.netmite.com/android/mydroid/frameworks/base/opengl/libGLES...)
>
> //-----------------------------------------
> EGLBoolean eglSwapBuffers(EGLDisplay dpy, EGLSurface draw)
> {
>     if (!validate_display_surface(dpy, draw))
>         return EGL_FALSE;
>     egl_display_t const * const dp = get_display(dpy);
>     egl_surface_t const * const s = get_surface(draw);
>     return s->cnx->hooks->egl.eglSwapBuffers(dp->dpys[s->impl], s->surface);
> }
>
> //-----------------------------------------
> The last line of this function I believe hooks into and makes a low
> level call into the GL driver library on the device. Somewhere in
> there might be a call that locks up and causes the main thread of my
> game to start spitting out those timeout (LockCondition) messages to
> logcat.
>
> On Sep 13, 6:24 pm, Jeremy Statz <[email protected]> wrote:
>
>
>
> > What I did to log this is to create a special logging class that can
> > output to SD card, then put a print at the top and bottom of every
> > function call in my wallpaper, with a fluch() after every line being
> > written.  This wrecks the framerate and results in a several hundred
> > meg log after a few hours, but means I have a reliable log of all
> > activity throughout the application.  Consistently, the crash occurs
> > in-between frames, and requires no special stimulation -- the phone
> > can just be sitting there running, then will suddenly die without
> > warning.  I'm confident saying that.
>
> > I've actually been using your GLWallpaperService code as a base,
> > though I've made a couple minor changes.  It's been quite stable
> > overall.  The big problem here, in my eyes, is that this appears to
> > affect any OpenGL using application, and is tremendously discouraging,
> > to the point that people pull things off the market and avoid using
> > OpenGL.  This hurts the platform as a whole, especially when comparing
> > its game library to the iPhone.
>
> > On top of that, while this seemed exceedingly uncommon on 2.1, I feel
> > like I've seen a whole lot more of it with 2.2 on my HTC Incredible.
> > Considering the free versions of my wallpapers have something like
> > three or four million downloads between them, that scares me a great
> > deal.  That's a lot of potentially affected users.
>
> > On Sep 13, 3:22 am, Robert Green <[email protected]> wrote:
>
> > > Jeremy,
>
> > > It happened to me quite a bit on both 2.1 and 2.2 on every device.
> > > It's surely a defect of Android but so far no one has been able to
> > > track it down.  It's a tough one, very difficult to reproduce
> > > reliably.  It tends to happen more often in certain cases for reasons
> > > unknown to me right now, such as on an EVO running 2.1 on the third
> > > level of Deadly Chambers if you play it straight through.  Since that
> > > takes 20 minutes, it's impossible to figure out by logging because
> > > it's just way too much.  Breakpoints don't work because the game
> > > doesn't run fast enough in debug mode and debugging doesn't work
> > > anyways once the lockup has happened.
>
> > > It's pretty much a worse case scenario - a very fatal, very
> > > intermittent bug that breaks debugging and is nearly impossible to
> > > isolate.
>
> > > I hope some genius out there can figure this one out.  I couldn't even
> > > get close.
>
> > > On Sep 12, 10:34 pm, Jeremy Statz <[email protected]> wrote:
>
> > > > There's been a couple threads about this in the last year or so, but I
> > > > wanted to draw attention to it again, as I've spent most of the last
> > > > couple of days trying to confirm it wasn't anything I'm doing.
>
> > > > The short story is, I'm pretty sure there's a very serious lock-up-the-
> > > > phone level bug that happens when using OpenGL applications.  When it
> > > > occurs the phone will freeze for at least two minutes before rebooting
> > > > itself, sometimes requiring that the battery be pulled.  This didn't
> > > > seem to happen often on 2.1, but I'm seeing it with a lot more
> > > > regularity on 2.2 (particularly on HTC phones, though that may be a
> > > > coincidence) and am really hoping to draw some attention from someone
> > > > in a position to fix it.  Typically, you'll end up seeing hundreds of
> > > > lines of this in the LogCat log:
>
> > > > WARN/SharedBufferStack(2005): waitForCondition(LockCondition) timed
> > > > out (identity=9, status=0). CPU may be pegged. trying again.
> > > > WARN/SharedBufferStack(71): waitForCondition(LockCondition) timed out
> > > > (identity=9, status=0). CPU may be pegged. trying again.
>
> > > > There won't be any callstack or error of any kind preceding this.  In
> > > > this case 2005 is my app (a live wallpaper), and 71 appears to be the
> > > > UI/system process.  I'm assuming this is a race condition within
> > > > SharedBufferStack.  I've seen cases where four or five process IDs
> > > > were all printing the message, all OpenGL apps of some kind.
>
> > > > Replicating the problem is very, very difficult and seems very
> > > > random... in the last case I had to leave the wallpaper running and
> > > > active on the home screen for about seven hours before it occurred.
> > > > Sometimes I've had it go days, other times it'll happen within a few
> > > > minutes.  My log output demonstrates no unusual activity at all before
> > > > the crash, loading/unloading/visibility/etc doesn't seem to be a
> > > > factor.  It's very rare.
>
> > > > I kind of feel like this got glossed over the last couple times people
> > > > have reported it here, but based on my reading it's a long-standing
> > > > problem with Android and OpenGL and is having a chilling effect on its
> > > > usage.  Is there anyone who can explain if there's anything I can do
> > > > to discourage problems here?  This is beyond my ken honestly, does
> > > > anyone know this part of the system well?

-- 
You received this message because you are subscribed to the Google
Groups "Android Developers" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/android-developers?hl=en

Reply via email to