On Fri, Mar 6, 2009 at 4:11 PM, Richard Thomas <[email protected]> wrote:
>
> On Mar 6, 7:00 pm, Tristam MacDonald <[email protected]> wrote:
> > On Fri, Mar 6, 2009 at 1:43 PM, Richard Thomas <[email protected]>
> wrote:
> > > On Mar 6, 12:51 am, Tristam MacDonald <[email protected]> wrote:
> > > > On Thu, Mar 5, 2009 at 5:42 PM, Richard Thomas <[email protected]>
> > > wrote:
> >
> > > > > On Mar 5, 9:10 pm, Tristam MacDonald <[email protected]> wrote:
> > > > > > On Thu, Mar 5, 2009 at 1:57 PM, Richard Thomas <
> [email protected]>
> > > > > wrote:
> >
> > > > > > > I am experiencing a very weird bug. My interpreter keeps
> > > segfaulting
> > > > > > > when I do any serious amount of GL. It almost always happens
> but
> > > not
> > > > > > > after a consistent amount of time. I have a simple lap timer
> > > program I
> > > > > > > wrote and to keep it from being boring I put a little expanding
> > > > > > > circles effect in the background. Now if I disable the label
> that's
> > > > > > > displaying the time or the circles then it'll run seemingly
> forever
> > > > > > > without crashing, if I do both at once it crashes within a
> minute
> > > or
> > > > > > > so.
> >
> > > > > > > I've exhausted almost every possibility I can think of. I ran
> > > memtest
> > > > > > > for half a dozen passes in cases I had a hardware fault.
> Pyglet's
> > > GL
> > > > > > > wrapper is simple enough that I can't really see ctypes being
> the
> > > > > > > problem or foreign function libraries would be giving up all
> over
> > > the
> > > > > > > place. Besides which nobody else I've tried has had this
> problem.
> >
> > > > > > > Is there anyone who's experienced anything like this?
> >
> > > > > > > Sometimes I get a bus error instead of a seg fault.
> >
> > > > > > What Mac hardware are you running?
> > > > > > If you have the Intel integrated X3100 GPU, present in 2008
> MacBooks
> > > and
> > > > > > iMacs, and possibly some Mac Minis, then any attempt to mix
> immediate
> > > > > mode
> > > > > > OpenGL with VBOs can cause both segfaults and bus errors.
> > > > > > There are a couple of posts on my blog (in signature) related to
> this
> > > > > issue.
> >
> > > > > Yes, I am using an X3100 GPU on an early 2008 MacBook. The
> conditions
> > > > > you describe make sense as labels are VBO based and my circle
> effect
> > > > > was immediate mode. However, I am already running 10.5.6. I
> downloaded
> > > > > and ran the test from your blog for 10 minutes with no sign of a
> > > > > problem so this may be another (albeit similar) issue. Sometimes
> the
> > > > > traceback is inside a glDrawArray (presumably from pyglets batch
> > > > > rendering) and sometimes a glVertex2f (my immediate mode
> rendering).
> >
> > > > > I'm not sure I know where to start with whittling out a narrow test
> > > > > case. Any advice?
> >
> > > > > Richard.
> >
> > > > My original test case doesn't seem to cover all the problems, in
> > > particular
> > > > Blender still crashes regularly. I would recommend that you try
> switching
> > > > your circle effect over to VBO (i.e. pyglet.graphics.triangle_list)
> > > > temporarily, and see if the problem still occurs.
> >
> > > Having switched to VBOs it still crashes. Maybe there's just a problem
> > > with VBOs...
> >
> > > I'd be grateful if anyone on similar hardware (early 2008 MacBook,
> > > X3100 GPU) could run this test for a few minutes and report what
> > > happens. There is a pair of lines one of which is commented to swap
> > > from immediate mode on the circles to VBOs.
> >
> > > Script:http://www.chard.ath.cx/~richard/files/opengl_crash.py<
> http://www.chard.ath.cx/%7Erichard/files/opengl_crash.py>
> >
> > Crashed after about 20 seconds on my MacBook (White, September 2008,
> X3100).
> > About a minute later the whole system went down, as has often happened in
> my
> > test of this bug. It is definitely the same issue I was testing before,
> but
> > you seem to have discovered a new facet of it.
>

> I've got it down to a much smaller test case, there's no effects or
> anything anymore, it just crashes within 10-15 seconds. There is a
> curious line though, commented out, that if uncommented will cause it
> not to crash. I will let the script speak for itself.
>
> http://www.chard.ath.cx/~richard/files/opengl_crash.py
>

By adding a print statement to the script, I was able to ascertain that the
crash always occurs when number=68. Beats me why that would be, though:
number = 0
def update(dt=None):
    global number
    number = min(100, number + 1)
    print number

-- 
Tristam MacDonald
http://swiftcoder.wordpress.com/

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"pyglet-users" 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/pyglet-users?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to