Geoff Reidy writes:
> Curtis L. Olson wrote:
> > Geoff, this would appear to be a potential problem in the tile freeing
> > code.  Can you give me a route (i.e. a series of waypoints) that will
> > show the problem (and the approximate place where things die.)  Also,
> > leave the visibility at the default for the test (or tell me what
> > visibility you are using) because the visibility value determines the
> > size of the tile cache which affects when tiles need to be removed.
> > 
> > Regards,
> > 
> > Curt.
> > 
> 
> OK to test this I have been flying the a4, taking off from Sydney (YSSY) 
>   and immediately going into autopilot with waypoint set to Katmandu 
> (VNKT). Using visibility=64000 and cruising at about 12000ft.
> Also I'm running in 16bpp if that makes any difference.
> The assert error occurs shortly after crossing 125�E and about halfway 
> between the Australian coast and Timor.
> Oh, and this is before I started mucking about with the code :)

Geoff,

I did some testing at home this weekend and I think there is something
going on here related to the random ground cover objects.

If I run with --disable-random-objects I have never seen a crash.  If
I run with the random objects I do see a similar crash after flying
3000 miles give or take a couple thousand.

So I will say the following:

For long haul flights, consider running without random objects.
Random objects have a couple issues:

- program crash after a few thousand miles.  (To be fair, this may not
  directly be the random object code's fault.  It could be an
  interaction issue with other portions of the program?)

- issues with freeing tiles ... this can lead to substantially reduced
  frame rates while the system is struggling to free memory associated
  with a tile that has just been removed from the cache.

For short hops, training, or any other sort of flight where you aren't
venturing too far from home base, then it should be safe to leave
random objects on.

Regards,

Curt.
-- 
Curtis Olson   IVLab / HumanFIRST Program       FlightGear Project
Twin Cities    [EMAIL PROTECTED]                  [EMAIL PROTECTED]
Minnesota      http://www.menet.umn.edu/~curt   http://www.flightgear.org

_______________________________________________
Flightgear-devel mailing list
[EMAIL PROTECTED]
http://mail.flightgear.org/mailman/listinfo/flightgear-devel

Reply via email to