Menu

#222 JSBSim NaNs on reset/reposition

2.4.0
Fixed
nobody
2011-07-23
2010-12-28
Anonymous
No

Originally created by: zakalawe@mac.com
Originally owned by: zakalawe@mac.com

(I can't find a specific bug report for this, so filing this one)

When doing reset or re-position, both of which ultimately end up calling fgReInitSubsystems, JSBSIm aircraft often experience NaN-related crashes/un-caught exception throws. During init of JSBSim for the second (or greater) time, it performs an update with dt=0, and during this update, NaNs are generated.

Since the problem never happens on the first init of JSBSIm, it is presumed to relate to some static data, or uninitalized memory.

The problem has not always existed, and has got markedly worse since the 2.0 release - it used to occur with some low probability on each reset/reposition, but now occurs with 100% probability on any reset or reposition, for me. (But not necessarily for other users)

As soon as JSBSIm generates NaNs for /position/longitude-deg and latitude-deg, all code that depends on those (OpenAL, OSG, tile-manager, terrain intersection) ceases to function correctly - this tends to be the symptoms reported.

Related

Tickets: #194
Tickets: #204
Tickets: #216
Tickets: #249

Discussion

  • Anonymous

    Anonymous - 2010-12-28

    Originally posted by: zakalawe@mac.com

    Note that bug 204 post-dates this, and is a much more specific bug that has been introduced into JSBsim. Even when bug 204 is fixed, this issue, which is currently masked by it, will presumably re-appear.

     

    Related

    Tickets: #204

  • Anonymous

    Anonymous - 2011-01-01

    Originally posted by: zakalawe@mac.com

    Pushed:
    http://gitorious.org/fg/flightgear/commit/738d653745142f61ddcf1f4845532135f29dc8fa

    which makes a considerable difference to this for me - so far, I was not able to produce a NaN on reposition or reset with the C172. This needs wider testing and feedback now. (If you're testing, a find a NaN that can only reproduced with a particular aircraft, or init option, then please file a separate defect, since that is probably a different bug to this)

    Status: Testing

     
  • Anonymous

    Anonymous - 2011-01-04

    Originally posted by: g...@av8n.com

    At the moment, this seems to be not entirely fixed ... but no longer 100%
    reproducible.  I'm seeing NaN-related disasters maybe 1 time out of 5 when
    repositioning to a ground (airport) location.

     
  • Anonymous

    Anonymous - 2011-01-24

    Originally posted by: bre... (code.google.com)@gmail.com

    Another fix is available which may finally solve this issue (hopefully!).
    See emails with descriptions by Andreas Gaeb on the mailing-list.
    I'm also attaching the two patches here - just in case...

     
  • Anonymous

    Anonymous - 2011-01-24

    Originally posted by: zakalawe@mac.com

    Andreas' fixes are merged, and seem to work well for me - if you were experiencing this bug, please re-test with latest Git code, and provide feedback here.

    Relevant commit:
    http://gitorious.org/fg/flightgear/commit/4f364af6d178d947eae1a5a751e3a9542b270069

     
  • Anonymous

    Anonymous - 2011-02-19

    Originally posted by: bre... (code.google.com)@gmail.com

    Considered fixed. Need a new report otherwise.

    Status: Fixed

     
  • Anonymous

    Anonymous - 2011-06-05

    Originally posted by: bre... (code.google.com)@gmail.com

    (No comment was entered for this change.)

    Status: Temp

     
  • Anonymous

    Anonymous - 2011-06-05

    Originally posted by: bre... (code.google.com)@gmail.com

    (No comment was entered for this change.)

    Status: Fixed

     
  • Anonymous

    Anonymous - 2011-07-23

    Originally posted by: bre... (code.google.com)@gmail.com

    (No comment was entered for this change.)

    Labels: Milestone-2.4.0

     

Log in to post a comment.

MongoDB Logo MongoDB