Viggen freeze during looping

[TOViper]

This is the second flight I got a freeze when doing just a looping over the runway, using Viggen as the only aircraft in a simple single player mission (starting from the mission editor and taking off).

 

It happens when passing the vertical.

 

DCS freezes, and there is not other solution than killing the process via the task manager.

It happend several times in the last days, so now I have enough of it

 

Edit: had to change title, because it is not related to Marianas Islands only.

 

The LOG is attached.

dcs.log

Edited July 20, 2021 by TOViper

[Saab87]

Having the same issue sometimes 

[TOViper]

I can reproduce the bug, happend just before. Let's see if I can make a video ...

https://www.youtube.com/watch?v=HPWkmNQgRSs

 

Edited July 14, 2021 by TOViper
link for video

[TOViper]

BTW: happens in Caucasus as well, just had a freeze two hours ago.

[Machalot]

Who in their right mind pulls up through vertical in the Viggen?  Get your nose back down in the mud where it belongs! 

[QuiGon]

7 hours ago, Machalot said:

Who in their right mind pulls up through vertical in the Viggen?  Get your nose back down in the mud where it belongs! 

 

right?!

[TOViper]

This bug seems to be happening not only on my machine ... at least I am not alone in the dark.

@RagnarDa, plz help!!

 

https://forums.eagle.ru/topic/280625-viggen-crash-to-desktop/?tab=comments#comment-4760018

 

[Spanueh]

I’ve had this loop-bug happen too; get to the top of a loop and crash.  Unfortunately the game crashes so hard there is no track-file or crash-log.

Turning the mirrors off seems to fix the problem.

I’ll be checking for the bug with a new computer, which is in the mail, this weekend and report back.

[TOViper]

1 hour ago, Spanueh said:

I’ve had this loop-bug happen too; get to the top of a loop and crash.  Unfortunately the game crashes so hard there is no track-file or crash-log.

Turning the mirrors off seems to fix the problem.

I’ll be checking for the bug with a new computer, which is in the mail, this weekend and report back.

Can't confirm this, since I flew the sorties with mirrors off anyway.

[MYSE1234]

39 minutes ago, TOViper said:

Can't confirm this, since I flew the sorties with mirrors off anyway.

Agreed, that's my experience as well.

[Spanueh]

Just a small update on the Loop-Crash.  With a new computer and my old video card gtx1070 (new card is defective), the crash/freeze does still happen.  Both with the mirrors on and off and the results are the same.

[TOViper]

Now, eventually after a few test flights today, I got a LOG file.

Single player, engine out, stalling the aircraft and pitching it violently in a strong nose down attitude (around -90° of pitch).

 

dcs.log

[TOViper]

New DCS update log from today:

 

  Fixed DCS Crash when looping the aircraft

 

Oh yeah baby yeah baby

Edited September 17, 2021 by TOViper

[TOViper]

Back to work: this is not fixed. Still got a freeze when going through 90° nose down during a looping.

Attached see log.

 

dcs.log

[hilmerby]

Same thing happened to me!

[RealDCSpilot]

Me too. Yesterday, 4YA Syria Open Beta server. Did a loop and DCS crashed. CTDs are usually very rare on my end.

[IronMike]

Thank you all, also for the logs! We hope we can get behind this asap.

[Rapierarch]

Just an update. I'm having the same problem. (last night twice)

Currently on open beta latest.

[CaptCrash]

Confirmed as well after 3 half cuban eights, the fourth crashed.
edit: I was able to duplicate by flying level ->pull up slightly past vertical -> push forward to slightly past vertical. So level -> ~+95 degrees pitch change -> -6 degrees pitch change and the sim crashed.
I am also in the latest open beta
edit 2: checked the log:

2021-10-29 05:20:26.653 ERROR   VIGGEN: HeatBlur: Assert failed: Value out of range. -. Was nan. Expected between 0.000000 and 6.283185 @ e:\dcs\viggen git\ajs_5\trunk\viggennew\sources\flightmodel\flightdataunit.cpp:695

Tacview-20211028-221902-DCS-04 - AJS37 Training Basic Navigation.txt.acmidcs.log

Edited October 29, 2021 by CaptCrash

[Machalot]

On 10/28/2021 at 9:58 PM, CaptCrash said:

Confirmed as well after 3 half cuban eights, the fourth crashed.
edit: I was able to duplicate by flying level ->pull up slightly past vertical -> push forward to slightly past vertical. So level -> ~+95 degrees pitch change -> -6 degrees pitch change and the sim crashed.
I am also in the latest open beta
edit 2: checked the log:

2021-10-29 05:20:26.653 ERROR   VIGGEN: HeatBlur: Assert failed: Value out of range. -. Was nan. Expected between 0.000000 and 6.283185 @ e:\dcs\viggen git\ajs_5\trunk\viggennew\sources\flightmodel\flightdataunit.cpp:695

Tacview-20211028-221902-DCS-04 - AJS37 Training Basic Navigation.txt.acmi 128 kB · 0 downloads dcs.log 63.5 kB · 1 download

 

The expected range of some variable is given as 0 to 2 pi.  I would bet money there is an issue with calculating the yaw, pitch, and roll Euler angles at vertical.  Based on the values 0 to 2 pi, it's likely yaw.  (Pitch is typically -pi/2 to +pi/2, and roll is typically -pi to +pi.)

There is a known singularity in the math at +/-90 deg pitch angle.  A "nan" result could be from trying to divide by zero [which is cos(90°), and appears in the denominator of the yaw equation] or by taking the inverse trig function of a value outside the range [-1, 1]. Both of those operations can appear in the relevant Euler angle equations, depending how they're written. The range exceedance could be the result of floating point rounding error.  These things should only occur very close to 90 deg pitch, which is why sometimes you don't get the crash, and sometimes you do. 

My own hypothesis, anyway.  That's where I would start debugging. 

Edited October 31, 2021 by Machalot

[CaptCrash]

3 hours ago, Machalot said:

The expected range of some variable is given as 0 to 2 pi.  I would bet money there is an issue with calculating the yaw, pitch, and roll Euler angles at vertical.  Based on the values 0 to 2 pi, it's likely yaw.  (Pitch is typically -pi/2 to +pi/2, and roll is typically -pi to +pi.)

There is a known singularity in the math at +/-90 deg pitch angle.  A "nan" result could be from trying to divide by zero [which is cos(90°), and appears in the denominator of the yaw equation] or by taking the inverse trig function of a value outside the range [-1, 1]. Both of those operations can appear in the relevant Euler angle equations, depending how they're written. The range exceedance could be the result of floating point rounding error.  These things should only occur very close to 90 deg pitch, which is why sometimes you don't get the crash, and sometimes you do. 

My own hypothesis, anyway.  That's where I would start debugging. 

 

My guess is that the attitude indicator is what is causing the nan since DCS would be using quaternions for the flight model

[Machalot]

8 hours ago, CaptCrash said:

My guess is that the attitude indicator is what is causing the nan since DCS would be using quaternions for the flight model

I agree. The error is occurring in flightdataunit.cpp, which seems like a plausible name for some avionics and instrument model(s).

Edited October 31, 2021 by Machalot

[TOViper]

Seems gone with update OB 2.7.8.16140.
Can someone confirm?

 

[MYSE1234]

I can not unfortunately. Just had a crash from it after 30 sec of testing.

[TOViper]

Yep, wrong alert. After 1 hour of flying a few sorties I got the freeze again.

2021-11-26 21:04:19.949 ERROR   VIGGEN: HeatBlur: Assert failed: Value out of range. -. Was nan. Expected between 0.000000 and 6.283185 @ e:\dcs\viggen git\ajs_5\trunk\viggennew\sources\flightmodel\flightdataunit.cpp:695

DAMN