Crash of 2600 meter

a mighty crash of 2600 meters.
Pixhawk 3.2RC2 mission with waypoints spline last command RTL
before DO-Jump WP 12 after the command do-jump to reach the copter was no longer to control. switch to RTL, Loiter. Stabilize and EKF, as well as turn off the transmitter has brought no response.
the copter I had already given up on him, crash after 14 minutes about 30 meters from the start.
I think it is a fault in the do-jump command.
I do not know what caused it. 'm already flown missions with this version without problems,
the configuration spline WP and Do-jump I had not previously been tested.
'm grateful for any help
LOGS … H8Os2sH8ga


I see a typo, I do jump would. WP12 on the WP2.
but why the loss of control?

At about line 145k , you can see it go bananas on roll axis.
It all starts when motor outputs are lowered.
Roll drops way below Desired roll, that correctly builds up soon after.
So it seems it wants to stay level, but overshoots wildly.

Your Roll and pitch I and D terms are all 0.15 - Unless this is a one-in-a-millon case, it does not seem right to me. My guess would be it’s badly tuned, and started to hunt.

Regarding your altitude, I doubt, you did that legally.

thank you for the quick response

at 145K the lipo is almost discharged, so the overshoot.

the error bug is that I have inadvertently placed the command do_jump to the last WP 12. WP must be 2.

This BUG is very dangerous to it.

the copter out of control. he rises to the Lipo is discharged.

It would be nice if a developer has to say something.

[color=#00BF00]Please note that Beta versions and release candidates should only be used by very experienced users and only to help the developers test the code.
In those versions, bugs are more likely to be present and the user should be aware of a higher operational risk and choose location and operation modes for flights accordingly.[/color]

thanks for the quick reply.

no test no errors are found.
I am aware of the matter.
work with Arducopter since the beginnings

I’ve answered this on the AC3.2 testing thread. It is indeed a very bad bug. A divide by zero in the spline controller. We have a fix that I’ll push into master today and it’ll go out with -rc3.

Sorry for the troubles!