Crazy Physics Bug

Hi, just wondering if anyone has a clue what’s going on here?

This is on a ZTE Blade, it doesn’t ever happen on my SGS2. It also only happens when in contact with something on a >45’ angle. I have tried changing the physics solver iteration count, when changing this to 100 it sends the bike straight to blue screen as soon as it moves (it blue screens when the bikes position becomes Nah due to it being blasted off the screen) it worked fine at 100 solver count on the SGS2.
So this has me very confused, any ideas of what I should try next?

Are you putting physics in fixedupdate?

Yer I am :slight_smile:

I just dont get how it will work on the galaxy s2 but completely glitch out on the ZTE Blade :s

Bump, anyone else got any ideas?

check your collision layers, ive had something like this happen with hitting a golf ball,

make sure your colliders that your driving on are on the same layer and make sure it is turned off for them to collide with each other,

when the collision between the same layer is turned on and they are intersecting from the start, anything that hits them will freak out and shoot away

Thanks for the suggestion Pat, just tried sorting them out and it made no difference :frowning: any other ideas?

if it works on one device but not the other that somehow really implies that there is one or more cases where forces or even worse position changes (should never happen) get applied in Update or LateUpdate instead of fixed update, leading to inconsistent non-continous physics which on a slower device will lead to ‘physical teleports’ which in case of a collision can have all kind of nightmare consequence because it might need to push the object out extremely far to resolve the collision and such pushes can lead to very strong forces up to ‘infinite forces’

Are you sure you are only using the physical functions and no modifications to positions and rotations otherwise?

I only use AddRelativeTorque and AddTorque to move the bike, the only exception to this is when the bike is reset, At which point it is set to kinematic and put to sleep. Also all AddTorque functions are called in fixedupdate.
What you have described dreamora is exactly what it looks like is happening, I just cant figure out why :frowning:

Bump, any more ideas?
It seems this is only happening on 600MHz devices, I just want to know what’s doing it!

Still suffering :frowning: really no one has a clue? will take any suggestions!!!

have you sent unity a bug report? what about a logcat report?

you can either use android sdk to get the logcat or download log collector from the market, its a free app that gives you a log cat report right on your phone, it almost seems as the phone is loosing memory?

can you replicate this problem in a simplified scene? try to recreate the problem area in a new scene and see if it continues to happen…

if it does in a very simple scene then it may be the setup? if it doesnt happen it may be something in your bigger scene or the phone dumping memory

do you have specific phone models this happens on vs. not happening on?

if you want me to test it on my phone, i have a droid x, you can PM me

im not sure if any of this will help, just tossing ideas out there to test on, let me know how it goes

You can get unstuck in physics by calculating a number in update, then applying that number in physics which are in fixedupdate. In cases like that, you would need deltatiming in update.

Can be replicated in a simple scene with just one metal girder.

It seems to only happen on phones with 600MHz ARM11 processor with Qualcomm MSM7227 chipset.

What do you mean by hippo? could you give a slightly wordier explanation? :slight_smile:

Finally found a solution!!! although no idea why it worked :s
I put a box collider inside all my objects that had a rotation vector = (0,90,0). Then just scaled the collider to fix the shape and the glitch is gone. Seriously weird, anyone know why it would do this?

no idea, glad you found it out :smile: keep us updated on the release!