Liniare Motor Bug

My prievious two posts harped a bit on the jerkyness (rubberbanding) of phyisics based scripted vehicles (boats in my case), that have been occuring since the last update. And looking at the known issues list, I see no mention of this item.

However I noticed something last night that might shed some light on whats happening here. I noticed that physics based scripts, which use Linden wind for movement dont suffer from this malady, nore do mulit-move (non-physics based scripts). Also, physics based scripts seem to work smoothly when the turning inputs are alternately, and continually used.

This tells me there is a bug in the new release that affects the liniar motor which governs forward and backwards movements of physics based scripts of vehicles.

I have read posts, and heard from others inworld that have encountered this same issue. Now the big question, how does one get this listed in the known issues list?

It doesn't seem to effect linear motor itself.

It effects the rate at which position updates are sent to the client for vehicles using linear and angular motors (that is, most of them). It only seems to send updates when you hit a control... any control, including throttle controls. On the sim side the motion is as smooth as ever, you just don't see it as smooth on the client unless you're sending control updates.

It seems to be based on hitting the control keys, because other people's vehicles seem smoother if you just move around a little, but I won't swear that I'm not fooling myself there.

I haven't noticed that on my vehicles. They seem as smooth on this first look release as before and they are autonomous with no control inputs at all.

Could you explain this a bit more?

Are you sure? I've watched jerky vehicles move with Show Updates enabled and they seem to be doing position updates (blue) pretty quickly - maybe +10/second or so. This is on a slow moving vehicle going less than 1m/second after a small linear push.

Or am I missing the point?

I guess what he is saying is that the viewed vehicles are moving smoothly, and what Im saying is that from the driver's view, they are getting delayed updates and so appear to have eratic movement unless a control input is applyed.

This condition was not noticed until the last update, so something in that last update caused this effect. The following patches failed to address this, and it is not listed in the known issues, so the Lindens need to be made aware of it.

Again, this only seems to effect vehicles using liniare motors in physics based scripts.

I don't think this can be narrowed down to just the linear motor like you've proposed. I notice the exact same choppiness of movement in my physical objects in Suffugium, which move around using llApplyImpulse, not the vehicle API.

I haven't tried it with show updates enabled, but that's just bizarre... because they don't behave like the jerkiness is sim-side... they recover to the right position when they finally *do* get rubber-banded back into place. Why would they bother sending updates for incorrect positions?
That's what I'm saying too... but it doesn't seem to matter which motor the control is applied to.

Hmm, what I have noticed from time to time is the vehicle is rendered way off course, then after a few seconds snaps back to where it should be.

Is that what you're seeing? I've been seeing this for awhile now though - several weeks.

Bizarre is a good word.

More and more, I'm thinking this some client side optimization/bugfix gone bad. I've recently noticed the problem go from bad-but-not-horrible when the camera is fairly close to being absolutely-horrible when I pull the camera way back and watch from a distance.

Between that and how things look with Show Updates on, I think this doesn't have anything to (directly) do with linear motors - it's all just client weirdness.

edit: or, if SL does client-side prediction with physical objects, it's a bug in that - they're trying to be sneaky with easy linear-only movements and something's getting lost or out of sync. When the angular motor is in effect, they don't try to be sneaky and just take the updates from the sim. I'm totally making this up but it almost sounds good..

Well if you read my original post on this, you'll note that the problem dosent appear with scripted items that use no liniar motor, such as Kranker Greenace's Takos which use Linden wind to move. Also the multimove scripts are much smoother in movement that the physics based ones at the moment.

So it does kind of point to the liniar motor thingy being borked.

The Tako doesn't really use the wind as such. It senses the wind in some modes and then uses the motors to move.

And if you read my post, I'm telling you that my observations aren't compatible with that, because my scripted items that use no linear motor still do exhibit this behavior. Check my police drones and cars in Suffugium sim for an example.

Incidentally, there's no such thing as "using Linden wind to move". All a script like that does is read the llWind() state and use llApplyImpulse to make it have an effect on their object. Perhaps the reason that Greenace's script doesn't exhibit this behavior is because it's polling super-fast and applying impulses many times per second.

Anybody notice the release notes for tomorrows update??



/me crosses her fingers.

Only took them 2 months, keeping fingers crossed as well.

I tested this in the beta last weekend. Flying is much improved, and they appear to have set up a testbed for the "unsit" bug. Haven't scotched it yet, though.

Still a little jerky but far better..