0005488: llLookAt not working on diva-r15402-b

(0020019)
justincc (administrator)
2011-09-23 18:41

Does this fail on current development code?

(0020349)
Digi Fly (reporter)
2011-11-13 18:05
edited on: 2011-11-13 18:06

I have tested llLookAt, with this script, same as above only lower strength. used it in sl & opensim. SL works fine.
and opensim its a dead cube. no movement at all.

Tested with git: 6b4ae4f30c2091f5f6fa99dfc904cfdd42a266ea [r/17162]

//Causes Object to look at nearest Avatar.
default
{
    state_entry()
    {
        llSensorRepeat("", "", AGENT, 50.0, PI, 0.5);
    }

    sensor(integer total_number)
    {
        llLookAt( llDetectedPos(0) + <0.0, 0.0, 1.0>, 1.0, 1.0 );
    }
}

(0021350)
Ferd Fred (reporter)
2012-05-02 10:47

This is a partial fix by using a replacement function:

// This is an OpenSim compatible llLookAt()!
face_target(vector lookat)
{
rotation rot = llGetRot() * llRotBetween(<0.0 ,0.0 ,1.0 > * llGetRot(), lookat - llGetPos());
llSetRot(rot);
}

(0021353)
cmickeyb (administrator)
2012-05-03 08:59

Can you tell me what behavior you are seeing and what behavior you think you should see. llLookAt() has been "fixed" multiple times. It does not use physics so it won't behave the same as SL specifically with the strength and dampening parameters.

(0021355)
Digi Fly (reporter)
2012-05-03 12:05

Quick test shows its working normal in opensim parameters :)
But right now i have test it on linux. Think it looks fixt,
only missing phyics :)

(0023771)
tglion (reporter)
2013-04-09 00:51

llLookAt does still not working correct. Whith or without physics,
it rotates often wrongly around z-axsis. Looks like a bit messy.
This is on latest git checkout of dev master.

To reproduce:
Create a rectangle-prim x=0.5m, y=0.4m, z=1.0m , put an eye-texture on the top.
Then put following script into this prim:

default {
    state_entry() {
        llSensorRepeat("", llGetOwner(), AGENT, 96.0, PI, 0.5);
    }

    sensor(integer n) {
        integer i;
        for (i=0; i<n; ++i) {
            vector p1 = llDetectedPos(i);
            llLookAt(p1, 1.0, 1.0);
            return;
        }
    }
}

walking around the object now, compare same on SL and you will see, whats going wrong...

(0023772)
Richardus Raymaker (reporter)
2013-04-09 06:27

opensim 0.7.5 final here.

Just tried the latest example , and saw something strange.
After some manual testing with singularity and imprudence it looks like
the Y axe is broken with rotation. If i use in the viewer the up or down arrow i can only set rotation on the Y axe from 270 -> 0 -> 90 degree.

It dont go round like the other axes from 0-360.

(0023774)
tglion (reporter)
2013-04-10 07:12

Other problems are:
1) the llLookAt don't respect the axis-locking (for eg. if set llSetStatus(STATUS_ROTATE_X, FALSE); in physics-mode).
2) The lookAt on physics objects should look with top of x-axis (like in SL) and not with the top of z-axis of the object.

Tested with PhysicsEngine ODE and Bullet and compared it with SL

(0023775)
cmickeyb (administrator)
2013-04-10 10:37

Just some feedback on this. The current implementation of llLookAt is a complete hack using explicit position updates and (at least used to) "works" for a small number of cases. It doesn't use any physics. Until recently we didn't have any of the machinery necessary to make it work. Some of the recent changes in Bullet might enable a "correct" implementation.

I'll include radams1 in this chain & talk to him about how to get it working with Bullet. If anyone would like to provide some code for the implementation, that would be great. :-)

Issue History
Date Modified	Username	Field	Change
2011-05-18 11:26	ebremer	New Issue
2011-05-18 11:26	ebremer	Git Revision	=> diva-r15402-b
2011-05-18 11:26	ebremer	SVN Revision	=> 0
2011-05-18 11:26	ebremer	Run Mode	=> Standalone (Multiple Regions)
2011-05-18 11:26	ebremer	Physics Engine	=> BasicPhysics
2011-05-18 11:26	ebremer	Environment	=> .NET / Windows64
2011-05-18 11:26	ebremer	Mono Version	=> None
2011-05-18 11:26	ebremer	Viewer	=> Second Life 2.6.6 -229662
2011-09-23 18:41	justincc	Note Added: 0020019
2011-11-13 18:05	Digi Fly	Note Added: 0020349
2011-11-13 18:06	Digi Fly	Note Edited: 0020349	View Revisions
2011-11-13 18:06	Digi Fly	Note Edited: 0020349	View Revisions
2012-05-02 10:47	Ferd Fred	Note Added: 0021350
2012-05-03 08:59	cmickeyb	Note Added: 0021353
2012-05-03 12:05	Digi Fly	Note Added: 0021355
2013-04-09 00:51	tglion	Note Added: 0023771
2013-04-09 06:27	Richardus Raymaker	Note Added: 0023772
2013-04-10 07:12	tglion	Note Added: 0023774
2013-04-10 10:37	cmickeyb	Note Added: 0023775
2013-04-10 10:38	cmickeyb	Assigned To	=> Robert Adams
2013-04-10 10:38	cmickeyb	Status	new => assigned
2015-01-15 15:58	justincc	Relationship added	has duplicate 0006888

Notes
(0020019) justincc (administrator) 2011-09-23 18:41	Does this fail on current development code?

(0020349) Digi Fly (reporter) 2011-11-13 18:05 edited on: 2011-11-13 18:06	I have tested llLookAt, with this script, same as above only lower strength. used it in sl & opensim. SL works fine. and opensim its a dead cube. no movement at all. Tested with git: 6b4ae4f30c2091f5f6fa99dfc904cfdd42a266ea [r/17162] //Causes Object to look at nearest Avatar. default { state_entry() { llSensorRepeat("", "", AGENT, 50.0, PI, 0.5); } sensor(integer total_number) { llLookAt( llDetectedPos(0) + <0.0, 0.0, 1.0>, 1.0, 1.0 ); } }

(0021350) Ferd Fred (reporter) 2012-05-02 10:47	This is a partial fix by using a replacement function: // This is an OpenSim compatible llLookAt()! face_target(vector lookat) { rotation rot = llGetRot() * llRotBetween(<0.0 ,0.0 ,1.0 > * llGetRot(), lookat - llGetPos()); llSetRot(rot); }

(0021353) cmickeyb (administrator) 2012-05-03 08:59	Can you tell me what behavior you are seeing and what behavior you think you should see. llLookAt() has been "fixed" multiple times. It does not use physics so it won't behave the same as SL specifically with the strength and dampening parameters.

(0021355) Digi Fly (reporter) 2012-05-03 12:05	Quick test shows its working normal in opensim parameters :) But right now i have test it on linux. Think it looks fixt, only missing phyics :)

(0023771) tglion (reporter) 2013-04-09 00:51	llLookAt does still not working correct. Whith or without physics, it rotates often wrongly around z-axsis. Looks like a bit messy. This is on latest git checkout of dev master. To reproduce: Create a rectangle-prim x=0.5m, y=0.4m, z=1.0m , put an eye-texture on the top. Then put following script into this prim: default { state_entry() { llSensorRepeat("", llGetOwner(), AGENT, 96.0, PI, 0.5); } sensor(integer n) { integer i; for (i=0; i<n; ++i) { vector p1 = llDetectedPos(i); llLookAt(p1, 1.0, 1.0); return; } } } walking around the object now, compare same on SL and you will see, whats going wrong...

(0023772) Richardus Raymaker (reporter) 2013-04-09 06:27	opensim 0.7.5 final here. Just tried the latest example , and saw something strange. After some manual testing with singularity and imprudence it looks like the Y axe is broken with rotation. If i use in the viewer the up or down arrow i can only set rotation on the Y axe from 270 -> 0 -> 90 degree. It dont go round like the other axes from 0-360.

(0023774) tglion (reporter) 2013-04-10 07:12	Other problems are: 1) the llLookAt don't respect the axis-locking (for eg. if set llSetStatus(STATUS_ROTATE_X, FALSE); in physics-mode). 2) The lookAt on physics objects should look with top of x-axis (like in SL) and not with the top of z-axis of the object. Tested with PhysicsEngine ODE and Bullet and compared it with SL

(0023775) cmickeyb (administrator) 2013-04-10 10:37	Just some feedback on this. The current implementation of llLookAt is a complete hack using explicit position updates and (at least used to) "works" for a small number of cases. It doesn't use any physics. Until recently we didn't have any of the machinery necessary to make it work. Some of the recent changes in Bullet might enable a "correct" implementation. I'll include radams1 in this chain & talk to him about how to get it working with Bullet. If anyone would like to provide some code for the implementation, that would be great. :-)