problem with depth sensor when running Linux in VM

[fferri]

I'm using V-REP 3.1.2 linux 64-bit, and I'm unable to get correct readings from the provided SICK S300 sensor.
It used to work in 3.1.0, but in 3.1.2, if you set script parameter showLaserSegments=true, you'll see that laser segments look like if they are "inside" the sensor mesh.

Screenshot:


EDIT: changed topic title: it seems that this is caused by graphic card issues when running ubuntu in a virtual machine.

[coppelia]

Hello,

are your vision sensors working fine otherwise? There might be a graphic card compatibility..

The laser scanner is set-up by two vision sensors. Can you try to somewhat adjust the Near / far clipping planes? Also, can you try to turn the Renderable flag of the casing (object SICK_S300_fast) off to see if that makes a difference?

Cheers

[fferri]

It seems the RGB information of a vision sensor is there, but the depth information is not.
(note: I'm running in a virtual machine, Ubuntu 14.04)

As a double check can you suggest me a scene where the depth sensor is demoed so that I'm sure to be testing it properly?

[coppelia]

Try with the model Models/components/sensors/Mesa SR4000.ttm.

You can also turn the depth buffer retrieval on/off via the item Ignore depth info (faster).

Finally, you can try to adjust some OpenGl settings, mainly the item Offscreen context type and/or FBO type.

Cheers

[fferri]

It seems a problem with the graphics driver when running ubuntu in a virtual machine (I updated topic title accordingly).

[coppelia]

Could you solve the problem? If yes, what was the solution?

[fferri]

Nope, I didn't solve it.

I tried all six combinations of {offscreen surface, invisible window, visible window} x {Qt, non Qt}, but none of them fixed the issue.

[coppelia]

Could you confirm that the other vision sensors are working fine? (e.g. the Mesa SR4000 model)

[fferri]

Nope, also the Mesa SR4000 sensor gives me a black depth image (but RGB image is OK) when running in a virtual machine.

Note:

I think I erroneously assumed it was a 3.1.0->3.1.2 regression, because it worked on my lab pc with ubuntu 13.04 and vrep 3.1.0, but the same scene running on virtual machine (my laptop) with ubuntu 14.04, vrep 3.1.2 didn't work. However I can boot that same ubuntu 14.04 installation on my laptop natively, and it works.

If you can give me access to a 3.1.0 linux 64 release, I can check if when running in a virtual machine I experience the same problem or not, but it is likely that the problem has always been there.