You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I recreated the bug with logging and found these lines:
[2018.01.10-01.08.25:449][ 0]LogTemp: Warning: Getting buffer of size 1048576
[2018.01.10-01.08.25:449][ 0]LogHolodeck: Subscribing sensor PrimaryPlayerCamera for uav0
It seems 1048576 should be height * width * sizeof(FColor).
FColor is 32 bits so that size is wrong: 480 * 640 * 32 = 9830400 so the allocated buffer is far too small.
Also, we need to have functionality to change the buffer size in Python. @mitchneutron and I discussed changing enums to individual sensor classes with member functions, instead of dictionaries from enum to value.
Handing over to @mitchneutron since the PrimaryPlayerCamera buffer is something he worked on before.
The crash due to the viewport size being different than expected should now be fixed. The fix is in the viewportBug branch of Holodeck. I'm still trying to narrow down the allocated buffer size bug.
Changing:
UrbanCity UAVAgent uav0 512 512 PrimaryPlayerCamera OrientationSensor LocationSensor VelocitySensor IMUSensor
to
UrbanCity UAVAgent uav0 480 640 PrimaryPlayerCamera OrientationSensor LocationSensor VelocitySensor IMUSensor
results in a silent hang - looks like the holodeck subprocess has died.
The text was updated successfully, but these errors were encountered: