search

dkavolis / Ferram-Aerospace-Research

Aerodynamics model for Kerbal Space Program
Other
80 stars 32 forks source link

Voxelisation happens correctly for exactly one launch, then fails thereafter (0s and NaNs everywhere) #119

Open aogden41 opened 3 years ago

aogden41 commented 3 years ago

Hello,

Running a fresh install of RO and the latest FAR binary from master branch. The install is than 24h old, and I've got some strange behaviour from FAR. I have also tried the release version 0.16.0.2 and also the previous version 0.16.0.1, same behaviour.

It appears that the voxelisation doesn't work properly after the first launch - sets everything to zero (drag coeffs, mach 1 wave drag, etc) and gets stuck on "Voxel waiting for chunks to be released". Removing the B9 winglets I had on my rocket unblocked this a little bit, but I was still getting no change on the FAR calculations and it does not report the FAR values in the console (such as CrossSection area and standard smoothing dev).

Additionally, the Display Voxel Debug button breaks as well and returns a nullref.

I am running KSP 1.10.1 with RO/RP1. It is a fresh steam install, and I made sure to clear my old one out before I re-installed.

Here is my log file: https://drive.google.com/file/d/1CzhubL87Cp1R1wJHPO0xYbct3vkT7s-l/view?usp=sharing

This is what happens when voxelisation fails (see 'vapor in feedlines' assumingly because it detects drag as zero). On the first launch, this rocket launches correctly. I tested using identical rocket models with exactly the same delta V and aerodynamic profile. image

This is what the FAR analysis window says in the VAB. On the first launch, this is populated correctly. image

I assume at this point that it's another mod fighting FAR to the death for superiority over the aerodynamic system, but I am not sure. I attached my mod list. I also found a csv file in the FAR logs from this morning (not same as log file) which demonstrates the behaviour of the aerodynamics (0 drag, 0 lift, etc. I changed extension to txt so github would accept it).

mods.txt Orion 1_2021_06_05_09_33_22.txt

dkavolis commented 3 years ago

Can you upload MM cache as well?

aogden41 commented 3 years ago

This one? https://drive.google.com/file/d/1A-amUqu9_PNTDUyeTfR96LCLjN5unIXr/view?usp=sharing

aogden41 commented 3 years ago

Any luck? @dkavolis

Alioth81 commented 3 years ago

@andrewogden1678 Do you have debug voxel display in flight scene enabled? Does this error also occur if you disable debug voxelation in flight?

I experienced a similar error when I tried to verify that some meshes are excluded from voxelation. I think the debug voxelation in flight gets stuck somewhere and waits for "chunks to be released". I did not bother to investigate further as disabling "debug voxelation" in flight did solve it for me. However I agree that if you have debug voxelation enabled in flight it can happen that voxelation is getting stuck and no aerodynamics are calculated. But I do not no how to reproduce it reliably. It can happen after one launch, but I also had it happen when just pressing the debug voxel button, when loading f the vessel was completed (with other vessels present) or having a part explode (which probably also triggers voxelation).

aogden41 commented 3 years ago

Sorry, forgot to update on this. It turns out it was the latest version of B9 Procedural Wings not playing nicely with FAR. I rolled back one version (0.91 I think?) and copied in the FAR GameData folder from the master branch and it works beautifully and voxelises properly.