Kerbal Space Program

I made two quicksaves - one after seating the Kerbal from a ladder, another after seating the Kerbal from the platform. Loading either quicksave leads to the crash, regardless how the Kerbal got into the seat. Also there is no significant difference between the contents of the two quicksaves. This means that restoring the game state with a Kerbal in seat is also done the "wrong" way, similar to seating the Kerbal from a ladder.

Great work, and thanks for looking into this.

Kasuha wrote:

I produced two output logs. I made effort to use exactly the same steps when starting up the game and setting up the scenario, so making a diff of them is possible. Messages

[KerbalSeat]: Being Boarded by Jebediah Kerman
[KerbalSeat]: Boarded by Crasher2

appear in both of them. These seem to be ok. The difference really starts with "Invalid parameter because it was infinity or nan." message in case of crash. In case of no crash (seated from platform) there is the following exception instead:

NullReferenceException
at (wrapper managed-to-native) UnityEngine.Component:InternalGetTransform ()
at UnityEngine.Component.get_transform () [0x00000] in <filename unknown>:0
at PQSCity.OnSphereReset () [0x00000] in <filename unknown>:0
at PQS.Mod_OnSphereReset () [0x00000] in <filename unknown>:0
at PQS.ResetSphere () [0x00000] in <filename unknown>:0
at PQS.OnDestroy () [0x00000] in <filename unknown>:0
(Filename: Line: -1)

The "[KerbalSeat]: Boarded by Crasher2" is unusual as it only happens when a Kerbal boards a command seat. Boarding all other pods results in the Kerbal being correctly identified, and not the vessel boarding itself. I also see the NRE, but have another entry that looks suspicious, and has been attributed to other bad behaviour once triggered.

[LOG 16:02:02.360] [KerbalSeat]: Being Boarded by Jebediah Kerman
[WRN 16:02:02.361] [PartJoint]: None of the provided nodes was valid!
[LOG 16:02:02.363] [FLIGHT GLOBALS]: Switching To Vessel Untitled Space Craft ---------------------- 

Are you able to confirm this?

The vessel is nothing more than a Mk1 pod with a command seat on the side.

The "[KerbalSeat]: Boarded by Crasher2" is unusual as it only happens when a Kerbal boards a command seat. Boarding all other pods results in the Kerbal being correctly identified, and not the vessel boarding itself.

Pods and command seats are very different.

When boarding a command pod, the Kerbal becomes part of the pod's inventory. No physical force can get a Kerbal out of the pod, it either survives or dies in explosion. It does not even affect mass of the pod (or at least never did).

When boarding a command seat, the Kerbal becomes a new part of the ship, kind of like if it docked with it - a separate part with a joint between the Kerbal and the seat and this joint can be broken by physical forces.

The two processes are very different.

The sequence

[KerbalSeat]: Being Boarded by Jebediah Kerman
[PartJoint]: None of the provided nodes was valid!
[FLIGHT GLOBALS]: Switching To Vessel Crasher2 ----------------------
Camera Mode: AUTO
stage manager resuming...
[KerbalSeat]: Boarded by Crasher2

appears in both my two logs from my update 2 and the result (black screen crash when boarded from ladder, exception and normal continuing when boarded from platform) does not seem to depend on its presence.

Kasuha wrote:

The "[KerbalSeat]: Boarded by Crasher2" is unusual as it only happens when a Kerbal boards a command seat. Boarding all other pods results in the Kerbal being correctly identified, and not the vessel boarding itself.

Pods and command seats are very different.

When boarding a command pod, the Kerbal becomes part of the pod's inventory. No physical force can get a Kerbal out of the pod, it either survives or dies in explosion. It does not even affect mass of the pod (or at least never did).

When boarding a command seat, the Kerbal becomes a new part of the ship, kind of like if it docked with it - a separate part with a joint between the Kerbal and the seat and this joint can be broken by physical forces.

The two processes are very different.

The sequence

[KerbalSeat]: Being Boarded by Jebediah Kerman
[PartJoint]: None of the provided nodes was valid!
[FLIGHT GLOBALS]: Switching To Vessel Crasher2 ----------------------
Camera Mode: AUTO
stage manager resuming...
[KerbalSeat]: Boarded by Crasher2

appears in both my two logs from my update 2 and the result (black screen crash when boarded from ladder, exception and normal continuing when boarded from platform) does not seem to depend on its presence.

You are of course correct in your observation of the Kerbal becoming a part of the vessel via the kerbalEVA placeholder part. However, I was exploring the possibility of an incorrect root part being the cause of the later black screen lockup caused by NaNs. It may well be that the vessel boarding the vessel is just an artefact of the reporting text system that is coincidental to the problem. The invalid nodes are also unique to the seat, and I can't help speculate that these both play a factor during the rapid series of events during the explosion. By using different impact velocities in my tests, the black screen occurs around the time that the seat and the Kerbal would be destroyed. I have managed to destroy the whole craft - including seat - to be left with the Kerbal alone and no lock up ( I think this needs exploring further)

This does look like there may be a race condition during the explosion that is leaving the vessel state, and therefore its orbital state indeterminate which are causing the NaNs. From observation of the behaviour in previous reports that included NaN (not a number) events, the subsequent behaviour of the game and it requiring a restart, follows the same pattern. This is of course not desirable, even in edge cases like this.

I've made a note to use this as a test case for the upcoming testing. It's certainly noteworthy.