-
Notifications
You must be signed in to change notification settings - Fork 450
fix: NetworkSceneManager exception when DontDestroyOnLoad NetworkObjects are being synchronized #1090
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
fix: NetworkSceneManager exception when DontDestroyOnLoad NetworkObjects are being synchronized #1090
Conversation
This should fix all scenarios with one scenario causing a warning message that might not be needed but I am leaving it in until I get a consensus on handling the scenario where a NetworkObject "will" be moved to the DontDestroyOnLoad scene but during the initial syncrhonization it does not exist.
Changed some of the names used, added and updated comments, and applied the changes to the assets based on the name changes.
…/github.com/Unity-Technologies/com.unity.netcode.gameobjects into fix/develop-dontdestroy-networkscenemanager
@@ -245,13 +245,41 @@ internal void SetTheSceneBeingSynchronized(int serverSceneHandle) | |||
// If the scene was not found (invalid) or was not loaded then throw an exception |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Maybe removing this comment since we are not throwing anymore
if (!SceneBeingSynchronized.IsValid() || !SceneBeingSynchronized.isLoaded) | ||
{ | ||
throw new Exception($"[{nameof(NetworkSceneManager)}- {nameof(ScenesLoaded)}] Could not find the appropriate scene to set as being synchronized!"); | ||
// Let's go ahead and use the currently active scene under the scenario where a NetworkObject is determined to exist in a scene that the NetworkSceneManager is not aware of |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
So in the older code, we halted and caught fire if 1) the scene being sync'd was invalid or 2) it was not loaded.
Case 1) Seems like we want to still halt and catch fire for situation 1). Changing to some other scene would seem to mask a problem that needs to be dealt with. I mean, why would we think we would want to pop back to the active scene if the scene we're going to is invalid.
Case 2) same kind of thing; if the scene is valid but not loaded yet, perhaps there is another handling - I'm just not sure why we'd just pick whatever scene happens to be loaded now
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is specific to in-scene placed NetworkObjects during the client side deserialization phase for:
- Newly joined client synchronization where all scenes are already loaded and the client is just running through the in-scene placed NetworkObjects (deserialization) to be instantiated.
- A level is loaded and the client is synchronizing in-scene placed NetworkObjects.
The only time the offending method (SetTheSceneBeingSynchronized) is truly pertinent is when you have the same additive scene loaded multiple times and there are in-scene NetworkObjects in each instance that need to be synchronized. We use scene handlers to distinguish between "duplicate in-scene NetworkObjects" in order to "find the right NetworkObject" so we can apply the appropriate NetworkObjectId to that specific instance.
For all other cases, the GlobalObjectIdHash should be unique and will not matter really which scene is the current scene being synchronized.
What scenario would cause us to use the currently active scene?
If you set NetworkManager.DontDestroy to false and then have a NetworkObject that moves itself into the DontDestroyOnLoad scene when it is instantiated, then when a client is trying to "synchronize" to this specific scenario the first NetworkObject the client is trying to deserialize with this kind of behavior will not have a DontDestroyOnLoad scene (yet) and so it is perfectly fine to return the currently active scene. Upon that first NetworkObject instantiating (after deserialization) the DontDestroyOnLoad scene will exist so any other NetworkObject being deserialized (with the same behavior) on the client from that point forward will set the DontDestroyOnLoad scene to be the "scene being synchronized".
It is one of those chicken or the egg scenarios:
DontDestroyOnLoad is only created when you move a GameObject into that scene using the GameObject.DontDestroyOnLoad method. If nothing has been moved into the DontDestroyOnLoad scene, then it wont exist and isn't something that can be instantiated. I had thought about adding a chunk of extra code that would tell the client to "pre-seed" the DontDestroyOnLoad scene (i.e. create a temporary GameObject before deserialization and then destroy it afterwards) but that seemed a bit over the top. :)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think I may just need to trust you on this one, not sure I can parse this out / come to understand all the related systems as quickly as we need
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This was refactored a bit and might make more sense at this point. But yeah.., there are a lot of different possible scenarios to consider for sure.
} | ||
else | ||
{ | ||
// The next scenario is if a user intentionally moves NetworkObjects into a DontDestroyOnLoad scene |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
When we say "user intentionally moves NO into a DDOL", is this like the user is manipulating objects manually in the editor while things are running? I guess I didn't think of this as something we needed to explicitly support. Or is this if code moves a NO in the DDOL scene?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It is if the user's code intentionally moves it into the DDOL scene.
If you had a NetworkBehaviour that had this code in it:
/// <summary>
/// When enabled, we move ourself to the DontDestroyOnLoad scene
/// </summary>
private void OnEnable()
{
DontDestroyOnLoad(this);
}
Then it would get moved into the DDOL Scene when it was instantiated... and that would apply to both dynamically spawned and in-scene placed NetworkObjects.
testproject/Assets/Tests/Manual/DontDestroyOnLoad/DontDestroyOnLoadSpawnHandler.cs
Show resolved
Hide resolved
testproject/Assets/Tests/Manual/DontDestroyOnLoad/ObjectToNotDestroyBehaviour.cs
Show resolved
Hide resolved
com.unity.netcode.gameobjects/Runtime/SceneManagement/NetworkSceneManager.cs
Show resolved
Hide resolved
com.unity.netcode.gameobjects/Runtime/SceneManagement/NetworkSceneManager.cs
Show resolved
Hide resolved
Removed the NetworkObject property and assignment code from the PlayerMovementManager as it is not needed. Updated comments in NetworkSceneManager regarding DDOL.
else // Otherwise, we have to create a GameObject and move it into the DDOL to get the scene | ||
{ | ||
|
||
#if UNITY_EDITOR || DEVELOPMENT_BUILD |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Use UNITY_INCLUDE_TESTS instead of these...
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@JesseOlmer I am going to fix those in PR-1080 if that is ok?
/// <summary> | ||
/// Handle cleaning up the DDOL GameObject | ||
/// </summary> | ||
public void Dispose() |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Is this called anywhere I could not see it being called?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
When it is destroyed?
public class NetworkSceneManager:IDisposable
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is removed with the most recent commit.
#endif | ||
// Create our DDOL GameObject and move it into the DDOL scene so we can register the DDOL with | ||
// the NetworkSceneManager. | ||
m_DDOLObject = new GameObject("DDOL-NWSM"); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
can we just cache off the scene handle and then destroy the object right away? thus not polluting the scene with a useless object ?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Let me see if that works.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It indeed did work to do it that way! ⚡ ZAPPED that code out of there!
…nsform * develop: (26 commits) fix: client connected InvokeOnClientConnectedCallback with scene management disabled (#1123) fix: removed `public` class `NetcodeObserver` (MTT-1157) (#1122) feat: add NetworkMessageSent/Received metrics (#1112) feat: snapshot. MTU sizing option for Snapshot. MTT-1087 (#1111) Add metrics for transport bytes sent and received (#1104) fix: Missing end profiling sample (#1118) chore: support standalone mode for netcode runtimetests (#1115) feat: Change MetricNames for a more complex value type (#1109) feat: Track scene event metrics (#1089) style: whitespace fixes (#1117) feat: replace scene registration with scenes in build list (#1080) fix: mtt-857 GitHub issue 915 (#1099) fix: NetworkSceneManager exception when DontDestroyOnLoad NetworkObjects are being synchronized (#1090) feat: NetworkTransform Custom Editor Inspector UI (#1101) refactor: remove TempGlobalObjectIdHashOverride (#1105) fix: MTT-1124 Counters are now reported in sync with other metrics (#1096) refactor: convert using var statements to using var declarations (#1100) chore: updated all of the namespaces to match the tools package change (#1095) refactor!: remove network variable settings, network behaviour cleanup (#1097) fix: mtt-1088 review. Safer handling of out-of-order or old messages (#1091) ... # Conflicts: # com.unity.netcode.gameobjects/Prototyping/NetworkTransform.cs
…cts are being synchronized (Unity-Technologies#1090) * fix This fixes the dont destroy on load issue in NetworkSceneManager * test This provides manual tests to verify that don't destroy on load will work when NetworkManger's don't destroy property is enabled or disabled. * fix This should fix all scenarios with one scenario causing a warning message that might not be needed but I am leaving it in until I get a consensus on handling the scenario where a NetworkObject "will" be moved to the DontDestroyOnLoad scene but during the initial syncrhonization it does not exist. * style and refactor Changed some of the names used, added and updated comments, and applied the changes to the assets based on the name changes. * test Adding the unit test to verify the baseline DontDestroyOnLoad scene synchronization is working when NetworkManager.DestroyOnLoad is enabled and disabled. * style updating comments based on PR suggestions. * fix Fixing renamed method. * fix Verifying alternate DontDestroyOnLoad test case mixed with checking when we can or cannot send RPCs to a client. * refactor Cleaning up our DDOL GameObject. * refactor and style Removed the NetworkObject property and assignment code from the PlayerMovementManager as it is not needed. Updated comments in NetworkSceneManager regarding DDOL. * refactor We don't need to keep the DDOL object around after we have the DDOL scene. Removed IDisposable from NetworkSceneManager Removed the private m_DDOLObject Removed the Dispose method. Co-authored-by: M. Fatih MAR <mfatihmar@gmail.com>
MTT-1136
This fix pertains to the following GitHub-1088 issue.
This fix includes a validating manual test check for Github-701 issue in testproject\Assets\Tests\Manual\SceneTransitioning\SceneTransitioningTest that demonstrates how the NetworkSceneManager.OnSceneEvent can be used to detect when clients are ready to receive RPCs.