- Fixed issue where only client would print default state of cdo used to initialize network state.
#rb LouisPhilippe.Seguin
[CL 36760738 by mattias hornlund in 5.5 branch]
CVD already attempted to stop the active recording in these situations, but there are cases where the active session might take more than one frame to disconnect, causing the error.
This CL also adds a way to restore the trace channels CVD disabled at the start of the recording
#jira UE-224727
#rb Benn.Gallagher
[CL 36759591 by sergio gardeazabal in 5.5 branch]
CVD needs to map each inner implicit object to a valid shape instance index so it can properly colorize and show the correct per shape data in the details panel.
We were already handling the case for nested Implicit Object unions, where the objects in the inner union share the same shape instance index, but we need to do the same for union clustered objects that might be the root.
#jira UE-225002
#rb Benn.Gallagher
[CL 36759353 by sergio gardeazabal in 5.5 branch]
- Added support for separate indices per NetTokenDataStore to make it possible to use fewer bits for the index for types with few exports.
#rb Peter.Engstrom
[CL 36757497 by mattias hornlund in 5.5 branch]
- HasDynamicInternalClusterParent and HasClusterUnionParent are mutually exclusive and testing both was resulting in getting always false for computing HasDynamicInternalClusterParent
[CL 36747843 by cedric caillaud in 5.5 branch]
- Fix out of bound related crashes when using the select by box node ( when there was more transforms than geometry in the collection )
- now properly access the centroid array by geometry index
- StaticMeshToCollection node now outputs collection with a single root when using the Split Element option
#rb gustav.melich
#jira UE-225106
#p4v-cherrypick 36466212
[CL 36485897 by cedric caillaud in 5.5 branch]
Add support for non-nanite fallback mesh for nanite geometry collections.
Strip geometry collection nanite render data when cooking non-nanite platforms.
#rb cedric.caillaud
[CL 36439857 by jeremy moore in 5.5 branch]
* Added PostInit callback to UNetObjectFactory
* Moved PostInit code from bridges to their respective factory
#rnx
#rb Mattias.Hornlund
[CL 36390399 by louisphilippe seguin in 5.5 branch]
- Enabled use of NetTokens/NetTokenStore and NetTokenDataStores outside of Iris to facilitate exporting data from shared serialization paths
- NetTokenStore and NetTokenDataStores are now owned by NetDriver.
- Added support for CustomExport bunches to carry data assoiciated with NetTokenExports
GameplayTags
- GameplayTagNetSerializer now respects settings from GameplayTags.ini
- Implemented placeholder GameplayTagNetTokenDataStore.
GameplayTags are now using NetTokens to export data if running in the new dynamic mode.
#rb jodon.karlik, Ryan.Gerleve
[CL 36380672 by mattias hornlund in 5.5 branch]
- Implemented widget to render all recorded logs
- Enabled the log trace channel in CVD
- Implemented a search bar and category & verbosity filter
- Search and other filter application is async
- Implemented two-way selection system. If you scrub to any given frame, all the log lines for that game frame or solver frame will be selected and the log window will scroll automatically to the line. If you click on any log line, CVD will jump to the closes recorded game frame for the selected log line.
#jira UE-203462
#rb Benn.Gallagher
[CL 36365541 by sergio gardeazabal in 5.5 branch]
Softs Evolution doesn't calculate Aerodynamics forces in two passes, so the Forces buffer is not calculated. Instead just calculate the force based on the current state. This will represent the force for next frame instead of current frame, but most of our debug display shows based on current (end of frame) state rather than last frame, so this is consistent with that.
Also add a user-scale menu option and set the Simulation Statistics debug display to true, and fix a DPI scale bug in the text visualization.
#rb kriss.gossart
#jira UE-224203
[CL 36352223 by alex mcadams in 5.5 branch]
Fix nanite geometry collection not visually rotating on gizmo driven rotations.
Needed to reapply modified instance transforms after primitive transform update.
This is because, for nanite, Geom Collection GPU instance transforms are stored relative to the relative (non-rotated) world transform.
This bug wasn't visible for non-gizmo use movement cases which go through full scene proxy recreate.
This bug wasn't visible for gizmo translation use case because this is in the relative transform.
#rb cedric.caillaud
[CL 36344776 by jeremy moore in 5.5 branch]
- Rely on FRewindData for both rewind and stepping through non-resim objects histories so we don't only rely on the ResimCache which holds duplicate data.
- Deactivate FEvolutionResimCache by default, API in FPhysicsSolverBase to toggle on and off and the normal EnableRewindData call will enable it if parameter for it is true.
- Put in CVar, enabled by default, to be able to disable FPushPhysicsData dirty proxies from being cached in history: p.Resim.CachePushDataDirtyProxies
- Put in CVar, enabled by default, to be able to disable FPushPhysicsData async inputs from being cached in history: p.Resim.CachePushDataAsyncInputs
- Put in CVar, enabled by default, to optimize RewindData: p.Resim.RewindDataOptimization
- Mark islands for resim if they have particles in them marked for FullResim
- Take actual physics tick-rate into account for physics tick buffering and rewind data history size, instead of value in project settings.
#rb tom.waterson
[CL 36341150 by markus boberg in 5.5 branch]
