The submitted read was completing almost as quickly as it took to return from BeginRead() so it was never in the PendingMaterializes map for collection. Consequently the OnRead() callback wasn't able to find the read in the PendingMaterializes map.
Implemented and preflighted by martin.ridgers, reviewed and submitted by pj.kack.
#rb pj.kack
#rnx
#[robomerge][valkyrie] dev-valkyrie
#ushell-cherrypick of 28340644 by Martin.Ridgers
[CL 28343041 by pj kack in ue5-main branch]
#rb pj
#rnx
#jira
(git:ae45cf1 - Martin Ridgers - 2023-09-27 22:10:32 +0200)
#ushell-cherrypick of 28290419 by Martin.Ridgers
[CL 28337121 by martin ridgers in ue5-main branch]
#rb PJ.Kack
#rnx
- Flipped cvar 's.IasGenerateOnDemandToc' to default to true.
- Added a specific line of logging when generating the OnDemandToc from disk.
[CL 28337057 by paul chipchase in ue5-main branch]
Remove children array of set from Dynamic Collection, it is now using the bit array from its parent.
#rb cedric.caillaud
[CL 28331755 by vincent robert in ue5-main branch]
- add setting struct to FMeshContactGenerator
- change capsule collision to use FMeshContactGenerator with custom settings
#rb vincent.robert, jaco.vandyk
[CL 28327309 by chris caulfield in ue5-main branch]
- Added support for sending attachments flagged as SendImmediate from PostDispatchTick
#jira UE-122456
#rb Peter.Engstrom
[CL 28318795 by mattias hornlund in ue5-main branch]
The max number of huge objects in transit is controlled via the cvar net.Iris.ReplicationWriterMaxHugeObjectsInTransit, defaulting to 16, but will be at least one. There's a tradeoff mainly between the connection characteristics to support and normal object replication scheduling when tweaking this value. On one hand you don't want to end up stalling object replication because the top priority objects are huge. So you want to be able to keep replicating huge objects during the maximum latency, including latency variation, and packet loss scenarios you want to provide the best experience possible for. On the other hand object deletion cannot be performed once the object is in the huge object queue. Consider this and how long time it will take to replicate the huge object queue depending on the average payload of a huge object.
#jira UE-123370
#rb mattias.hornlund
[CL 28283681 by peter engstrom in ue5-main branch]
E.g., Objects would pop out of the ground even when InitialDepenetrationVelocity was set to zero. We don't save solver results for contacts that did not generate an impulse so we were throwing away initial penetration data. We must calculate and store MinimumInitialPhi on the previous frame when the data is still available.
#rb none
[CL 28278627 by chris caulfield in ue5-main branch]
