The reason why the BVTree's appeared under the polygon was because we were rendering and calculating them incorrectly. After fixing this, they now appeared above the polygons. This offset is not necessary as it causes BVTree's to build incorrectly.
Larger levels require watershed as some complex geometry (buildings with multiple floors) will exceed the maximum number of allowed sweep scans when building regions.
dtNavMesh::getTileCount returns the total number of added tiles, the issue with this is that if the navmesh gets pruned, that means disabling unlinked polygons and removing dead tiles, then some tile instances in the m_tiles array may become empty. Less tiles will be installed and since there are null tiles in between, we won't be traversing each tile.
Deduplicated mesh load function and implemented a simple type handler instead. Code can now load OBJ and PLY geometry from the user interface and project files.
dtNavMeshQuery::findNearestPoly can return an error flag, and the function cal also succeed while returning a NULL polyref. Both need to be checked before the algorithm is called. Also moved the memory allocation for flags inside this new check to avoid allocating for nothing.
These patches are no longer needed; the server and client already have the fatal error flag removed (process continues instead of hard erroring since script errors aren't fatal to the engine). The server also has its own routine within the function so client code never gets called.
The server needs these lumps in order to work. Any other lump not marked as SV should not be packaged into server VPK's as this will increase the install size (the server never parses them, as it doesn't need to render light maps for example).
Add "mp_rr_desertlands_holiday", remove "mp_rr_desertlands_64k_x_64k_tt". "mp_rr_desertlands_holiday" is the same as "mp_rr_desertlands_64k_x_64k_tt", but the BSP originates from a later version of the game, and the map has been fully ported over to R5Reloaded rendering "mp_rr_desertlands_64k_x_64k_tt" redundant.
Changing the load/unload system to FILO caused a crash when loading 3 level paks, either through request load or level settings, and then unloading them in reverse order. This is a regression caused when the pak loading and unloading system was reworked. Changing the system back to FIFO fixes the issue.
IsEULAUpToDate will always fail on dedicated servers as there is no user interface to display and accept the EULA (dedi's can also be ran without a terminal). The dedicated server users agree with the EULA upon launching the dedicated server.
Typically for dedicated servers you don't want any message boxes as they hang the process. The '-nomessagebox' helps solve this problem. Also removed extraneous check on MessageBoxA in EngineLoggerSink, we should always terminate regardless if MessageBoxA fails or not. The error will always be logged anyways.
Always terminate the process, MessageBoxA hangs the process anyways. If for some reason MessageBoxA fails then it will still terminate rather than continue and crash.
If the crash occurs during the initialization of systems, and Tier1 wasn't initialized yet, the system will display a prompt asking the user whether or not to send the reports. If the crash occurs after Tier1 was initialized, the system will check if the user has crash reporting enabled, and if the user has accepted the EULA before submitting it automatically.
The engine executable has been patched to no longer install this, therefore this function will never be called as we install a vectored exception filter anyways. This solves the issue where the crash message application gets called more than once when a crash occurs before the SDK detours are initialized.
Increase nesting limit to 128, this should make sure it will always be enough while also making sure no infinite recursions occur. Also improved error print formatting.
New command '_setClassVarClient' only sets the class var on the client, similar to how '_setClassVarServer' only sets it on the server. Command 'set' still has its old behavior, setting it on the server and client. The netmsg SVC_SetClassVar now only runs the _setClassVarClient handler, and if the netmsg was successfully sent, the code runs the _setClassVarServer handler directly instead of relying on the client to NET_StringMsg it in. This is a lot more reliable, secure and performant than the previous approach and drops the requirement of setting sv_cheats to function.
Netmessage SVC_SetClassVar allows the server to change class settings securely on the client. This was implemented due to popular demand, and previous approaches using a combination of ClientCommands (Cbuf and NET_StringCmd) were deemed insecure and unreliable.
First change callback of cvar 'mp_gamemode' must be removed as its client only, calling this on the server causes a crash as we never initialize client specific systems, therefore, the globals used in the change callback are null.
