Slightly mapped out the structure more.
The function should now be compatible with earlier game dll's.
CStudioHWDataRef::SetFlags has to be rebuild as these are declared inline for earlier game dll's.
Note: this does not work reliably still.
The only way we could make this work 100% reliable would be to fully rebuild '0x140341D40' in the SDK and load our pak files among with the pre-existing g_pakLoadApi->AsyncLoad() call in this function, as this will ensure everything will be ran synchronously.
The current approach by taking some JT fifolock wrapper will only work reliably between one level change, unsure why the second and up fail.
Fixed rare crash in CMDLCache when buffer itteration returns invalid pointer (not null!) for handle.
Made all variables and functions used by BuildPropStaticFrustumCullMap searched dynamically (these should also work on s0, s1 and s2, they are compared but at the moment untested).
TODO: Fix deref crash properly (see bsplib.cpp l291).
Only works for prop_static (for the time being)
Only works for the S3 game_dll.
In order to get this to work: you must set 'old_gather_props' to nun-null and (for certain maps) perform the patch to prevent the frustum culling code to deref a bad pointer (see command 'opcodes test').
In 'datacache/mdlcache.cpp' the function 'CMDLCache::FindMDL' attempts to find 'mdl/error.rmdl' and assigns the studiohdr and handle to the members of CMDLFallback.
In 'CMDLCache::FindUncachedMDL' we check if a model exists, if a model does not exist, we replace the studiohdr with the one of error.rmdl we stored in the CMDLFallback structure.
This does actually work (on the dedicated server it doesn't crash at all!), but on the client it crashes when trying to gather props (right before rendering), setting the ConVar* 'old_gather_props' does interesting things (check IDA around this ConVar*). setting this to 1 causes it to crash in another CMDLCache method when trying to deref something in the global cache pool.
This method is easy to rebuild. I will do this soon and attempt to return error.rmdl parts from here as well if results are null (this might actually just work).
Leaving 'old_gather_props' to 0 causes it to crash in the middle of the function which is supposed to gather the props the 'new' way.
The gather props functions are kind of daunting
NOTE: Currently only confirmed to work somewhat properly on the dedicated server for prop_static. prop_dynamic is unconfirmed. And (almost?) works on the client.