This allocated some memory and never freed it again. A TArray would have been better - but since we know the maximum size is 4 we may just use a static array here to keep things as efficient as possible.
Due to C++ conversion rules this was a bit too volatile. There's really not enough places where being able to pass a string directly into the sound API was beneficial - the two most frequent functions now got overloaded variants.
This works similar to MAPINFO: The first definition in a file declares the format, all later ones must be the same.
This change in syntax increases robustness significantly because it avoids the problems with the original syntax not being able to detect badly formatted names.
*Added Font and TextureID types to Option and Image Scroller items (List items already have these)
*Added "Class" instruction for Image Scrollers
*Added ability for Option items to read in their OptionMenuDescriptor (List and Image Scroller items already have this)
*Added "ForceList" instruction for Lists so that skill, playerclass, and episode menus don't get overridden
Changed SendConsoleEvent to SendInterfaceEvent to make functionality clearer. Added InterfaceProcess virtual to EventHandlers. Added CCMD for sending interface events.
*Added ONLYSLAMSOLID flag to ignore non-solid Actors that aren't shootable while SKULLFLY is on.
*Added Slam state that gets entered when slamming an Actor.
This is mainly for new games which have been designed for hardware rendering.
Note that this does not remove the software renderer's code - all it does is to disable the vid_rendermode CVAR by turning it into a static constant with the value 4.
- -noextras: disable loading of any extras (lights, brightmaps, widescreen)
- -nolights: disable loading of lights
- -nobrightmaps: disable loading of brightmaps
- -nowidescreen: disable loading of included widescreen assets
- Have a real finalizer stage for destroying objects instead of mixing them
in to the sweep stage.
- Base GC rate on a running average of the allocation rate instead of basing
it on the amount of time taken since the last sweep started.
- More GC stats for better tweaking.
- Improved performance of YZ axis swap matrix calculations by calculating them with the initial matrix calculations.
- IQMs with no bones will no longer crash
* use a different subfolder for each IWAD's saves.
* do not allow load and save CCMDs to escape the save folder. Absolute paths and '..' are being blocked now.
* unified savegame path and filename generation in one single function. All ad-hoc file name generation was replaced.
* -loadgame will also use the designated savegame folder now.
This matches modern software design standards, if a portable install is desired a config named "gzdoom_portable.ini" needs to be created in the program directory.
This may be an empty file for a first launch.
* deleting some unused code
* turned several class methods into static local functions in cases where they never were used outside this file.
* inlined the dangerous assignment operator in the only place where it was used.
There was one special case allowing to let an actor die on spawn, but this could call script code on an incompletely set up map which resulted in crashes.
- Removed bone manipulation code
- Implemented an index in calculateBones to optimize multi-armature actors
- Moved the bone storage object's creation to RenderModels so that the armature array can be sized there
- Factored in parent bone rotations to check if a bone needs updating
- Implemented multiply Quaternion functions to TVector4
- Converted Euler rotations in A_ManipulateBone to degrees
- Refactored IQM and calculateBones to process TRS at runtime which resolves some of the faulty animations with large rotations. Will also make bone manipulations much easier to do
As it currently stood, only traces that never found anything in traversal had their distance and final position corrected.
Meanwhile, traces that skipped everything would return the distance and position of the last thing they crossed instead.
This change makes both cases consistent by "filling out" the remainder of the trace line for the latter.
This solves two problems:
* The linked list is too slow, a map is better. A map cannot be used with statically allocated CVARs because order of initialization is undefined.
* The current CVAR system is an unordered mishmash of static variables and dynamically allocated ones and the means of identification are unsafe. With this everything is allocated on the heap so it can all be handled the same by the cleanup code.
