The reason this was never added was the hard dependency on the line trigger types. This implements some modified logic that does not try to find all potential lifts in the map.
Also moving the MBF flags to compatflags so that they are easier to control by the user as these must be part of compatibility presets.
This way it can be directly used as a native ZScript export.
Like SuggestMissileAttack the change to a method was for virtual overrides that have been turned into flags since then.
This cannot use the last color found while reading ahead - it must pick the last color of the string part that just got broken out into its own line when starting the next one.
Especially in the thinker code this is needed for software rendering.
Strictly speaking, the software renderer should do the same, but it checks r_dynlight in so many places deep in the logic where the level is not available.
Status bar was already freed while `FCajunMaster::RemoveAllBots()` function may still access it
Status bar is destroyed a bit later in `P_Shutdown()` function
https://forum.zdoom.org/viewtopic.php?t=72843
Prompted by 'Hurt' which has > 4000 lights and runs into both performance issues and unpredictable light selection for the limited amount of shadowmap slots.
* added handlers for the missing upper flags in the first flag word.
* refactored some code related to MF_BOUNCES to adapt to other flags' changes to avoid constant property updates for too many flags.
* treat anything with RF_ZDOOMTRANS as non-translucent for the purpose of the flag checks.
* let CASTSPRITESHADOW act as an override for render style checks. If this is set we can assume that a shadow is desired regardless of style.
* reordered code to do the more costly checks only when needed.
* exclude sprites which are not either opaque or use regular blended translucency from casting shadows.
* exclude wall and flat sprites from casting shadows.
In both of these cases the shadow math is unable to produce anything useful that doesn't look off.
src/common/engine/serializer_internal.h:241: warning: format specifies type 'long long' but the argument has type 'int64_t' (aka 'long') [-Wformat]
src/common/engine/serializer_internal.h:250: warning: format specifies type 'long long' but the argument has type 'int64_t' (aka 'long') [-Wformat]
src/common/rendering/vulkan/system/vk_device.cpp:364:29: warning: format specifies type 'unsigned long long' but the argument has type 'uint64_t' (aka 'unsigned long') [-Wformat]
src/common/rendering/vulkan/system/vk_framebuffer.cpp:701:57: warning: format specifies type 'unsigned long long' but the argument has type 'VkDeviceSize' (aka 'unsigned long') [-Wformat]
Due to how the MBF21 flag checkers work the first flag word may not repurpose any unused flags in it. We still need MF_SLIDE to be usable by them, despite being totally unused.
src/playsim/p_maputl.cpp:1691:42: error: too many arguments provided to function-like macro invocation
src/gamedata/d_dehacked.cpp:2385:62: error: cannot pass non-trivial object of type 'FString' to variadic function;
A dangling pointer in corpse queue may appear if actor is added to the queue when GC is in propagation state.
Enqueued corpse actor remains white, and if it’s destroyed and garbage collected before dequeue, a dangling pointer will be accessed during its removal from the queue.
In console, do `summon CorpseSpawner` and `gc now` with the following script loaded. Without a write barrier, it will crash in two seconds.
```
class TestCorpse : Actor
{
States
{
Spawn:
POSS U 1 A_Die;
Death:
POSS U 1 A_QueueCorpse;
Stop;
}
}
class CorpseSpawner : Actor
{
override void Tick()
{
A_SpawnItem("TestCorpse");
}
}
```
https://forum.zdoom.org/viewtopic.php?t=69842
These extend the parameter set of Scroll_Texture_Offsets, so in Hexen/UDMF format the advanced scroll types are also available for the tag-less variant.
While 'block players' was just a simple remap, 'block land monsters' required quite a bit of work...
This also fixes the bug that BLOCK_FLOATERS was never checked for sliding against a wall.
When built against 10.15 SDK or earlier, macOS 11.0+ returns 10.16 for compatibility, and external process is needed to output the actual version for us
There is no `TVector2<>` constructor that accepts a pointer to float. However, there is such constructor in `TVector3<>`, so `TVector2<>` can be constructed from `float*` implicitly via temporary `TVector3<>` object.
- Just set init for SoundClass to empty.
- Removed code block from SetActorProperty for APROP_Soundclass that does nothing
- Lower-cased soundclass in FSerializer
- Created a new const char to read the player's soundclass. If the playerpawn returns NAME_None for it's default, then it will set defaultsoundclass to "player". After running the skin code, the function now returns defaultsoundclass or soundclass, depending if soundclass is empty or not.
- Renamed GetSoundClass to S_GetSoundClass
- SoundClass is instantiated to "" by default. Since this property is only used when it is not empty (otherwise GetSoundClass just defaults to player), we can get away with this.
- We may want the soundclass to remain the same if we explicitly set it to the same one that is currently used (say, we set SoundClass to "Caleb" so all other skins can use it)
- GetActorProperty for APROP_SoundClass just calls GetSoundClass,
- CheckActorProperty also just runs GetSoundClass
- GetSoundClass is no longer a static method. We needed to access it in other places.
- Made renovations to GetSoundClass. First of all, SoundClass is no longer instantiated there. Secondly, skinned sounds are now returned if SoundClass is empty. Thirdly, "sclass" in this method will return the default soundclass of the player pawn or SoundClass, depending on if SoundClass is empty. Finally, sclass will retrieve "player" if it is empty.
- I couldn't simply init SoundClass to NAME_SoundClass, even after converting it to the appropriate type. Probably because NAME_SoundClass hasn't been parsed from decorate yet. Instead, I change it to NAME_SoundClass through GetSoundClass if it's valid and currently "player".
- The skin checker code in GetSoundClass now checks if the SoundClass is equal to NAME_SoundClass. This mechanism exists so that way reverting the SoundClass to NAME_SoundClass processes the skin soundclass code. If it's different, the code is not processed.
- Just returns sclass. This is never null, so there's no need to check if so.
- S_FindSkinnedSound just uses GetSoundClass. This makes sure skins are checked.
- Gave default init value to SoundClass as "Player"
- Changed SoundClass detection to use the if/else structure
- Checked for null pointer in S_FindSkinnedSound when reading the player's sound class
It always reused the previous content of the 'Pos' field which was either undefined or an older position where the actor was located when last spawning a light.
This eliminates nearly all palette dependencies, most importantly font translation will now be done on True Color data, making translations on True Color font less destructive.
- useBloodColor: when true, the sprayed decal will be shaded to match the calling actor's blood color.
- decalColor: when defined, the sprayed decal will be shaded to the specified color. Note that this will take precedence over useBloodColor. It is recommended to use only one parameter.
Note that due to how decals work in the engine, the "decalColor" parameter will only properly colorize the decal if the image is grayscale.
This also gives the caller the information whether the LineAttack
hitscan passed through a portal regardless of whether it actually hit an
actor or not as a nice bonus.
This reverts commit 668f8f2cf6.
Revert "- added a proper error message when '--' or '++' gets used on 8 or 16 bit values."
This reverts commit a94f5dd1b3.
None of this was necessary - the triggered assert was bogus and had to be removed.
This requires SDL >= 2.0.4.
Since <https://bugzilla.libsdl.org/show_bug.cgi?id=2464> was fixed in
SDL 2.0.4, SDL behaves as a CMake "config-file package", even if it was
not itself built using CMake: it installs a sdl2-config.cmake file to
${libdir}/cmake/SDL2, which tells CMake where to find SDL's headers and
library, analogous to a pkg-config .pc file.
As a result, we no longer need to copy/paste a "find-module package"
to be able to find a system copy of SDL >= 2.0.4 with find_package(SDL2).
Find-module packages are now discouraged by the CMake developers, in
favour of having upstream projects behave as config-file packages.
This results in a small API change: FindSDL2 used to set SDL2_INCLUDE_DIR
and SDL2_LIBRARY, but the standard behaviour for config-file packages is
to set <name>_INCLUDE_DIRS and <name>_LIBRARIES. Use the CONFIG keyword
to make sure we search in config-file package mode, and will not find a
FindSDL2.cmake in some other directory that implements the old interface.
In addition to deleting redundant code, this avoids some assumptions in
FindSDL2 about the layout of a SDL installation. The current libsdl2-dev
package in Debian breaks those assumptions; this is considered a bug
and will hopefully be fixed soon, but it illustrates how fragile these
assumptions can be. We can be more robust against different installation
layouts by relying on SDL's own CMake integration.
When linking to a copy of CMake in a non-standard location, users can
now set the SDL2_DIR or CMAKE_PREFIX_PATH environment variable to point
to it; previously, these users would have used the SDL2DIR environment
variable. This continues to be unnecessary if using matching system-wide
installations of CMake and SDL2, for example both from Debian.
This is based on the patch written by Simon McVittie for OpenJK:
52030235f0
The wiki currently says they should be fixed point but the code
currently treats them as integers. Given that they are placed with
double precision horizontally, it is odd that the same precision
cannot be applied vertically. Changing this would break existing maps
so add two flags to support both types.
* Remove threading from the drawers
* Fix some r_scene_multithreaded related bugs
* Fix some r_scene_multithreaded crashes
* Fix fullbright shade bug
* Fix truecolor drawer crash
* Add debug code for showing the render slices
* Fix texture loading race condition and improve performance by only locking the load mutex if data hasn't already been updated for this frame
Co-authored-by: Magnus Norddahl <dpjudas@users.noreply.github.com>
Previously Vel would be reset only if the target was moving. This change
ensures that Vel is always reset, as seems to be the intent of the
oldvel parameter.
It wasn't respecting the map's pixel ratio setting. This wouldn't have
been very noticeable with the default pixel ratio of 1.2 but it was
very obvious with 1.8.
Both synth and sound font can be very relevant here for the final volume so using the same song with different settings needs to create different strings.
# draw upper and lower textures when the back sector does not properly match
# do not draw middle textures on portal lines
# minor optimization to 'is***Portal' functions to avoid memory access in the most common case of no portal being present.
Re-worded error messages which were unprecise or unfitting before (model index below 0 was not acknowledged at all, or grouped together with a "too many models" message).
modelIDs are given a default value of -1.
Important: A MODELDEF's FrameIndex lines can no longer refer to model indices that are beyond the number of models of that MODELDEF entry. There is in fact a check to avoid going beyond the number of an actor's models which would abort program operation at startup, but it never caught any such occurances.
surfaceSkinIDs was two-dimensional and is now a one-dimensional TArray as well, elements are accessed via [Row + Column * NumRows], in this case surfaceIndex + modelIndex * MD3_MAX_SURFACES]
Used TArray.Alloc to make TArrays have the correct size depending on the number of models.
Also removed MAX_MODELS_PER_FRAME.
Edited skinSurfaceIDs access for one-dimensional TArray
Added MD3_MODELS_MIN
To ensure compatibility with mods, all model-related TArrays (four in total) have a minimum size of 4, defined by MD3_MODELS_MIN.
The problem I discovered is that these backends do not respect the GUI-side system setting for swapping the buttons, so this needs to be done in-application so that swapped buttons can behave the same in the menu and the game.
