* 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.
* Changed model-related arrays to TArrays.
* Update models.cpp
Used TArray.Alloc to make TArrays have the correct size depending on the number of models.
surfaceSkinIDs was two-dimensional and is now a one-dimensional TArray as well, it's size is now (models * MD3_MAX_SURFACES)
surfaceSkinIDs was two-dimensional and is now a one-dimensional TArray as well, elements are accessed via [Row + Column * NumRows], in this case sSurfaceIndex + modelIndex * MD3_MAX_SURFACES]
* Edited skinSurfaceIDs access for one-dimensional TArray
* Edited skinSurfaceIDs access for one-dimensional TArray
* Edited skinSurfaceIDs access for one-dimensional TArray
* Changed model-related arrays to TArrays.
Also removed MAX_MODELS_PER_FRAME.
* Used TArray.Alloc to make TArrays have the correct size depending on the number of models.
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.
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]
* Update models.h
* Edited MAX_MODELS_MD3
* Update models_obj.cpp
The ConversationNPC and ConversationPC fields were not cleared if the conversation was closed by selecting the goodbye option, or if the conversation was closed from the item checks.
https://forum.zdoom.org/viewtopic.php?f=2&t=53655
This was the result of an issue which stumped the entire Discord, which led to me having to debug from source to find why some relatively standard publicly available PBR materials work. GZDoom is the ONLY program in the typical development stack (GIMP, Slade, UDB) with this narrow of support for the PNG format. As such, the average developer will have no other way to figure out what's going wrong without these: these CANNOT be allowed to fail silently. As things like PNG-compression and 64-bit color become more common in royalty-free PBR materials, support should be an eventual target. Even then, these warnings should remain to prevent this from being an issue the next time things change.
Prototype of overridden function with optional argument(s) missing could extend unrelated prototype of previously defined function when their arguments and return value(s) match
https://forum.zdoom.org/viewtopic.php?t=71340
- vid_aspect is now set to change to 21:9's ACTUAL aspect - 64:27
- screen elements now scale to 64:27 properly. to restore classic behavior, set cvar `vid_allowtrueultrawide` to false
There was one VMCall in DLevelScript::ScriptCall that didn't use default arguments. This was causing assertion failures in the GZDoom debug build when trying to run Wolfenstein: Blade of Agony on any map that wasn't TITLEMAP, because ACSTools.FindInventoryClass returns a boolean, and it has a default argument.
* Added SBSLETTERBOX to menudef.txt
* Changed PresentSideBySide in gl_renderer
Needed for new side-by-side mode
* Added VR_SIDEBYSIDELETTERBOX to gl_stereo3d.cpp
For side by side letterbox display mode
* Added VR_SIDEBYSIDELETTERBOX to hw_modes.h
For new side by side letterbox display mode
* Added VR_SIDEBYSIDELETTERBOX to hw_vrmodes.cpp
For Side By Side Letterbox display mode
* - modify Shape2D so that it manages its own vertex buffer
* - fix the last commit failing on Shape2Ds that were modified after AddShape was called
* - make buffers an array of F2DVertexBuffers instead of an array of pointers
* - fix AddShape with the same VBO but different transformation crashing the game
* - formatting fixes
- Always allow skyboxes by default
- Add option to disable skyboxes through mapinfo
Squashed commit of the following:
commit d4c4d9310d
Author: Marisa Kirisame <marisa@sayachan.org>
Date: Wed Dec 30 09:58:39 2020 +0100
Disable AO for skybox portals (can be forced back with MAPINFO flag).
commit bbfb934e80
Author: Rachael Alexanderson <madame-rachelle@users.noreply.github.com>
Date: Sun Dec 6 23:50:15 2020 -0500
- and now it builds! but ... does it run?
commit 67096c8966
Author: Rachael Alexanderson <madame-rachelle@users.noreply.github.com>
Date: Sun Dec 6 22:25:13 2020 -0500
- one more step closer to compile
commit 5a0c84dd2d
Author: Rachael Alexanderson <madame-rachelle@users.noreply.github.com>
Date: Sun Dec 6 22:13:39 2020 -0500
- set up dynamic opengl load for windows
commit 6ef8118b80
Author: Rachael Alexanderson <madame-rachelle@users.noreply.github.com>
Date: Sat Dec 5 16:38:53 2020 -0500
- allow compiling for win-arm64
commit 7d6f3797c4
Author: Rachael Alexanderson <madame-rachelle@users.noreply.github.com>
Date: Fri Dec 4 05:18:29 2020 -0500
- allow targeting ARM64 on Windows
Co-authored-by: Rachael Alexanderson <madame-rachelle@users.noreply.github.com>
With fixed point math barely being used anywhere with the vast majority of calls in the software voxel drawer it makes sense to reduce the function interface to the minimum possible and move the shift value into the function call as an argument.
include\algorithm(7419,17): error C3892: '_First': you cannot assign to a variable that is const (compiling source file src\rendering\hwrenderer\scene\hw_drawlist.cpp)
Such invisible decals were not taken into account when calculating their total number
Changing cl_maxdecals CVAR may not work as expected because of limit counter's inconsistent value
https://forum.zdoom.org/viewtopic.php?t=70457
src/common/console/c_console.cpp:1032:4: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/common/console/c_console.cpp:958:4: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/common/scripting/backend/codegen.cpp:2969:4: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/common/scripting/frontend/zcc_compile.cpp:2656:4: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/common/scripting/frontend/zcc_compile.cpp:521:5: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/common/scripting/jit/jit_math.cpp:966:19: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/d_main.cpp:1538:3: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/d_main.cpp:1549:3: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/d_main.cpp:3820:3: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/gameconfigfile.cpp:544:25: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/gamedata/p_xlat.cpp:193:11: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/gamedata/p_xlat.cpp:200:11: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/gamedata/p_xlat.cpp:210:11: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/g_game.cpp:1145:20: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/maploader/maploader.cpp:2142:3: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/menu/doommenu.cpp:159:28: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/m_cheat.cpp:109:3: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/playsim/mapthinkers/a_ceiling.cpp:254:24: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/playsim/mapthinkers/a_floor.cpp:322:18: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/playsim/mapthinkers/a_floor.cpp:330:18: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/playsim/mapthinkers/a_floor.cpp:344:10: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/playsim/mapthinkers/a_plats.cpp:129:47: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/playsim/mapthinkers/a_plats.cpp:284:8: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/playsim/p_acs.cpp:6302:18: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/playsim/p_acs.cpp:8111:16: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/playsim/p_spec.cpp:344:6: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/scripting/decorate/thingdef_parse.cpp:1343:9: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/scripting/zscript/zcc_compile_doom.cpp:298:12: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/wi_stuff.cpp:463:14: warning: this statement may fall through [-Wimplicit-fallthrough=]
src/common/menu/menudef.cpp:785:19: warning: comparison of integers of different signs: 'int' and 'unsigned int' [-Wsign-compare]
src/playsim/p_pspr.cpp:178:3: warning: field 'processPending' will be initialized after field 'HAlign' [-Wreorder]
Syntax: AddListMenu "MenuName" [<before|after> "SubMenuName"] { ... }
By default, new items will be added to the end of the menu. Specifying before/after allows inserting items before or after a specific submenu item.
The real issue is that the number of unattenuated sounds was unchecked and the near limit never kicked in.
To do this properly it is necessary to adjust the limit distance by the attenuation - zero attenuation must mean infinite distance and for high attenuations the distance must be lowered for limiting to work as intended.
The limit for the Doom boss sounds was increased to 4 to compensate for this change.
# Conflicts:
# src/common/audio/sound/oalsound.cpp
- Setting WOF_INTERPOLATE on any of the functions with flags (except A_OverlayFlags) will enable one tic for interpolating.
- Setting PSPF_INTERPOLATE with A_OverlayFlags will permanently enable it for that layer, but requires manual calling.
A_OverlayPivotAlign(int layer, int halign, int valign)
- Aligns the pivot point origin to a corner of the PSprite before applying offsets.
- - halign: Horizontal. Applicable constants are PSPA_<LEFT/CENTER/RIGHT>.
- - valign: Vertical. Applicable constants are PSPA_<TOP/CENTER/BOTTOM>.
- - Default is top left.
A_OverlayVertexOffset(int layer, int index, double x, double y, int flags)
- Allows offsetting the corners of the psprite, granting modders the ability to skew weapon sprites as they see fit.
- - index: The index of the vertice. Valid ranges are between [0,3].
- - x/y: Offsets of vertices.
- - flags: Takes WOF_ flags.
Other changes:
- Removed pivot point interpolation since it's pointless.
- Removed PSPF_PIVOTSCREEN due to complications with it having relativity. This will be revisited either later or in another submission.
- Added ResetPSprite() to be called with BringUpWeapon(), A_Lower(), and morph weapon raising to reset all the new properties to 0.
Nearly ready now. Just some final testing needed.
- Fixed PSPF_PIVOTPERCENT not being applied at start by default.
- Pivots now start in the upper left corner. Alignment flags are planned next for pivot point starting.
Significant progress made, but more testing still required.
- Fixed A_OverlayPivot carrying over copying of X to Y if Y is zero. This is meant to be exclusive to A_OverlayScale only.
- Added WOF_ZeroY to prevent copying X to Y for A_OverlayScale when desired, mirroring A_SetScale's behavior.
The idea behind this is to outright remove the relative position adding from the engine side and let it happen with A_OverlayOffset instead. Still more work to do.
* it is now possible to switch back to the main intro loop from the fullscreen console.
* do not distort the background of the fullscreen console. Always draw the console on top of what the main loop currently displays with a translucent black overlay.
* do not use the CONPIC for the fullscreen console as it tends to be more of a distraction than an asset when filling the entire screen.
* cleaned up c_console.cpp and moved several pieces of code unrelated to actual console management out of it.
I think it is inevitable that such status bars will appear in mods sooner or later and with the old code it is virtually impossible to create a wide status bar that displays properly but gets subjected to this code.
This only concerns the actual horizontal scroller. The vertical one still needs work and the "The End" screen only works if the second picture of the scroller is the full widescreen image because this page is done as a regular single image page which does not know anything about widescreen asset replacements.
The option still remains for cases where it does not work but the cases where mode 3 causes real problems are rare.
All I could find were a handful of poor intermission pics in some old mods which aren't worth keeping this artificial limitation.
This should only flag the tree as new if it is different from the last one.
This not only caused issues with dynamic updates but also made the renderer recreate the tree's data structures repeatedly.
Mainly to allow using 'if constexpr'.
Older CI compiler setups like GCC 5 for Linux were removed as a consequence.
Windows 32 bit was also removed because there are no plans to do any more 32 bit releases of GZDoom.
Rationale:
1. Now to reset a CVAR to default, the user doesn't need to remember the default value.
2. If a modder wants to reset a CVAR via menu command, they don't need to keep menudef and cvarinfo in sync.
src/common/filesystem/file_ssi.cpp:78:20: warning: comparison of integers of different signs: 'int' and 'uint32_t' (aka 'unsigned int') [-Wsign-compare]
src/common/filesystem/file_ssi.cpp:130:38: warning: comparison of integers of different signs: 'size_t' (aka 'unsigned long') and 'int' [-Wsign-compare]
This allows proper dimensions in the stock menus and should make menu layout for mods a lot easier because coordinates are more predictable than with the clean factors.
The feature is opt-in for custom content. As long as only stock elements are used, the menu will be able to switch between both on its own, but as soon as custom classes or custom elements are used, the virtual size must be declared explicitly, defaulting to clean scaling.
In this mode the clean scaling factors are ignored and content is always scaled to fit the screen as efficently as possible.
For the default summary screens an option was added to use this mode, which is a lot closer to the original look of this screen.
It is not 100% identical because it still factors in the author, if given and long level names that may be broken into multiple lines of text.
The background size wasn't taken into account for the splats and the widescreen check never verified the height of the background image.
It will now unconditionally use the new widescreen expansion mode for everything to reduce problems, use the background's display size for positioning overlays and only in the case of 200 pixel tall images perform widescreen positioning adjustment.
* a few shorter fields were moved into alignment gaps
* the visible angles are now stored as floats. Since these are only used for rendering the loss of precision is negligible, but this shortens AActor by 16 bytes alone.
Sensitivity scaling of both axes is now exposed as a raw factor to the user instead of obscuring it behind an unclear 'prescale' boolean.
This also consolidates the coordinate processing code to prevent such discrepancies as were present here from happening again.
Migration code for old config settings is present so that this change does not affect existing configurations.
Sound stopping was done in the wrong place - this killed the sound before they could be serialized for hub travel.
The sound may only be stopped *after* calling G_DoCompleted.
This reverts commit 8fc9f1e5ef.
As I already feared, this does not work right. The entire linking/unlinking code is simply too fragile. So no protection here against inept tinkering.
(for some reason it was unable to figure out the overrides for 'lumpname' in the function g_level.cpp/CalcMapName, even though this is a very old function)
- DirectInput axes are now X first, Y second.
- Menu axes controls swapped/fixed.
- Added SDL DualShock3 buttons in menu actions.
- Fixed SDL mouse system cursor; it could appear in-game.
This is the last parameter for A_RadiusThrust() DECORATE function.
If it is omitted or none, then A_RadiusThrust will behave as usual.
If it is set for some species name, it witt thrust only that species.
Of course, these species should be +VULNERABLE to be thrustable.
* support for SSI container format - this is an obscure format used by a few Duke Nukem mods - added to have this subsystem identical between both engines.
* removed some dead wrapper code
* made the Open methods local to their implementations, this was never called virtually.
* new texture format: ANM - this reads the first frame of a Build-ANM movie as a texture.
* some preparations for indexed (paletted) rendering.
* optimization of the patch texture checker - do not read in the entire file if checking the initial header is sufficient for rejecting it.
* removed refreshFreq variable and related code. This only got into the backend because of stupid interpolation code in some of the Build games which has long been removed.
* save FixedBitArrays as base64 in savegames.
* allow indirections in the string table - by prefixing the language string with '$$' the remaining text is interpreted as another string label to resolve.
* constexpr in palette code, also replacing #defines with enums.
Backported from Raze - without this there is no way to implement sounds that can unload themselves, because the existing ChannelEnded function is called when the sound is still active.
If the loader is allowed to go on, these objects will be linked into game structures and may do unpredictable things before the loader has a chance to abort at the end.
Some people apparently have to toy around with the engine's innards without fully understanding them. :(
When properly used the saveguards should never be triggered.
* Add FALLDAMAGE flag and add property to properly apply falling damage to the monsters
* Change name of propermonsterdamage property to propermonsterfallingdamage
It's possible for an actor to call Destroy() in a ZScript method (such
as Tick()) and then subsequently call A_StartSound() to play a sound.
Generally speaking this doesn't happen within a given class, but with a
class hierarchy, Destroy() may be called unbeknownst to a mod developer.
Even though checking bDestroyed is likely good practice, this ensures
that sounds won't be started on actors flagged for cleanup.
- If the hint text portion of a STRINGS lump entry begins with a '$', the value is treated as a string table lookup
Allows SECRETS strings to be translated via LANGUAGE lump. Reference: https://forum.zdoom.org/viewtopic.php?f=15&t=69827
Replaced explicit save of configuration file, which doesn't work after recent backend refactoring, with passing of CVARs we need to store to a new GZDoom process
https://forum.zdoom.org/viewtopic.php?t=69790
It's not obvious where we can reset blocking line in a more generic way, and whether it's possible at all, because this member is exposed to scripting
https://forum.zdoom.org/viewtopic.php?t=69747
src\common\audio\sound\s_sound.cpp(556,27): warning C4244: '=': conversion from 'double' to 'float', possible loss of data
src\scripting\vmthunks.cpp(2678,13): warning C4244: 'return': conversion from 'time_t' to 'int', possible loss of data
src/console/c_cmds.cpp:690:11: warning: format string is not a string literal (potentially insecure) [-Wformat-security]
src/scripting/vmthunks.cpp:2698:21: warning: comparison of integers of different signs: 'int' and 'unsigned long' [-Wsign-compare]
* Added ViewAngle/Pitch/Roll properties to actors.
- These are offsets for camera angles that allow turning the camera without affecting aim or movement direction.
- Added A_SetView<Angle/Pitch/Roll>, which will set the view direction.
- Added ABSVIEWANGLES flag, used to make the view absolute instead of an offset.
* Converted functions to be direct-native.
* Simplified some stuff, made up-to-date
* Dealt with DST problems
* Made SystemTime.Format clearscope, as there is no reason for this function to be limited to the ui
- If the actor passed into the function has THRUACTORS, PIT_CheckThing won't be called anymore.
- Moved THRUACTORS check to just under the self clipping check in PIT_CheckThing.
Bogus pointers and offsets were serialized silently
This will to find other cases of dangling pointers, and protect from loading of broken saved games
https://forum.zdoom.org/viewtopic.php?t=67494
Inventory items had dangling floor and ceiling sector pointers after transition to a new level
This could cause random issues with serialization
https://forum.zdoom.org/viewtopic.php?t=67494
This is a quick fix for an error in which Animated Doors crash the game by trying to check "actor->player" when "actor" itself is NULL. Deleting the check entirely also worked, but I worried it might be there for some higher-level scripting reason. This just puts in a check to make sure actor isn't NULL before checking actor->player, and keeps the behavior in that case the same.
I think this was happening because I had doors being opened by projectiles (like in Metroid) which were being despawned into NULL pointers when they hit the doors, as this was an issue when initially programming said doors..
- 1: The missile will continue moving through the actor, and it's down to the modder to handle bouncing.
- 0: The missile will explode.
- Any other value will process a bouncing actor as normally done in the engine.
src/common/engine/stats.h:139:31: error: invalid output constraint '=A' in asm
src/common/platform/posix/osx/iwadpicker_cocoa.mm:379:1: warning: non-void function does not return a value [-Wreturn-type]
- Sets the direct pitch of the sound to the specific float. Default is 0.0, meaning do not set a specific pitch. Regular pitch is 1.0.
- Only works for direct sound definitions.
- Overrides $PitchShift unless value is <= 0.0
- Overridden by A_StartSound's pitch parameter if the value > 0.0.
* 0: previous behavior (completely MAPINFO controlled)
* 1: always show level tally at the end of an episode
* 2: always show level tally even when changing levels within a hub
With introduction of localization support, several strings are no longer identical to the original messages
Double spaces were removed, and letter case was changed
The latter is already handled, extra whitespaces are now deleted to deal with the former
https://forum.zdoom.org/viewtopic.php?t=69236
Negative coordinates are treated as "disable scissor", so if some transformation results in something off-screen it needs to be handled before passing it to the backend.
The option effectively disables per frame mouse checks.
The motivation to add it was that many wireless mice 10-15 years ago had update rates of less than 35Hz, and on these it was necessary to sync mouse input with the playsim to properly interpolate between updates.
Today this is totally useless and even counterproductive because modern mice have significantly higher update rates, so this option no longer smoothes things but instead makes the mouse feel more choppy.
