- ACTION_MENU* not sent
- allow using the game UI with keyboard only
- allow mapping ACTION_MENU to controller inputs
- fix joystick axis (-) not working due to typo
- Support joystick hotplugging
- Support axes as buttons (means xbox360 shoulder triggers can be used as buttons)
- Show pretty joystick axis & button names if possible
- Tabify input actions UI
- Add 'mouse' column to input actions UI
- Allow to configure form hotkeys
- Allow ALL keys, get rid of internal key remapping
- SDL2: Use scancodes instead of keycodes since they are layout independent
- Allow extra mouse buttons (if present)
- Remove "lmbd" & "lmbu" actions in favor of "PrimaryAction" & "SecondaryAction"
Makes the configuration less redundant and doesn't send each action twice,
which happened if both were set to the same key.
- Fix Regressions from prev commits (menu not opening on Esc)
Still has a few minor bugs/issues that need to be fixed,
but pushing this now before the commit gets too large again.
This untangles some of the gigantic kitchen sink headers
in an attempt to split things into smaller files.
Also don't include gl.h, glext.h, windows.h,
and other such nonsense *everywhere*.
Lots of cleanups on the way too. More dead/unused code removal.
Remove incrFlag(), decrFlag() Lua functions.
Started working on a tiny input/ActionMapper refactor,
then everything fell apart and i ended up doing this.
I'm sorry.
Pretty much untested because input mapping is broken right now,
will fix that next.
Another vector that was always empty, optionally filled when specified
by a skeletal file, but none actually did this. Should be safe to remove.
Also small related optimization in Game::collideSkeletalVsCircle().
This was always (0, 0) and optionally used for bones if defined in skeletal files.
None of the official skeletal files used this, so this should be safe to remove.
The game cannot work with those anyhow...
```c
#undef BBGE_BUILD_IRRKLANG
#undef BBGE_BUILD_OPENALOGG
#undef BBGE_BUILD_XINPUT
#undef BBGE_BUILD_DELAYXINPUT
```
Should cause less GL pipeline stalling / driver spinlocking if enabled.
Disable AQUARIA_USE_GLM (default: true) in CMake to use the old GL pipeline version.
Based on the implementation by Matt Bierner:
https://bitbucket.org/mattbierner/ios-aquaria
- add obj_getUpdateCull()
- add obs parameter to getWallNormal(x, y, dist, obs) to control which obs to consider
- add getPerformanceCounter(), getPerformanceFreq()
- Now, the camera will now no longer be pulled towards the upper left
or lower rigtht corner.
- cameraPos was never used as an interpolating vector, cleaned up related code.
- Core::invGlobalScale is now updated whenever Core::globalScale is changed
(Not a nice solution but it does work)
- Increase cull radius by 10%, should prevent tiles from disappearing when
zoomed in a lot.
- entity_hasSkeletal()
- entity_getNumAnimLayers()
- filterNearestEntities() to fill an internal filter buffer
- getNextFilteredEntity() to iterate over the filter buffer
Minor changes:
- clean out some Lua functions registered twice
- Correct bool cast in quad_setSegs()
- Show nested mains in debug console
- One less //HACK comment
This avoid the problems with AL_INVERSE_DISTANCE_CLAMPED (doesn't reach 0),
and sounds way better than AL_LINEAR_DISTANCE_CLAMPED, which has an
all-too-sudden decay when a sound goes out of audible range.
With this patch, entities can serve as a sound source whose position
updates relative to the listener. This implements the last missing
feature of my previous positional audio patch, which is moving sound sources.
Previously, all sounds were relative to the listener with the listener
centered at (0, 0, 0), and once a sound started playing, the position could
not be changed. Volume was set as an estimation of the distance to the listener.
These restrictions are now gone; OpenAL will handle the volume & panning
based on the distance.
The remaining problem is that stereo sounds are not attenuated, at all.
Lua additions:
- playSfx() takes an additional parameter now.
- entity_setStopSoundsOnDeath()
This changes all typical vector interpolation functions that get
value, time, loopType, pingPong, ease as args from the Lua state;
use unified functions for all value type occurances.
This fixes the slightly annoying problem that it was not possible to
pass a boolean true to pingPong or ease, because lua_tonumber() would
always return 0 in that case.
Also changed:
- obj_getWorldPosition() now takes optional x,y-vector
New:
- obj_getWorldPositionAndRotation() is a more efficient shortcut,
as often both position and rotation are retrieved together,
and the underlying matrix chain operation is rather expensive.
- entity_getEmitter()
- entity_getNumEmitters()
- setSceneColor2()
- getSceneColor2()
The fix as suggested by Jonas Kulla on the mailing list is also in.
There hasn't been any extensive testing, and it may not yet compile
on platforms other than win32.
- quad_setRepeatTexture()
- quad_setRepeatScale()
- quad_isRepeatTexture()
(bone, entity, ... as well)
- getOldDT() - return unmodified dt
- getDT() - as passed to update(), for complicated edge cases
where dt is unavailable but required.
This commit is mainly intended to ease packaging for linux.
Unless environment variable AQUARIA_DATA_PATH is set, there are two
directories which are checked by the game:
If AQUARIA_DEFAULT_DATA_DIR is defined, it will chdir there for main
operation. If it's not defined, it chdirs into the directory where
the executable is located.
Then, if AQUARIA_EXTRA_DATA_DIR is defined, it will mount this directory
and all contents into the working path, so that the files present there
will override those from the working directory when accessed by the game.
Setting the environment variable AQUARIA_DATA_PATH will disable this
behavior altogether and use *only* AQUARIA_DATA_PATH as working dir.
This has the side effect that elements on the dark layer will now properly
wiggle when affected by water ripples.
Before this patch, the dark layer was always drawn above the postprocessed
screen content, unaffected.
This allows attaching multiple shader programs, processing the entire
screen as texture. Chaining shaders works, each shader will use
the output of the previous shader as input.
This repairs pixel/vertex shader code that was seemingly
experimented with during testing, but it was never used and no
shaders ever made it to the data files.
GL_CULL_FACE is effectively always disabled, except for quad strip
rendering, and keeping GL_CULL_FACE disabled there does not seem
to cause any visual artifacts.
Instead, it enhances skeletal animation possibilities
(e.g. flipping bone strips over without it disappearing)
This commit introduces a second pause mode:
Full game pause (as in menu), and world pause (as in spirit form).
All related checks are no longer done against WT_* constants,
but against the new world pause which is functionally equivalent,
but more flexible. Continuity::worldType is now only used to toggle
world pause correctly, and to apply some related graphical effects.
The world pause can also be controlled via script without actually
switching forms.
Added Lua functions:
+ node_setSpiritFreeze()
+ quad_setPauseLevel()
+ isWorldPaused()
+ setWorldPaused()
The first one caused the music to stop randomly on scene changes,
reproducible by changing maps with different music while holding 'G'
button to speed up time.
For some reason this caused the music to stop playing during the crossfade.
Wasn't able to find the root cause, so I added a little hack to kick the
music back on track if it stops unexpectedly.
The second one was a simple string case issue; mostly noticable because of
map background sounds not playing in loop mode.
(this is possibly a regression from an earlier patch, not sure)
Also small config addition to optionally increase memory allocation size
for temporary texture pixel data; the nouveau drivers seem notoriously
awkward and keep writing over the buffer.
This was caused by the VFS caching file sizes,
and not noticing that the file size had changed,
which confused zlib because the data stream ended too early.
- HTTP networking support, mods can be downloaded via the builtin downloader.
All network activity runs in a seperate thread, which is started
as soon as any network activity is requested.
- The master server is hard-coded to fg.wzff.de/aqmods/ if not specified otherwise;
this setting can be overridden in the config file.
- The mod selector screen is now a grid-view for much better navigation;
also works with joystick.
- VFS code is functionally similar to the old molebox-packed release
for win32. The game could also have its data shipped in a Zip file
or any other kind of archive.
- It is still possible to build without VFS support, but then the mod
downloader and soft-patching will not be available.
The full commit history can be found here:
https://github.com/fgenesis/Aquaria_clean/compare/master...vfs
The most important commit messages follow:
[...]
This replaces all std::ifstream with InStream, and fopen(), ... with vfopen(), ...
Some code is #ifdef'd for better performance and less memory-copying.
VFILE is defined to whatever type of file is in use:
- FILE if BBGE_BUILD_VFS is not defined
- tttvfs::VFSFile if it is.
Other changes:
- [un]packFile() is now unused and obsolete. That code has not been adjusted to use VFILE.
- glpng can now load from a memory buffer.
- TinyXML uses the VFS for reading operations now.
- The rather clunky binary stream loading of glfont2 got replaced with ByteBuffer,
which gets its data in one block (necessary to use the VFS without implementing
a somewhat STL-compliant std::ifstream replacement.)
-------------
Implement loading mods from zip files.
-------------
Implement soft-patching game data files. (Replacing textures/audio/... on the fly)
-------------
Misc bits:
- Extended GUI focus handling a bit
- Fixed weirdness in texture loading... not sure but this seems more correct to me.
Actually, considering that the texture will have its native size after restarting the game,
the lines removed with this commit seem pretty useless.
This was removed in HG changeset 7ec478d993b7, and is now implemented
in a way that is better than before: voice overs and music
no longer cause decoding lag, as they are always decoded on-the-fly.
The additional memory use (~40 MB) should be no problem for anyone.
The default is still to decode everything on the fly.
