- CSceneNodeAnimatorCameraFPS uses now SKeyMap instead of SCamKeyMap (structs were identical which was confusing and there wasn't any explanation in comments, so I decided to simplify it).
- Add some workaround to MeshViewer to show how we can currently fix the FPS-cam when users to alt-tab while moving. We can improve that some day when we have focus-events, but this works for now.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@3920 dfc29bdd-3216-0410-991c-e03cc46cb475
Added an example showing a very simple occlusion query setup.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@3628 dfc29bdd-3216-0410-991c-e03cc46cb475
ITriangleSelector::getSceneNodeForTriangle, ISceneNodeAnimatorCollisionResponse::getCollisionNode, ISceneCollisionManager::getCollisionPoint and ISceneCollisionManager::getCollisionResultPosition.
As collision functions often are followed by changing node positions users where so far forced to using const_casts (found by Greenya).
(This will lead once more to many post about correct getCollisionResultPosition usage, but can't be helped - those consts were too restricitive, sorry)
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@3442 dfc29bdd-3216-0410-991c-e03cc46cb475
- CFileList
added Offset Parameter to SFileListEntry and removed the private array from the archive loaders.
CFileList::addItem now uses automatic incremental id if id = 0
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@3336 dfc29bdd-3216-0410-991c-e03cc46cb475
- BurningVideo
- add Normalmap Rendering ( one light only), pushed Burningvideo to 0.46
- internal vertexformat changed
- changed fixpoint from 9 to 10 bit fract resolution
- renamed createBurningVideoDriver to createBurningVideoDriver and uses SIrrlichtCreationParameters like opengl
- internal interfaces for the trianglerenders unified.
- Example 11.
changed the light billboards to use the light color. ( green light, green particle, red light red particle )
allow to disable the bump/parallax on the earth like in the room ( with transparency )
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@3333 dfc29bdd-3216-0410-991c-e03cc46cb475
Same childs => children in a few places in documenatation and variables.
Removed a compile warning in CCubeSceneNode.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@3330 dfc29bdd-3216-0410-991c-e03cc46cb475
We use multicolor hardware cursors, but so far only Linux and Windows 32 are supported.
For Linux application must now link additionally with Xcursor unless they disable _IRR_LINUX_XCURSOR_.
Example 24 is added to show how to use those cursors and also as a general test for mouse-behavior.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@3302 dfc29bdd-3216-0410-991c-e03cc46cb475
- Further cleanup/improvements to example 22 MaterialViewer
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2974 dfc29bdd-3216-0410-991c-e03cc46cb475
- Break up EventHandler into more functions
- Add enums for all ID's
- Fix scaling display
- Add quick-scale buttons for *10 and * 0.01
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2953 dfc29bdd-3216-0410-991c-e03cc46cb475
- Added TAR archive loader.
- Renamed the following functions-
IFileArchive::getArchiveType to getType
IFileSystem::registerFileArchive to addFileArchive
IFileSystem::unregisterFileArchive to removeFileArchive
IFileArchive::openFile to createAndOpenFile
- New enum, E_FILE_ARCHIVE_TYPE. getType on IArchiveLoader and IFileArchive now both return this.
- IFileSystem::addFileArchive takes a parameter to specify the archive type rather always using the file extension. IFileSystem::addZipFileArchive, addFolderFileArchive and addPakFileArchive now use this but these functions are now marked as deprecated. Users should now use addFileArchive instead.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2447 dfc29bdd-3216-0410-991c-e03cc46cb475
- PixelBlend16 and PixelBlend16_simd are working for the new rules.
- bugfix. CLightSceneNode didn't correctly update it's attributes
Lighting Linear Attenuation. = 1.f / radius
The Example loadirr files set the lightscene radius to 1000.f but
stays on the previous default attentuation with the older radius 100 -> 1.f / 100
so the examples looks golden-brown.
Now the radius is correctly!! set to the attenuation of 1.f/1000.f because the
file doesn't have special attenuation. and now it looks more yellow.
can anybody show me a correct screenshot for this file;-)? Niko?
Or is this behavior the default lighting?. then it would be
a fixed constant linear attenuation of 0.01f;-). Please clearify
For now i didn't fixed it
I encountered this behavior because i ( burning video ) used the original radius
for calculations and so i've found that radius != 1.f / linearAttenuation but
in the LightSceneNode this formula was used.. confused;-)
- vector template and equals tests
as working with the test suits i cleaned the template behavior (mixed types are
used in the templates) and added all missing special math function with their coressponding type
I also set the equal test for s32 to behave like the f32 routine.
The function equals always implements a weak test.
that means a tolerance MUST always be used if you use the equal function. default is 1.
you can set it to zero a==b-> equals ( a, b, 0 ) but do it explicit like you have to
for floating compare. This is important when irrlicht is going to use special hardware
math acceleration on a per function base, like sse2, or the other way round fixpoint.
- VideoDriver drawPixel
The HW renderes are using the alpha components for blending.
The Software Renderes and image loaders are using CImage::setPixel copy.
so setPixel is engaged to either blends or copy the pixel
default: false
- Burningvideo
added RenderMaterial EMT_SPHERE_MAP
pushed burningsvideo to 0.43
added RenderMaterial EMT_REFLECTION_2_LAYER
pushed burningsvideo to 0.44
set EMT_TRANSPARENT_ALPHA_CHANNEL_REF
to use AlphaRef 0.5 like Direct3D
One Note: in OpenGL there is know difference between sphere_map and reflection layer
both using GL_TEXTURE_GEN_MODE GL_SPHERE_MAP, whereas in d3d one time using camera_normal
on sphere and reflection on refletcion_layer.
The visual difference is that on sphere map the "image is not moving" when you rotate the
viewer. For Buring i took the opengl visual. always moving
- rename quake3 SEntity to IEntity to be confom with IShader
even IShader and IEntity are none pure virtual interfaces
like most irrlicht objects
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2207 dfc29bdd-3216-0410-991c-e03cc46cb475
- fixed createMeshWith2TCoords
normals were missing during copy.
- addded
//! Creates a copy of the mesh, which will only consist of S3DVertex vertices.
IMesh* CMeshManipulator::createMeshWith1TCoords(IMesh* mesh) const
- added io::IFileSystem* CSceneManager::getFileSystem()
for preparing to remove the (mostly) unnecessary double member variables
in many loaders
- added virtual const c8* ISceneManager::getAnimatorTypeName(ESCENE_NODE_ANIMATOR_TYPE type);
to the SceneManger Interface. just like getTypeName is public
- added CSceneNodeAnimatorFlyCircle::radiusEllipsoid.
if radiusEllipsoid == 0 the default circle animation is done
else radiusEllipsoid forms the b-axe of the ellipsoid.
-> gummiball bouncing
- added ISceneManager::createFlyStraightAnimator variable bool ping-pong
used in loop mode to device if start from beginning ( default ) or make ping-pong
-> straight bouncing
- changed IFileSystem::registerFileArchive
remove the index of the hiarchy and added a new interface method
//! move the hirarchy of the filesystem. moves sourceIndex relative up or down
virtual bool moveFileArchive( u32 sourceIndex, s32 relative ) = 0;
- bugfix and changes in
SViewFrustum::SViewFrustum
wrong size of Matrices copy. This bug must be ages old... (typo)
detected during resizing the Matrices. removed obsolute Matrices
renamed E_TRANSFORMATION_STATE_2 to E_TRANSFORMATION_STATE_FRUSTUM
therefore also changed SViewFrustum::setTransformState to not tap
in the pitfall again of wrong memory...
and renamed it to getTransform, like in the driver
and Matrices private
- OpenGL:
Specular
- moved
//! EMT_ONETEXTURE_BLEND: has BlendFactor Alphablending
inline bool textureBlendFunc_hasAlpha ( E_BLEND_FACTOR factor ) const
from the material renderes ( 3x declared ) to SMaterial.h
- updated managed light example to use standard driver selection
- BurningsVideo
- LightModel reworked.
Point Light & Direction Light works for Diffuse Color as aspected
Specular and Fog still have problems ( needs new pixel shader )
pushed burningsvideo to 0.42 for this major step
- removed obsolete matrix transformations
renamed E_TRANSFORMATION_STATE_2 to E_TRANSFORMATION_STATE_BURNING
- cleaned line3d.h vector3d.h template behavior.
many mixed f32/f64 implementations are here. i'm not sure if this should be
the default behavior to use f64 for example for 1.0/x value, because they
benefit from more precisions, but in my point of view the user is responsible
of choosing a vector3d<f32> or vector3d<f64>.
- added core::squareroot to irrmath.h
-> for having candidates for faster math in the same file
- added AllowZWriteOnTransparent from SceneManager to burningsvideo
Following SceneManger guideline
-added hasAlpha() to ITexture
This info can be used for e.q to downgrade a transparent alpha channel blit
to add if the texture has no alpha channel.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2199 dfc29bdd-3216-0410-991c-e03cc46cb475
- Triangle selectors created from animated mesh scene nodes will update themselves as required to stay in sync with the node.
- ISceneCollisionManager::getSceneNodeAndCollisionPointFromRay() allows selection by BB and triangle on a heirarchy of scene nodes.
Example 07 updated to show the usage of ISceneCollisionManager::getSceneNodeAndCollisionPointFromRay(), used on animated meshes.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2177 dfc29bdd-3216-0410-991c-e03cc46cb475
Derive COctTreeSceneNode from IMeshSceneNode instead ISceneNode, and have it return the IMesh that was used to create its octtree. Tested with a slightly modified example 07.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2173 dfc29bdd-3216-0410-991c-e03cc46cb475
example 21.update. save/load game data on driver change
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2157 dfc29bdd-3216-0410-991c-e03cc46cb475
Adapted some more methods to the new string API.
renamed FileSystem method.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2150 dfc29bdd-3216-0410-991c-e03cc46cb475
- FileSystem 2.0 SUPER MASTER MAJOR API CHANGE !!!
The FileSystem is know build internally like for e.q the texture-, and the meshloaders.
There exists a known list of ArchiveLoader, which know how to produce a Archive.
The Loaders and the Archive can be attached/detached on runtime.
The FileNames are now stored as core::string<c16>. where c16 is toggled between char/wchar
with the #define flag _IRR_WCHAR_FILESYSTEM, to supported unicode backends (default:off)
I replaced all (const c8* filename) to string references.
Basically the FileSystem is divided into two regions. Native and Virtual.
Native means using the backend OS.
Virtual means only use currently attach IArchives.
Browsing
each FileSystem has it's own workdirectory and it's own methods to
- create a FileTree
- add/remove files & directory ( to be done )
Hint: store a savegame in a zip archive...
basic browsing for all archives is implemented.
Example 21. Quake3Explorer shows this
TODO:
- a file filter should be implemented.
- The IArchive should have a function to create a filetree
for now CFileList is used.
Class Hiarchy:
IArchiveLoader: is able to produce a IFileArchive
- ZipLoader
- PakLoader
- MountPointReader ( formaly known as CUnzipReader )
IFileArchive:
-ZipArchive
-PakArchive
-MountPoint (known as FolderFile)
IFileSystem
- addArchiveLoader
- changed implementation of isALoadableFileExtension in all loaders
to have consistent behavior
- added a parameter to IFileList * createFileList
setFileListSystem
allows to query files in any of the game archives
standard behavior listtype = SYSTEM ( default)
- CLimitReadFile
added multiple file random-access support.
solved problems with mixed compressed & uncompressed files in a zip
TODO:
- Big Big Testing!!
- Linux Version ( minor )
- remove all double loader interfaces where only the filename differs
(IReadFile/const char *filename). This blows up the the interface
- many loaders use their own private filesearching
we should rework this
- there are a lot of helper function ( getAbsolutePath, getFileDir )
which should be adapted to the virtual filesystem
- IrrlichtDevice
added:
virtual bool setGammaRamp( f32 red, f32 green, f32 blue, f32 brightness, f32 contrast ) = 0;
virtual bool getGammaRamp( f32 &red, f32 &green, f32 &blue ) = 0;
and calculating methods to DeviceStub.
implemented in Win32, TODO: other Devices
- irrlicht.h
changed exported irrlicht.dll routines createDevice, createDeviceEx, IdentityMatrix
to extern "C" name mangling.
for easier dynamically loading the irrlicht library and different versions
- ParticleSystem
removed the private (old?,wrong?) interface from the ParticleEffectors
to match the parent class irr::io::IAttributeExchangingObject::deserializeAttributes
TODO:
please test if the serialization works!
- Generic
- vector3d<T>& normalize()
#if 0
f32 length = (f32)(X*X + Y*Y + Z*Z);
if (core::equals(length, 0.f))
return *this;
length = core::reciprocal_squareroot ( (f32)length );
#else
const T length = core::reciprocal_squareroot ( (X*X + Y*Y + Z*Z) );
#endif
Weak checking on zero?!?! just to avoid a sqrt?. mhm, maybe not;-)
added reciprocal_squareroot for f64
- dimension2d
added operator dimension2d<T>& operator=(const dimension2d<U>& other)
to cast between different types
- vector2d
bugfix:
vector2d<T>& operator+=(const dimension2d<T>& other) { X += other.Width; Y += other.Width; return *this; }
to
vector2d<T>& operator+=(const dimension2d<T>& other) { X += other.Width; Y += other.Height; return *this; }
- C3DMeshLoader renamed chunks const u16 to a enum
removing "variable declared but never used warning"
- added a global const identity Material
changed all references *((video::SMaterial*)0) to point to IdentityMaterial
removed warning: "a NULL reference is not allowed"
- modified IRRLICHT_MATH to not support reciprocal stuff
but to use faster float-to-int conversion.
gcc troubles may they are. i'm using intel-compiler..;-)
- core::matrix4
USE_MATRIX_TEST
i tried to optimize the identity-check ( in means of performance)
i didn't succeed so well, so i made a define for the matrix isIdentity -check
for now it's sometimes faster to always calculate versus identity-check
but if there are a lot of scenenodes/ particles one can profit from the
fast_inverse matrix, when no scaling is used. further approvement could
be done on inverse for just tranlastion! ( many static scenenodes are not rotated,
they are just placed somewhere in the world)
one thing to take in account is that sizeof(matrix) is 64 byte and
with the additional bool/u32 makes it 66 byte which is not really cache-friendly..
- added buildRotateFromTo
Builds a matrix that rotates from one vector to another
- irr::array. changed allocating routine in push_back
okt, 2008. it's only allowed to alloc one element, if
default constructor has to be called.
removes existing crashes. ( MD3 Mesh ) and possible others ones.
A new list template should be made.
one with constructor/destructor calls ( safe_array ) and
one without. like the array since the beginning of irrlicht.
currently the array/string is extremly slow..
also a hint for the user has to be done, so that a struct T of
array<T> must have a copy constructor of type T ( const T&other ).
i needed hours to track that down...
added a new method setAllocStrategy,
safe ( used + 1 ), double ( used * 2 + 1)
better default strategies will be implemented
- removed binary_search_const
i added it quite a long time ago, but it doesnt make real sense
a call to a sort method should happen always. i just wanted to safe
a few cycles..
- added binary_search_multi
searches for a multi-set ( more than 1 entry in the sorted array)
returns start and end-index
- changed some identity matrix settings to use core::IdentityMatrix
- added deletePathFromFilename to generic string functions in coreutil.h and
removed from CZipReader and CPakReader
- s32 deserializeAttributes used instead of virtual void deserializeAttributes in
ParticleSystem ( wrong virtual was used)
- strings & Locale
- started to add locale support
- added verify to string
- added some helper functions
- XBOX
i have access to a XBOX development machine now. I started to compile
for the XBOX. Question: Who did the previous implementation?. There
is no XBOX-Device inhere. maybe it's forbidden because of using the offical
Microsoft XDK. I will implement a native or sdl device based on opendk.
irrlicht compiles without errors on the xbox but can't be used.
TODO:
- native XBOX Device
- Windows Mobile
reworked a little. added the mobile example to the windows solution for
cross development.
added maximal 128x128 texture size for windows mobile ( memory issues )
- Collision Speed Up
The Collision Speed Up greatly improves with many small static child-nodes
- added COctTreeTriangleSelector::getTriangles for 3dline from user Piraaate
- modified createOctTreeTriangleSelector and createTriangleSelector
to allow node == 0, to be added to a meta selector
- CSceneNodeAnimatorCollisionResponse has the same problem as CSceneNodeAnimatorFPS
on first update:
Problem. you start setting the map. (setWorld). First update cames 4000 ms later.
The Animator applies the missing force... big problem...
changed to react on first update like camera.
- add Variable FirstUpdate. if set to true ( on all changes )
then position, lasttime, and falling are initialized
-added #define OCTTREE_USE_HARDWARE in Octree.h
if defined octtree uses internally a derived scene::MeshBuffer which has
the possibility to use the Hardware Vertex Buffer for static vertices and
dirty indices;-)
if defined OCTTREE_USE_HARDWARE octree uses internally a derived scene::CMeshBuffer
so it's not just a replacement inside the octree. It also in the OctTreeSceneNode.
#if defined (OCTTREE_USE_HARDWARE)
driver->drawMeshBuffer ( &LightMapMeshes[i] );
#else
driver->drawIndexedTriangleList( &LightMapMeshes[i].Vertices[0], LightMapMeshes[i].Vertices.size(),
d[i].Indices, d[i].CurrentSize / 3);
#endif
#define OCTTREE_PARENTTEST is also used. It's skip testing on fully outside and takes everything on fully inside
- virtual void ISceneNode::updateAbsolutePosition()
- changed
inline CMatrix4<T> CMatrix4<T>::operator*(const CMatrix4<T>& m2) const
all two matrices have to be checked by isIdentity()
to let the isIdentity work always
-changed inline bool CMatrix4<T>::isIdentity() const
on full identityCheck->
to look first on Translation, because this is the most challenging element
which will likely not to be identity..
- virtual core::matrix4 getRelativeTransformation() const
Hiarchy on Identity-Check
1) ->getRelativeTransform -> 9 floating point checks to be passed as Identity
2) ->isIdentity () -> 16 floating point checks to be passed as Identity
- inline void CMatrix4<T>::transformBoxEx(core::aabbox3d<f32>& box) const
added isIdentity() check
- changed CSceneNodeAnimatorCollisionResponse
- added CSceneNodeAnimatorCollisionResponse::setGravity
needed to set the differents Forces for the Animator. for eq. water..
- added CSceneNodeAnimatorCollisionResponse::setAnimateTarget
- added CSceneNodeAnimatorCollisionResponse::getAnimateTarget
- changed CSceneNodeAnimatorCollisionResponse::animateNode to react on FirstUpdate
- changad Gravity to
- TODO: set Gravity to Physically frame independent values..
current response uses an frame depdended acceleration vector.
~9.81 m/s^2 was achieved at around 50 fps with a setting of -0.03
may effect existing application..
- SceneNodes
- CSkyDomeSceneNode
moved radius ( default 1000 ) to constructor
added Normals
added DebugInfo
added Material.ZBuffer, added SceneMaanager
- CVolumeLightSceneNode:
changed default blending OneTextureBlendgl_src_color gl_src_alpha to
EMT_TRANSPARENT_ADD_COLOR ( gl_src_color gl_one )
which gives the same effect on non-transparent-materials.
Following the unspoken guide-line, lowest effect as default
- added LensFlareSceneNode (from forum user gammaray, modified to work )
showing in example special fx
- changed SceneNode Skydome f64 to f32,
- AnimatedMesh
-Debug Data:
mesh normals didn't rotate with the scenenode fixed ( matrix-multiplication order)
- Camera SceneNode setPosition
Camera now finally allow to change position and target and updates all
effected animators..
a call to OnAnimate ( ) lastime < time or OnAnimate ( 0 ) will reset the
camera and fr. the collision animator to a new position
- Device:
added the current mousebutton state to the Mouse Event
so i need to get the current mouse state from the OS
-a dded to CIrrDeviceWin32
TODO:
- Linux and SDL Device
- GUI
- CGUIFont:
- added virtual void setInvisibleCharacters( const wchar_t *s ) = 0;
define which characters should not be drawn ( send to driver) by the font.
for example " " would not draw any space which is usually blank in most fonts
and saves rendering of ususally full blank alpha-sprites.
This saves a lot of rendering...
default:
setInvisibleCharacters ( L" " );
- added MultiLine rendering
should avoid to us CStaticText breaking text in future
- CGUIListBox
- changed Scrollbar LargeStepSize to ItemHeight
which easy enables to scroll line by line
- CGUIScrollBar
bug:
Create a Window and inside a listbox with a scrollbar or
a windowed irrlicht application
Click & hold Scrollbar Slider. move outside it's region.
Release Mouse. Go Back to Scrollbar.. it's moving always...
it's generally missing the event PRESSED_MOVED, which
leads to problem when an element is dragging, has a focus, or position loose
and gets focus back again. ( think of a drunken mouse sliding left&right during tracking )
so added the mouse Input Buttonstates on every mouse event
IrrDeviceWin32:
added event.MouseInput.ButtonStates = wParam & ( MK_LBUTTON | MK_RBUTTON | MK_MBUTTON );
TODO:
Linux & SDL
so now i can do this
case irr::EMIE_MOUSE_MOVED:
if ( !event.MouseInput.isLeftPressed () )
{
Dragging = false;
}
- bug:
Scrollbar notifyListBox notify when the scrollbar is clicked.
- changed timed event in draw to OnPostRender
Why the hell is a gui element firing a timed event
in a draw routine!!!!!. This should be corrected for all gui-elements.
- added GUI Image List from Reinhard Ostermeier, modified to work
added GUI Tree View from Reinhard Ostermeier, modified to work
shown in the Quake3MapShader Example
TODO: Spritebanks
- FileOpenDialog
changed the static text for the filename to an edit box.
- changed the interface for addEditBox to match with addStaticText
- changed the interface for addSpinBox to match with addEditBox
- added MouseWheel to Spinbox
- changed CGUITable CLICK_AREA from 3 to 12 to enable clicking on the visible marker
- CGUISpritebank
removed some crashes with empty Sprite banks
- IGUIScrollBar
added SetMin before min was always 0
changed ScrollWheel Direction on horizontal to move right on wheel up, left on wheel down
- IComboBox
-added ItemData
- removed IsVisbile check in IGUIElement::draw
- Image Loaders
- added TGA file type 2 ( grayscale uncompressed )
- added TGA file type (1) 8 Bit indexed color uncompressed
ColorConverter:
- added convert_B8G8R8toA8R8G8B8
- added convert_B8G8R8A8toA8R8G8B8
- Media Files
- added missing shaders and textures to map-20kdm2.
Taken from free implementation
- ball.wav. adjusted DC-Offset, amplified to -4dB, trim cross-zero
- impact.wav clip-restoration, trim cross-zero
- added gun.md2, gun.pcx to media-files
copyright issues!. i don't know from where this file came from...
i hope this is not from original quake2..
- added new irrlicht logo irrlicht3.png
i've taken the new layout. i should ask niko to use it.
- added Skydome picture to media files (skydome2.jpg) half/sphere
- OctTree
-added
#define OCTTREE_PARENTTEST ( default: disabled )
used to leave-out children test if the parent passed a complete frustum.
plus: leaves out children test
minus: all edges have to be checked
- added MesBuffer Hardware Hint Vertex to octtree
- CQuake3ShaderSceneNode:
- removed function releaseMesh
Shader doesn't copy the original mesh anymore ( saving memory )
so therefore this (for others often misleading ) function was removed
- changed constructor to take a (shared) destination meshbuffer for rendering
reducing vertex-memory to a half
- don't copy the original vertices anymore
- added deformvertexes autosprite
- added deformvertexes move
- added support for RTCW and Raven BSPs ( qmap2 )
- added polygonoffset (TODO: not perfect)
- added added nomipmaps
- added rgbgen const
- added alphagen
- added MesBuffer Hardware Hint Vertex/Index to Quake3: static geometry, dynamic indices
- added Quake3Explorer examples
- added wave noise
- added tcmod transform
- added whiteimage
- added collision to Quake3Explorer
- renamed SMD3QuaterionTag* to SMD3QuaternionTag* ( typo )
- updated quake3:blendfunc
- added crouch to Quake3Explorer
(modifying the ellipsiodRadius of the camera animator )
added crouch to CSceneNodeAnimatorCameraFPS
still problems with stand up and collision
- Quake3MapLoader
modified memory allocation for faster loading
- Quake3LoadParam
added Parameter to the Mesh-Loader
- added
The still existing missing caulking of curved surfaces.
using round in the coordinates doesn't solve the problem.
but for the demo bsp mesh it solves the problem... (luck)
so for now it's switchable.
TJUNCTION_SOLVER_ROUND
default:off
- BurningVideo
- pushed BurningsVideo to 0.40
- added blendfunc gl_one_minus_dst_alpha gl_one
- added blendfunc gl_dst_color gl_zero
- added blendfunc gl_dst_color src_alpha
- modified AlphaChannel_Ref renderer to support alpha test lessequal
- addded 32 Bit Index Buffer
- added sourceRect/destRect check to 2D-Blitter ( slower, but resolves crash )
- added setTextureCreationFlag video::ETCF_ALLOW_NON_POWER_2
Burning checks this flag and when set, it bypasses the power2 size check,
which is necessary on 3D but can be avoided on 2D.
used on fonts automatically.
- added Support for Destination Alpha
- OpenGL
- Fixed a bug in COpenGLExtensenionHandler where a glint was downcasted to u8!!!!!!
MaxTextureSize=static_cast<u32>(num);
- TODO: COpenGLMaterialRenderer_ONETEXTURE_BLEND to work as expected
- Direct3D8
- compile and links again
- added 32 Bit Index Buffer
- D3DSAMP_MIPMAPLODBIAS doesnt compile!. it is d3d9 i think.
- compile for XBOX
- Direc3D9
- fixed crash on RTT Textures DepthBuffer freed twice.
added deleteAllTextures to destuctor
- NullDriver
- removeallTextures. added setMaterial ( SMaterial() ) to clean pointers for freed textures
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2148 dfc29bdd-3216-0410-991c-e03cc46cb475
- FileSystem 2.0 SUPER MASTER MAJOR API CHANGE !!!
The FileSystem is know build internally like for e.q the texture-, and the meshloaders.
There exists a known list of ArchiveLoader, which know how to produce a Archive.
The Loaders and the Archive can be attached/detached on runtime.
The FileNames are now stored as core::string<c16>. where c16 is toggled between char/wchar
with the #define flag _IRR_WCHAR_FILESYSTEM, to supported unicode backends (default:off)
I replaced all (const c8* filename) to string references.
Basically the FileSystem is divided into two regions. Native and Virtual.
Native means using the backend OS.
Virtual means only use currently attach IArchives.
Browsing
each FileSystem has it's own workdirectory and it's own methods to
- create a FileTree
- add/remove files & directory ( to be done )
Hint: store a savegame in a zip archive...
basic browsing for all archives is implemented.
Example 21. Quake3Explorer shows this
TODO:
- a file filter should be implemented.
- The IArchive should have a function to create a filetree
for now CFileList is used.
Class Hiarchy:
IArchiveLoader: is able to produce a IFileArchive
- ZipLoader
- PakLoader
- MountPointReader ( formaly known as CUnzipReader )
IFileArchive:
-ZipArchive
-PakArchive
-MountPoint (known as FolderFile)
IFileSystem
- addArchiveLoader
- changed implementation of isALoadableFileExtension in all loaders
to have consistent behavior
- added a parameter to IFileList * createFileList
setFileListSystem
allows to query files in any of the game archives
standard behavior listtype = SYSTEM ( default)
- CLimitReadFile
added multiple file random-access support.
solved problems with mixed compressed & uncompressed files in a zip
TODO:
- Big Big Testing!!
- Linux Version ( minor )
- remove all double loader interfaces where only the filename differs
(IReadFile/const char *filename). This blows up the the interface
- many loaders use their own private filesearching
we should rework this
- there are a lot of helper function ( getAbsolutePath, getFileDir )
which should be adapted to the virtual filesystem
- IrrlichtDevice
added:
virtual bool setGammaRamp( f32 red, f32 green, f32 blue, f32 brightness, f32 contrast ) = 0;
virtual bool getGammaRamp( f32 &red, f32 &green, f32 &blue ) = 0;
and calculating methods to DeviceStub.
implemented in Win32, TODO: other Devices
- irrlicht.h
changed exported irrlicht.dll routines createDevice, createDeviceEx, IdentityMatrix
to extern "C" name mangling.
for easier dynamically loading the irrlicht library and different versions
- ParticleSystem
removed the private (old?,wrong?) interface from the ParticleEffectors
to match the parent class irr::io::IAttributeExchangingObject::deserializeAttributes
TODO:
please test if the serialization works!
- Generic
- vector3d<T>& normalize()
#if 0
f32 length = (f32)(X*X + Y*Y + Z*Z);
if (core::equals(length, 0.f))
return *this;
length = core::reciprocal_squareroot ( (f32)length );
#else
const T length = core::reciprocal_squareroot ( (X*X + Y*Y + Z*Z) );
#endif
Weak checking on zero?!?! just to avoid a sqrt?. mhm, maybe not;-)
added reciprocal_squareroot for f64
- dimension2d
added operator dimension2d<T>& operator=(const dimension2d<U>& other)
to cast between different types
- vector2d
bugfix:
vector2d<T>& operator+=(const dimension2d<T>& other) { X += other.Width; Y += other.Width; return *this; }
to
vector2d<T>& operator+=(const dimension2d<T>& other) { X += other.Width; Y += other.Height; return *this; }
- C3DMeshLoader renamed chunks const u16 to a enum
removing "variable declared but never used warning"
- added a global const identity Material
changed all references *((video::SMaterial*)0) to point to IdentityMaterial
removed warning: "a NULL reference is not allowed"
- modified IRRLICHT_MATH to not support reciprocal stuff
but to use faster float-to-int conversion.
gcc troubles may they are. i'm using intel-compiler..;-)
- core::matrix4
USE_MATRIX_TEST
i tried to optimize the identity-check ( in means of performance)
i didn't succeed so well, so i made a define for the matrix isIdentity -check
for now it's sometimes faster to always calculate versus identity-check
but if there are a lot of scenenodes/ particles one can profit from the
fast_inverse matrix, when no scaling is used. further approvement could
be done on inverse for just tranlastion! ( many static scenenodes are not rotated,
they are just placed somewhere in the world)
one thing to take in account is that sizeof(matrix) is 64 byte and
with the additional bool/u32 makes it 66 byte which is not really cache-friendly..
- added buildRotateFromTo
Builds a matrix that rotates from one vector to another
- irr::array. changed allocating routine in push_back
okt, 2008. it's only allowed to alloc one element, if
default constructor has to be called.
removes existing crashes. ( MD3 Mesh ) and possible others ones.
A new list template should be made.
one with constructor/destructor calls ( safe_array ) and
one without. like the array since the beginning of irrlicht.
currently the array/string is extremly slow..
also a hint for the user has to be done, so that a struct T of
array<T> must have a copy constructor of type T ( const T&other ).
i needed hours to track that down...
added a new method setAllocStrategy,
safe ( used + 1 ), double ( used * 2 + 1)
better default strategies will be implemented
- removed binary_search_const
i added it quite a long time ago, but it doesnt make real sense
a call to a sort method should happen always. i just wanted to safe
a few cycles..
- added binary_search_multi
searches for a multi-set ( more than 1 entry in the sorted array)
returns start and end-index
- changed some identity matrix settings to use core::IdentityMatrix
- added deletePathFromFilename to generic string functions in coreutil.h and
removed from CZipReader and CPakReader
- s32 deserializeAttributes used instead of virtual void deserializeAttributes in
ParticleSystem ( wrong virtual was used)
- strings & Locale
- started to add locale support
- added verify to string
- added some helper functions
- XBOX
i have access to a XBOX development machine now. I started to compile
for the XBOX. Question: Who did the previous implementation?. There
is no XBOX-Device inhere. maybe it's forbidden because of using the offical
Microsoft XDK. I will implement a native or sdl device based on opendk.
irrlicht compiles without errors on the xbox but can't be used.
TODO:
- native XBOX Device
- Windows Mobile
reworked a little. added the mobile example to the windows solution for
cross development.
added maximal 128x128 texture size for windows mobile ( memory issues )
- Collision Speed Up
The Collision Speed Up greatly improves with many small static child-nodes
- added COctTreeTriangleSelector::getTriangles for 3dline from user Piraaate
- modified createOctTreeTriangleSelector and createTriangleSelector
to allow node == 0, to be added to a meta selector
- CSceneNodeAnimatorCollisionResponse has the same problem as CSceneNodeAnimatorFPS
on first update:
Problem. you start setting the map. (setWorld). First update cames 4000 ms later.
The Animator applies the missing force... big problem...
changed to react on first update like camera.
- add Variable FirstUpdate. if set to true ( on all changes )
then position, lasttime, and falling are initialized
-added #define OCTTREE_USE_HARDWARE in Octree.h
if defined octtree uses internally a derived scene::MeshBuffer which has
the possibility to use the Hardware Vertex Buffer for static vertices and
dirty indices;-)
if defined OCTTREE_USE_HARDWARE octree uses internally a derived scene::CMeshBuffer
so it's not just a replacement inside the octree. It also in the OctTreeSceneNode.
#if defined (OCTTREE_USE_HARDWARE)
driver->drawMeshBuffer ( &LightMapMeshes[i] );
#else
driver->drawIndexedTriangleList( &LightMapMeshes[i].Vertices[0], LightMapMeshes[i].Vertices.size(),
d[i].Indices, d[i].CurrentSize / 3);
#endif
#define OCTTREE_PARENTTEST is also used. It's skip testing on fully outside and takes everything on fully inside
- virtual void ISceneNode::updateAbsolutePosition()
- changed
inline CMatrix4<T> CMatrix4<T>::operator*(const CMatrix4<T>& m2) const
all two matrices have to be checked by isIdentity()
to let the isIdentity work always
-changed inline bool CMatrix4<T>::isIdentity() const
on full identityCheck->
to look first on Translation, because this is the most challenging element
which will likely not to be identity..
- virtual core::matrix4 getRelativeTransformation() const
Hiarchy on Identity-Check
1) ->getRelativeTransform -> 9 floating point checks to be passed as Identity
2) ->isIdentity () -> 16 floating point checks to be passed as Identity
- inline void CMatrix4<T>::transformBoxEx(core::aabbox3d<f32>& box) const
added isIdentity() check
- changed CSceneNodeAnimatorCollisionResponse
- added CSceneNodeAnimatorCollisionResponse::setGravity
needed to set the differents Forces for the Animator. for eq. water..
- added CSceneNodeAnimatorCollisionResponse::setAnimateTarget
- added CSceneNodeAnimatorCollisionResponse::getAnimateTarget
- changed CSceneNodeAnimatorCollisionResponse::animateNode to react on FirstUpdate
- changad Gravity to
- TODO: set Gravity to Physically frame independent values..
current response uses an frame depdended acceleration vector.
~9.81 m/s^2 was achieved at around 50 fps with a setting of -0.03
may effect existing application..
- SceneNodes
- CSkyDomeSceneNode
moved radius ( default 1000 ) to constructor
added Normals
added DebugInfo
added Material.ZBuffer, added SceneMaanager
- CVolumeLightSceneNode:
changed default blending OneTextureBlendgl_src_color gl_src_alpha to
EMT_TRANSPARENT_ADD_COLOR ( gl_src_color gl_one )
which gives the same effect on non-transparent-materials.
Following the unspoken guide-line, lowest effect as default
- added LensFlareSceneNode (from forum user gammaray, modified to work )
showing in example special fx
- changed SceneNode Skydome f64 to f32,
- AnimatedMesh
-Debug Data:
mesh normals didn't rotate with the scenenode fixed ( matrix-multiplication order)
- Camera SceneNode setPosition
Camera now finally allow to change position and target and updates all
effected animators..
a call to OnAnimate ( ) lastime < time or OnAnimate ( 0 ) will reset the
camera and fr. the collision animator to a new position
- Device:
added the current mousebutton state to the Mouse Event
so i need to get the current mouse state from the OS
-a dded to CIrrDeviceWin32
TODO:
- Linux and SDL Device
- GUI
- CGUIFont:
- added virtual void setInvisibleCharacters( const wchar_t *s ) = 0;
define which characters should not be drawn ( send to driver) by the font.
for example " " would not draw any space which is usually blank in most fonts
and saves rendering of ususally full blank alpha-sprites.
This saves a lot of rendering...
default:
setInvisibleCharacters ( L" " );
- added MultiLine rendering
should avoid to us CStaticText breaking text in future
- CGUIListBox
- changed Scrollbar LargeStepSize to ItemHeight
which easy enables to scroll line by line
- CGUIScrollBar
bug:
Create a Window and inside a listbox with a scrollbar or
a windowed irrlicht application
Click & hold Scrollbar Slider. move outside it's region.
Release Mouse. Go Back to Scrollbar.. it's moving always...
it's generally missing the event PRESSED_MOVED, which
leads to problem when an element is dragging, has a focus, or position loose
and gets focus back again. ( think of a drunken mouse sliding left&right during tracking )
so added the mouse Input Buttonstates on every mouse event
IrrDeviceWin32:
added event.MouseInput.ButtonStates = wParam & ( MK_LBUTTON | MK_RBUTTON | MK_MBUTTON );
TODO:
Linux & SDL
so now i can do this
case irr::EMIE_MOUSE_MOVED:
if ( !event.MouseInput.isLeftPressed () )
{
Dragging = false;
}
- bug:
Scrollbar notifyListBox notify when the scrollbar is clicked.
- changed timed event in draw to OnPostRender
Why the hell is a gui element firing a timed event
in a draw routine!!!!!. This should be corrected for all gui-elements.
- added GUI Image List from Reinhard Ostermeier, modified to work
added GUI Tree View from Reinhard Ostermeier, modified to work
shown in the Quake3MapShader Example
TODO: Spritebanks
- FileOpenDialog
changed the static text for the filename to an edit box.
- changed the interface for addEditBox to match with addStaticText
- changed the interface for addSpinBox to match with addEditBox
- added MouseWheel to Spinbox
- changed CGUITable CLICK_AREA from 3 to 12 to enable clicking on the visible marker
- CGUISpritebank
removed some crashes with empty Sprite banks
- IGUIScrollBar
added SetMin before min was always 0
changed ScrollWheel Direction on horizontal to move right on wheel up, left on wheel down
- IComboBox
-added ItemData
- removed IsVisbile check in IGUIElement::draw
- Image Loaders
- added TGA file type 2 ( grayscale uncompressed )
- added TGA file type (1) 8 Bit indexed color uncompressed
ColorConverter:
- added convert_B8G8R8toA8R8G8B8
- added convert_B8G8R8A8toA8R8G8B8
- Media Files
- added missing shaders and textures to map-20kdm2.
Taken from free implementation
- ball.wav. adjusted DC-Offset, amplified to -4dB, trim cross-zero
- impact.wav clip-restoration, trim cross-zero
- added gun.md2, gun.pcx to media-files
copyright issues!. i don't know from where this file came from...
i hope this is not from original quake2..
- added new irrlicht logo irrlicht3.png
i've taken the new layout. i should ask niko to use it.
- added Skydome picture to media files (skydome2.jpg) half/sphere
- OctTree
-added
#define OCTTREE_PARENTTEST ( default: disabled )
used to leave-out children test if the parent passed a complete frustum.
plus: leaves out children test
minus: all edges have to be checked
- added MesBuffer Hardware Hint Vertex to octtree
- CQuake3ShaderSceneNode:
- removed function releaseMesh
Shader doesn't copy the original mesh anymore ( saving memory )
so therefore this (for others often misleading ) function was removed
- changed constructor to take a (shared) destination meshbuffer for rendering
reducing vertex-memory to a half
- don't copy the original vertices anymore
- added deformvertexes autosprite
- added deformvertexes move
- added support for RTCW and Raven BSPs ( qmap2 )
- added polygonoffset (TODO: not perfect)
- added added nomipmaps
- added rgbgen const
- added alphagen
- added MesBuffer Hardware Hint Vertex/Index to Quake3: static geometry, dynamic indices
- added Quake3Explorer examples
- added wave noise
- added tcmod transform
- added whiteimage
- added collision to Quake3Explorer
- renamed SMD3QuaterionTag* to SMD3QuaternionTag* ( typo )
- updated quake3:blendfunc
- added crouch to Quake3Explorer
(modifying the ellipsiodRadius of the camera animator )
added crouch to CSceneNodeAnimatorCameraFPS
still problems with stand up and collision
- Quake3MapLoader
modified memory allocation for faster loading
- Quake3LoadParam
added Parameter to the Mesh-Loader
- added
The still existing missing caulking of curved surfaces.
using round in the coordinates doesn't solve the problem.
but for the demo bsp mesh it solves the problem... (luck)
so for now it's switchable.
TJUNCTION_SOLVER_ROUND
default:off
- BurningVideo
- pushed BurningsVideo to 0.40
- added blendfunc gl_one_minus_dst_alpha gl_one
- added blendfunc gl_dst_color gl_zero
- added blendfunc gl_dst_color src_alpha
- modified AlphaChannel_Ref renderer to support alpha test lessequal
- addded 32 Bit Index Buffer
- added sourceRect/destRect check to 2D-Blitter ( slower, but resolves crash )
- added setTextureCreationFlag video::ETCF_ALLOW_NON_POWER_2
Burning checks this flag and when set, it bypasses the power2 size check,
which is necessary on 3D but can be avoided on 2D.
used on fonts automatically.
- added Support for Destination Alpha
- OpenGL
- Fixed a bug in COpenGLExtensenionHandler where a glint was downcasted to u8!!!!!!
MaxTextureSize=static_cast<u32>(num);
- TODO: COpenGLMaterialRenderer_ONETEXTURE_BLEND to work as expected
- Direct3D8
- compile and links again
- added 32 Bit Index Buffer
- D3DSAMP_MIPMAPLODBIAS doesnt compile!. it is d3d9 i think.
- compile for XBOX
- Direc3D9
- fixed crash on RTT Textures DepthBuffer freed twice.
added deleteAllTextures to destuctor
- NullDriver
- removeallTextures. added setMaterial ( SMaterial() ) to clean pointers for freed textures
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2147 dfc29bdd-3216-0410-991c-e03cc46cb475
Revert my previous ITriangleSelector patch and implement garritg's great idea about retrieving an ISceneNode by index. Add an out parameter to ISceneCollisionManager::getCollisionPoint() to return the hit scene node. This is an API breaking change; the test and example apps have been updated.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2125 dfc29bdd-3216-0410-991c-e03cc46cb475
Change from dimension2d<s32> to dimension2d<u32> where practical. IImage and ITexture now have unsigned dimensions, as does screen size. This has had a significant knock-on through the code base, but it is more technically correct - the BEST KIND of correct.
GUI elements and draw2DImage methods still have and take signed dimensions, but we can look into this in future. Regression tested with the test suite and the example apps. Note that createDevice() now takes dimension2d<u32>!
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2092 dfc29bdd-3216-0410-991c-e03cc46cb475
If a light manager is not registered, the existing behaviour (activate the closest lights to the camera) will be retained unchanged. New example project added to demonstrate the usage, and regression tests and existing samples run. No regression test for the light manager yet, since I am more interested in not regressing the default behaviour.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2064 dfc29bdd-3216-0410-991c-e03cc46cb475
Change ray/bounding box selection so that the nearest node is determined by the actual collision point with the bounding box (or the distance to the box centre if the ray starts inside the box). Since this is a functional change from the old behaviour, I'm only committing to the trunk. This will be slightly slower than the old method, but should return the correct node; I consider accuracy to take priority over efficiency for collision.
tests/sceneCollisionManager.cpp has been updated, as has examples/07.Collision/main.cpp.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@2047 dfc29bdd-3216-0410-991c-e03cc46cb475
1) The active camera.
2) Receiving input.
This allows example 09.Meshviewer (with two cameras) to work as expected. Replace the magic numbers in that example with enumerated values.
This patch also exposes ISceneNode::SceneManager through a public getSceneManager() API.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@1974 dfc29bdd-3216-0410-991c-e03cc46cb475
Significant internal change to the way that FPS camera jump speed and collision response animator gravity interact. The behaviour is now much more realistic, but it will require you to adjust your jump speed and gravity.
git-svn-id: svn://svn.code.sf.net/p/irrlicht/code/trunk@1811 dfc29bdd-3216-0410-991c-e03cc46cb475
nadro
cutealien
hybrid
bitplane
engineer_apple
Rogerborg