This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "Gem - Graphics environment for multimedia".
The annotated tag, 0.90.0 has been created at 2cdd46a1926468edc1c894dc6ea94a3da39642fa (tag) tagging 92168fa160125848eb109f31e3873cd54704034b (commit) tagged by IOhannes m zmölnig on Wed Jan 23 13:17:15 2019 +0100 - Log ----------------------------------------------------------------- reconstructed fake release 0.90.0 (which never happened) (no author) (1): This commit was manufactured by cvs2svn to create branch 'v0_90'. cclepper (172): test for yuv update fixed/added some yuv stuff changed the color channel order for OSX fixed YUV code in pix_gain added yuv processing added yuv processing added missing yuv objects optimized pix_gain RGBA code added pix_blur improved quicktime performance and DV quality added pix_blur added blending added altivec code hq for DV added added Altivec cache function added pix_mix added altivec function added altivec function added YUV added altivec function initial commit added VBLsync for OSX added yuv and optimizations added secondscreen message to enable/disable fullscreen on a 2nd display added secondscreen message to enable/disable fullscreen on a 2nd display added secondscreen message to enable/disable fullscreen on a 2nd display added quality modes added quality modes removed post() in postrender new 3x3 yuv code added yuv code put OSX ifdef back in added LoadRam() added LoadRam() added LoadRam() added fps timing outlet added YUV function removed memset added yuv initial commit fixed a potential crash initial commit initial commit added buffer zeroing for YUV textures new pix_objects added the rest of the GL drawing modes changed speed message to take all float values added fps message changed secondscreen to use a minimum of 640x480 added fps message and outlet added #include <string.h> <stdlib.h> fixed the OSX format #ifdef initial commit added optimizations to YUV and a first try at an altivec function added optimizations to YUV and a first try at an altivec function initial commit - replacement for pix_blur with a much better algo changed the Mac format to GL_BGRA_EXT added 'mode' message to switch between power-of-2 and rectangle textures removed pointless macro added vec_dst for dst stream removed unused variable removed unused variables, added registers initial 'working' version fixed the OSX image loading code initial OSX version of mem2image() now works on OSX now works on OSX made CLAMP_TO_EDGE the repeat for TEXTURE_RECTANGLE_EXT fixed TEXTURE_RECTANGLE_EXT code for OSX added Altivec new integer processing mode - 4x faster than float initial commit tweaked some altivec removed debug post() improvements to YUV code added newfilm to state to tell when a film has been changed - mostly for faster rectangle texturing at this point fixed inaccurate estimation of film length, trying yet another way to do getFrame() fixed Ian's prefetch code added newfilm to state->image improved rectangle texturing improved altivec code improved altivec code + PPC scheduling more Altivec tweaks added some Altivec added Altivec YUV function now does color-space switching fixed up some altivec changed #ifdef MACOSX to __APPLE__ changed #ifdef MACOSX to __APPLE__ and commented out yuv_setup() added a few more drawTypes changed #ifdef MACOSX to __APPLE__ changed #ifdef MACOSX to __APPLE__ initial commit initial commit initial commit initial commit initial commit initial commit changed #ifdef MACOSX to __APPLE__ changed #ifdef MACOSX to __APPLE__ added GL_LINE_SMOOTH changed #ifdef MACOSX to __APPLE__ changed #ifdef MACOSX to __APPLE__ added #ifdef for PPC970/G5 added colorspace message added ram message, made YUV default colorspace changed #ifdef MACOSX to __APPLE__ changed #ifdef MACOSX to __APPLE__ changed #ifdef MACOSX to __APPLE__ fixed commenting error changed #ifdef MACOSX to __APPLE__ changed the colorspace message to match pix_film, dimension message now works changed gain removed processGrey added a better description added a better description changed #ifdef MACOSX to __APPLE__ change tv_ to pix_ added FSAA changed Altivec defines changed Altivec defines added RGB Altivec added RGB Altivec changed from float to int processing of grey value changed from float to int processing added RGB Altivec added RGB Altivec added Altivec slight change to RGB vector code a dst tweak and bug fix removed the extraneous RGB_Altivec prototype added Altivec function for both RGB and YUV new pbproj fixed error in code - it now compiles changed in to GLint to match prototype of csMess made client_storage a user defined option. this is a temporary workaround to some texture corruption issues. default is now set to YUV colorspace changed internal texture format back to RGBA so now alpha works in textures added hiquality message to deal with DV streams (yeah i know it's ass backwards to disable hiquality to make DV streams look good) rewrite to make playback using 'auto' more efficient. 'rate' message sets the playback speed when using 'auto' mode. the movie rate is properly set for correct qt playback. new doDebug for handling internal state info display to the console. added the real frame counting method and made improvements to using the counter method of playback - no double processed frames yet another rewrite of the QT tasking code. this should hopefully give the 'best of both worlds'. added support for groups added support for rectangle textures to glm fixed a little problem with newImage being multiply defined new pix_movie for OSX. uses an optimized path for movie playing and texturing. this should fix the window bounds problem. changed float to int processing in the YUV code - now 10X faster. split processing into RGBA and YUV portions. added Altivec for YUV. an improved window bounds fix. still need to find a workaround for the fullscreen on 10.3 though slightly better altivec. added some Altivec for YUV slightly tweaked Altivec (load hoisting) just a cleaning I think this fixes the majority of playback problems on OSX. Movie loading is faster, playback is smoother using 'rate $1'. added some more cache hints - 45% percent faster new fullscreen code for 10.3 which is good. the window deletion routine does crash every 10-20 iterations though, which is bad. fixed #ifdefs initialized the Y,U,V variables to 0. no more random color fields on loading a patch with pix_offset. changed the render function to only calculate the geometry when the geometry changes. results in at least a 3X speedup. now has functioning code on OSX removed most of the debug messages. added OSX specific handling of the 'fullscreen' message to accommodate multiple displays client_storage is now off by default to avoid texture corruption problems when using pix_film in 'auto 1' mode on OSX. setting individual frames still works fine with client_storage however. removed commenting out of m_colorspace changed GetLuminance() coefficients to correct vaules YUV smoothing now works fixed up the YUV code - rolled a new black function and tests for proper pixel values fixed the 'it's oh so green' YUV problem on OSX changed the shifts for YUV output. now 100% less green. #ifdef to remove YUV code on OSX. maybe someone will write a version that doesn't crash? First step at separating the color-space functions in order to have them work properly. RGBA is still ok, and now YUV works on 'uniform' adjustments. dheck (25): Fixed Win32 compilation issues. Updated Visual Studio project and workspace to build current CVS source. Incorporated source structure changes in CVS (TV->pix) and new source files Added 31 bytes to imageStruct buffer allocation to allow realignment without causing buffer overruns. Reverted GL_BGR[A] to GL_BGR[A]_EXT to allow compilation on Windows (openGL 1.1) Excluded config.h from Windows builds Added pix_chroma_key to the pix setup function and added the source files to the Visual Studio project. Added pix_videoDS, a DirectShow based video input object for Windows Added VC++ 7 workspace and project. Modified pix_filmNT to provide more robust error handling and to support recent CVS change in pix_film frame loading behaviour Added topmost message to keep gemwin on top of other windows. Implemented for win32 only. Fixed topmost message implementation for that topmost properly would be applied to windows created after a previous was destroyed. Added a "render" message to gemwin to allow a single graph rendering to be done. Also changed "frame" message to allow a frame rate of 0. Added "clear_mask" method to [gemwin]. Uses with the GLdefine object to generate GLbitfields for the gemwin's glClear. Fixed return in void function. Changed automatic non-const array allocation to heap allocated array for compiler compatibility. Fixed problems with previous commit: avoid memory leaks due to alternate return paths. Added DirectShow graph exporting for debugging with GraphEdit More error checking tweaks Added alias to old external name (interim measure... should remove it eventually) Made GEM release build include DirectShow support by default. Added pix_snap2tex for snapping directly to a texture. This may need some experimentation with appropriate pixel formats for each platform to achieve good performance. Minor changes for Win32 compilation. Added a Quicktime for Windows film loader. Added project builder project for Mac OS X Added explicit reset of column index to 0 to fix invalid access bug. ggeiger (26): changed "for (int i=0; ..)" scoping config.h will be autogenerated from config.in moved glm.c into model_loader.cpp included particle library stuff particle and glm are in Gem now removed search for particle and glm added new FFMPEG version of pix_film added pix_filmFFMPEG setup added HAVE_LIBAVCODEC,HAVE_LIBAVFORMAT remove MMX routines checked in int/unfolded pix operations added FFMPEG library checks fixed avcodec check bug Moved from GemLibs moved orb into Gem fix compilation bugs, removed ifdefs for ffmeg version disable depth test for alpha blended render chains. I think this is the correct way how alpha should work, still there is a message *auto* to turn it off bang on gemhead triggers a render cycle (was broken) change is in pd already added auto message description added numberbox to rotate light (cube seems to be very dark with 180) distribute particles according to their velocity if there is more than one particle per frame removed my name from startup re-used continueRender() function remove setup() function this is done through a static class structure now, that is created for each object Gem objects are added automatically to pd as soon as they are compiled in fixed alpha depthtest default value tigital (161): add MACOSX specific changes added support for YUV processing added support for MACOSX and YUV processing added support for MACOSX add support for MACOSX add support for MACOSX initial commit initial commit added MACOSX support added MACOSX support initial contribution added MACOSX support added MACOSX support; processYUVImage() added MACOSX support standardized to processDualGray() & processDualYUV() standardized to processDualGray() & processDualYUV() added MACOSX, YUV support via #ifdef's added MACOSX added MACOSX; initial YUV support added for MACOSX; initial YUV support added MACOSX support; cleaned up unused variables added MACOSX & YUV support reordered variable to shut up gcc 3.1 added MACOSX support; initial YUV support removed unused variable added MACOSX support added MACOSX support; initial YUV support added MACOSX support; initial YUV support removed unused variables added MACOSX support; initial YUV support added MACOSX support; initial YUV support *** empty log message *** GL_EXTENSIONS formatting same for all platforms reordered variables to shutup gcc3.1 added support for MACOSX added MACOSX support removed unused variable "count" cleanup unused variables added yuv video input for MacOSX; cleanup unused variables added processYUVImage(), so it works with yuv now! fixed a coupla possible leaks corrected rangeMess inputs made s_windowClock=NULL on DestroyWindow to address possible memory leak added gl.h & glext.h to allow compile on OS X fixed bug which prevented non-powerOfTwo textures from texturing correctly (m_coords are now set and passed into state->texCoords) removed HAVE_LIBGLUT for MACOSX cleanup added tv_delay_setup updated to current "IMAGE_CLASS" memory functions (ie. buffer.clear() & buffer.allocate( dataSize) removed redundant m_linewidth call removed redundant m_linewidth call; added glBlend added glBlend now has controls for size, speed, and mode; currently only works with [pix_texture]; also requires rotateXYZ, scale, transposeXYZ for centering (see example/texture) ripple: bang to inlet causes image rippling, center of ripple is movable via x and y input: see example/07.texture/ slideSquare: creates a field of squares which reaveal texture while moving left to right rubber: bang "grabs" image, x/y input the can pull/distort image, 2nd bang releases image, which rebounds back to flat added setup for ripple, rubber, and slideSquares changed gl setup calls to float (from integer): they should all be one or the other, right? example patch for [newWave] controls example patch for [ripple] controls - doesn't require textures now ;-) - finally centered! - doesn't require textures now ;-) - added glBlend message moved from tv_* to pix_* added *_setup() for objects from tv_* to pix_* reordered member initializers to "...datasize(0),pdata(0)..." to remove gcc 3.1 warnings is "Base/config.h" used by anyone anymore? is "Base/config.h" used by anyone anymore? #ifdef'd for linux because it was stopping OSX build is "Base/config.h" used by anyone anymore? #ifdef'd for linux because it was stopping OSX build added #ifdef DEBUG to post() in processYUV_YUV() added #ifdef for OpenGL/gl.h on OSX fixed OSX crash: needed check for GLcontext before executing glGenLists() switched order of getFrame() and "m_curFrame = m_reqFrame"...also added an idle MoviesTask() on OSX OSX only: changed getFrame() for non-auto-processing, so now we get much smoother playback using metro/counter, in addition to allowing scrubbing/random frame access via a number box! Finally! OSX only: forgot to set one flag correctly OSX only: turns out that SampleNumToMediaTime() only worked when 1 movie was loaded...switched to a much simpler method of calculating m_reqFrame's m_movieTime...which works with at least five simultaneously running films! rearranged initializers to stop compile warnings forgot to remove getFrame() variable duration! Thanks to the vigilent cgc uncommented pix_dot_setup stuff, because now it works! now it works! RGBA only now, but YUV on it's way...from the EffectTV collection cleaned up unused code, including removal of media timescale calculation, which could cause crashes, and is not currently necessary cleaned up unused code, including removal of media timescale calculation, which could cause crashes, and is not currently necessary removed loadBang so the patches will not crash on OSX: again, OSX needs a glContext before making any glCalls! removed unused (now) media variable removed unused variables, started work on processYUVimage switched ordering of variables to appease compiler removed unused variables reordered variables to reduce compiler warnings added glext.h to compile on OS X added #define for sqrtf() on OS X added #define for logf(), expf(), sinf(), & cosf() on OS X (sin() et. al. are already done w/double precision) #elifdef doesn't compile on OSX, but #elif def does correct casting in sprintf's to shutup compiler warning initial port: RGB only (so far) :-( cleanup a bit added pix_rds_setup() needed for some example patches (like pix_gravity) add #ifdef __APPLE__ to correctly include headers commented out call to pix_film_setup(): is this needed on any platform (seems like we went thru this before)?... added "scale" message to inc/dec size of dots, plus some cleanup into the destructor added "scale" message to inc/dec size of dots, plus some cleanup into the destructor; also made inline_RGB2Y() protected reordered declarations to shutup gcc vector long has been deprecated: now it's vector int's added OSX glcontext check to destructor to stop crashes when patch is opened/manipulated, but no glcontext is created stupid cleanup of commented-out stuff cleanup, changed to __APPLE__, solved casting problem added PeteHelpers.h stuff (defines, inline functions) fixed cast problem (UInt8* instead of Str255) alphabetized functions; also added new setup functions: pix_colorreduce, pix_halftone, pix_kaleidoscope, pix_levels, pix_lumaoffset, pix_metaimage, pix_refraction, and pix_backlight initial import of Pete's_Plugins ports (RGBA only so far): pix_colorreduce, pix_halftone, pix_kaleidoscope, pix_levels, pix_lumaoffset, pix_metaimage, pix_refraction, and pix_backlight initial commit added return to DestroyGemWindow() to prevent crashing on window close in some circumstances added blend message small typo fix to allow compiling on OSX small typo fix to allow compiling on OSX (space between FTFont* & makeFont(...) - added texture_range usage on OSX, where available: speed gain? - fixed problem with power of 2 textures not taking fastest paths (rectangle_textures), if available - added texture_range_supported flag - added texture_range_supported flag (default=0, ie. off) and extension check reordered initializers to shut up gcc warning changed comment from "mousePressed" to "KeyBoardPressed" removed unused variables added glFinish() & several glPixelStorei()'s for __APPLE__ for speed up?: is this applicable to all platforms? added default: break; to _pCallActionList(), because PAHeaderID isn't taken care of added default: break; to PAAvoid::Execute() & PABounce::Execute, because PDPoint, PDLine, PDBox, PDCylinder, PDCone, and PDBlob aren't taken care of...also replaced __APPLE__ defines with macosx_math.h replaced __APPLE__ define with macosx_math.h fixed "comparison between signed & unsigned int" warning on line 117, in processRGBAImage() added __APPLE__ "macosx_math.h", some unused cleanup removed some unused code removed some unused code, cleanup more work on yuv processing, but it still doesn't work initial commit added camera_setup() corrected text added Cyrille's changes: now it's more controllable (and useful): see examples/07.texture/05.newWave.pd added Cyrille's changes: now it's more controllable (and useful): see examples/07.texture/05.newWave.pd removed unused vector to shut up gcc 3.3 rearranged variables to shut up gcc 3.3 about initialization added (int *) casts to fscanf() calls to shutup compiler (someone had to do it!) values should be "TimeValue", not int's (shuts up compiler warnings) OSX only: added event handler support for window, mouse and keyboard events (ie. [gemmouse] and [gemkeyboard]), plus the window now behaves in a standard gui manner added mouse wheel support OSX only: added support for correct mouse buttons added mouse wheel support, though there is no outlet (yet) OSX only: added carbon event processing to dispatchGemWindowMessages() (Finally!) initialized variables to shutup compiler warnings OSX only: forgot to "return noErr" in event loop! [OSX]-only: [gemwin] now correctly responds to the following messages: border & title; also fixed keyboard character input for [gemkeyname] [OSX]-only: standardized mouse location to "window relative": however, remember that (0,0) is always the top left corner, whether or not there is a title bar/border [OSX] shut up some warnings shut up a warning: can't have /* inside a comment [OSX] finally figure out how to get mouse coords to be the same whether or not there is a title bar, and no matter where the window is on the desktop...also solves the "go to {0,0} when mouse leaves window" problem! [OSX] now the mouse can be hidden/shown via the cursor 0/1 message [OSX] cleaned up warnings and consolidated a previously used function added YUV support (kinda): it works, but isn't precisely correct yet added YUV support (kinda): it works, but isn't precisely correct yet (especially in pix_objects that are color based; displacement objects are less-obviously incorrect) quick n'dirty camera help patch...will flesh out tonight? [OSX] added header for definition of UInt32 after the mass movement of functions YUV is almost fixed: at least we're dealing with true luma...but there still seems to be a pixel skipping problem YUV works better, except for "smooth" line filling YUV processing functions correctly now removed void to shut up gcc YUV works better: but it's about half the width resolution as it should be...at least it's not green ;-) here's what I came up with: luma seems more sensitive, and makes a better image, tho different than the rgba version *** empty log message *** zmoelnig (618): This commit was generated by cvs2svn to compensate for changes in r4, which included commits to RCS files with non-trunk default branches. fixed the install-things applied Daniel Heckenberg's patch for DualPix-Objects - edited some example-patches (very minor changes) - rewrote the render-chain mechanism: now the render-chain is not built statically when the rendering is started but dynamically (each gem.object is "banged" (with some gem_state pointers) at run-time gosh: why do we have a message-system in pd, if it is not used ? - the priority of the [gemhead]s can now be changed dynamically applied guenter's MMX changes to some pix things (?) oops, last checkin i forgot some things.... ok: applied guenter's MMX changes to pix_* don't remember: added some example-patches i had forgotten before fixed the pix_subtract bug (which nobody had reported yet) added erich's [cuboid] object, so he doesn't have to worry any more... Makefiles should be distclean on CVS ...? *** empty log message *** made the CVS compile under linux again made CVS compile under linux again added a lot of openGL stuff... *** empty log message *** a new (simpler?) setup-logistic new image-processing calls (processRGBAImage) added framewise delay! rewrote the <processImage>-callers. processImage is now used for images of any(!) format. thus the programmer has to care for the format him/herself the format specific functions are processImageRGBA and similar the same goes for PixDualObjects (processRGBA_Gray etc...) deleted pix_fx the pix_fx functionality is now in GemPixObj this functionality is: save the image-state before it is processed and reconstruct it in the postrender function so we can change size/format and even bend the data-pointer and objects "before" will not notice (and crash) revised 3 openGL objects (urgh: lots of work to do still) GemPix*Obj: hopefully fixed the bug that appeared when deleting pix_fx now we cache the whole pixBuf-structure and return it on postrendering GemPixUtil: now provides a function for copying imageStruct-information GemGlutObj: if !HAVE_GLUT, we only throw an error at object-creation teapot: if !HAVE_GLUT we still support [teapot], only the rendering is void text2d: the mac-#ifdefs didn't compile on linux, so i fixed them pix_info: get information on pixes (xsize, ysize, csize, format, data) pix_yuv : convert images to YUV (for now only RGBA); needs testing pix_rgba: provides a function for YUV; needs testing * : new process*Image() functions for RGBA, YUV,... yuv_*: (think) i changed these to the new process*Image() functions added some examples for 04.pix; they were missing? added movement_detection videoRTX *** empty log message *** rewrote (regenerated) the openGL-wrapping stuff added openGL-wrapper objects (mainly openGL-1.1) rewrote the configure-thing hopefully it will work now... rewrote the configure-thing hopefully it works now... added an "m_modified" flag. this might double some code, but it was surely quickto do there are new ifdef's because of the new configure spit out a warning when compiled without GLUT (is this new ?) added the openGL-stuff added the anything2grey converter because of the new configure, some ifdef's have changed there names (HAVE_MPEG becomes HAVE_LIBMPEG etc) *** empty log message *** added this README, so you know what the openGL-stuff is about... changed the identifier for the list-ID from "list" to "gllist", because pd would complain about this ("list" is reserved for pd-lists) changed the behaviour of $(CXXFLAGS): you can now set CXXFLAGS via the command-line and it it added to the c++-flags $(cxxflags) added optimization to the Makefiles: "-O3 -funroll-loops -ffast-math" redefined _GL_UNDEFINED to -1 (instead of -10) undef'ed the MacOS-specific Carbon-include GLdefines maps symbols (like GL_FLAT) to corresponding floats (like 7424) This commit was generated by cvs2svn to compensate for changes in r139, which included commits to RCS files with non-trunk default branches. updated the README.build (for the new configure) added a reallocate() message to imageStruct. This does only allocate memory, if the previous buffer is smaller than the needed one. there is a "int datasize" that describes the length of the buffer. removed the default width/height from the constructor (this is done in the cpp-file) another IIR-filter (time-domain) for pixes: very much like [pix_blur] and [pix_biquad], but more flexible *** empty log message *** removing TV-class removed TV-class removing TV polygon_smooth enables GL-polygon-smoothing (like it has been proposed by the mac-people) now the alpha-blending function can be set by hand: 0 .. GL_ONE_MINUS_SRC_ALPHA 1 .. GL_ONE updated a link added libavifile-support under linux somehow these were not properly updated ... added AVIPLAY (for linux) and (not yet functioning) GLTT defines added avifile support fixed a bug with SYSTEMDOCS and made pd/extra the default installation-dir (not pd/externs) the object file for venus was missing... pix_crop takes sub-images of images updated configure.... checked in Makefile again... some bug-fixes; esp [pix_buf] bugfixed separator changed text-object, so they can be disabled... ??? [pix_depot] can store images (any size, fixed number) (TODO: add a way to load decompress movies directly into [pix_depot]). it is like [table] [pix_get] gets frames from [pix_depot] (like [tabread]) [pix_put] writes frames to [pix_depot] (like [tabwrite]) removed to redundant lines (mainly a label "BREAK") *** empty log message *** some ifdefs to compile on my version of ffmpeg really nothing (maybe some blank lines) nothing that matters (just a boolean cast at autoMess) allocate() now allocates data according to xsize/ysize/csize type is automatically set (depending on OS) some casts to bool (vc6 didn't like it the way it was done) *** empty log message *** added a boolean cast (for vc6) set the state->image to zero before rendering changed a type from bool to GLboolean (why ? i guess for vc6) a "new" concept of how os/library dependant code should look like. do not make the pd-object-class OS-dependent but a helper class. a threaded class for video capturing under linux GemPixObj: i think there was a bug with not initializing pointers the rest is no real changes (only "visiting" the files) nothing i know of the parent class for os/lib dependent video-grabbing (childs include videoV4L, hopefully will include video1394, videoDS, videoWDM,...) changed the base class from GemPixObj to GemBase (so we have info on every image). maybe this was not a great idea ...( but not that bad either) set the notowned flag to 0 (because we own the rgba-image) just a debugging output dos2unix added support for other colour-spaces (though YUV might produced interlaced signals) can't remember (maybe only some tab-indents) fixed a bug in the destructor the capturing is now threaded *** empty log message *** the quicktime4linux movie-loader class added newlines at EOF fixed a bug on close added some casts to integer (warnings on gcc-3.2) double default argument fixed fixed the float2int bug (haven't i submitted this 4 times today ?) added filmQT4L though i fixed the linux-Video thing but someone did this for me... there was a <stdlib.h> missing in pix_dot.cpp added #include <string.h> new startup-logo with more authors added a dos2unix on make distclean oh i think pix_videoNEW is really buggy. this is BUGGY ? (i think) (refined) RGBA_Gray-support for PixDualObjects made an explicit cast from float to integer during the processing what to say ? making it work changed the API a little bit... suddenly i discovered that gcc-3.2 (?) didn't want to compile this... [part_render] let you render any gemlists as particles [part_info] is similar, but doesn't actually rotate/translate/... the gemlist, but outputs the data on the right outlets initialized some values (gcc-3.2 would complain) (nothing) removed some unused variables removed unused variables and initialized used ones swapped 2 lines (GLTTBitmapFont and GLTTPixmapFont) because gcc-3.2 would complain new indentation ;-) initialized a variable made texture_rectangle_supported work with nvidia... found a bug @ parsing the openGL-extension when no openGL-context was yet valid added the continueRender() function (important to output several gemlists (like with [part_render]) uncommented the stoprender() thing. what is it for ? (nothing), like dos2unix... tried to add the nvidia-specific defines for yuv422 and rectangle-texturing (however, the yuv-thing still doesn't work) cursorMessage should return something. it does so now. a new flag "--with-ieee1394" for linux-compiles (doesn't do anything yet...) ??? nothing, but made the constructor virtual (since the class is virtual too) made film-stuff compile with gcc-3.2 without warnings... (hope that is all) *** empty log message *** added a method for changing the color-space ifdef'ed GL_UNPACK_CLIENT_STORAGE_APPLE removed unused variables and initialized used ones. added a comment fixed things with old libaviplay, that doesn't work (since it only decodes audio) new indentation ;-) added a method for Gray played with the colour-spaces merged pix_texture and pix_texture2 and made rectangle-textures work with nvidia (hope it works...) a not yet working ieee1394-videoIN under linux examples for user-defined particles stupid bug when adding to grey images added some comments (is this enough documentation ?) upgraded to libparticle-1.20 upgraded to 1.20 upgraded to libarticle-1.20 this effects some of the [part_*] objects (eg: [part_info] now has another output for the age of the particle,...) i once uncommented the "realStopRendering" things, because i didn't know what it was for. trying to compile gem2pdp i noticed that this was needed. yes, i don't like it either, when changes are made to the repository wihtout documentation used to crash when the data-pointer was zero since datasize in imageStruct is now private, we use setBlack() to delete the image-buffer added colourConversion routines to GemPixUtils (for now only non-optimized) maybe they should rather be in GemPixConvert ? datasize is now private there is now an alias for [pix_separator] added Grey-Grey support first attempt to support my Philips Quickcam (720/40) added alias for [pix_separator] set output format to GL_ARGB_EXT on macOS. i hope this works, but maybe not added a deviceMess for setting which device to use renamed [pix_depot] and fellows to [pix_buffer] et al. [pix_put]: the frame# is now deleted after an image has been written into the depot. the next incoming image will not be stored until another frame# is given. added aliases for [pix_depot] and friends added convert and emboss yet another colour-space converter ??? don't know whether this makes sense (remove it LATER) copy and pasted the yuv_emboss into the Pixes source. for now, only yuv is supported on platform that do not support yuv422-openGL-renderings (i haven't managed to make my geforce behave the way i expected, and of course there are other cards that will never support yuv-textures), we now have implicit yuv2rgb-conversion for testing i had hardcorded requests for some "wrong" colour-spaces. hopefully this is removed now oops, missed one thing: when yuv was requested, we didn't directly captured to yuv422 but rather to yuv420p. fixed ifdef'ed the call for openGL-extension "rectangle-texture", since there were problems when Gem was compiled without rectangle-texture-support, but the drivers supported it (namely nvidia-driver) added a topmostMess (a nop) added data-acquisition "fromYUV422" added offsets chY0, chY1, chU, chV for UYVY interleaved data added a string2format converter (like 'rgba' to GL_RGBA), only Grey, YUV, and RGBA are supported removed tv_* help-patches, since they are not used any more ifdefed the use of GL_UNPACK_CLIENT_STORAGE_APPLE obviously some other changes, which i hope do not break anything, but i guess most of it is new indentation switched back from PALETTE_YUV422 to PALETTE_YUV420P. it seems as if video4linux means something different with YUV422 (or similar formats). too bad, conversions (copying!) again copied yuv_posterize to pix_posterize. YUV only. hope this works as it should moved macors for image-processing from GemPixUtil to GemFuncUtil (since they are functions, although they probably have not much use for things other than PixProcessing (GemPixFun.h ?) commited result of merge This commit was generated by cvs2svn to compensate for changes in r380, which included commits to RCS files with non-trunk default branches. fixed a typo that searched for "searchlibs" instead of "$searchlibs" fixed a typo that searched in "search_libs" instead of expanding $search_libs #ifdef'ed everything with HAVE_DV (so it won't compile if there are no libdv/dv.h headers) added a BUGS file file bugs under BUGS, and things that have to be done under TODO moved BUGS to Gem/src (../) updated "#elif def" doesn't compile under linux, but "#elif defined" does... bug-fix: there was weird-behaviour, when the offset was bigger than the size of the cropped(!) image added method for grey-images we now have chY1,.... for YUV-offsets, i have used this now since movement is stored in the Alpha-channel, i h?\008changed processImage to processRGBA image. added greyscale-support for pix_blur put the [pix_grey] back in again (why isn't it there ?) shouldn't the "snap" message be replaced by a "bang". (at least i have *add*ed it) commented out the ALTIVEC after "#endif" (gcc doesn't like this) set the include to "Base/GemPixDualObj.h" (the Base/ was missing) added some more yuv-conversions. (following a new naming-scheme after the FOURCC) fromUYVY, fromYUY2 (=YUYV), fromYVYU, fromYV12; fromYUV422() is now an alias for fromUYVY() fromYUV420P() is now an alias for fromYV12() fixed the dv-checking. force the headers to be there, when the library is found too changed the error-message (like for setting the device) slightly hopefully this is a real support for video over ieee1394 (firewire/dv) under linux! it seems like this is only used under linux for now. however, i have added the videoDV4L support. and there is now a "driver" message, that let's you choose between various drivers (like video4linux and video1394) yes, the way the modules are registered is really BAD. have to change this soon use image.reallocate() rather than image.clear() + image.allocate() removed the Makefile.depend . they are not really needed in CVS oops, made a type "libdv" instead of "dv" oops, i'm getting tired. another typo added some advice for video1394 under linux added a flag to WindowInfo, whether a constant context has been established. the destroyWindow depends partly on this (because my laptop (with an ati mobility l6) crashed when destroying a non-constant context. also changed the line "do not continue" to "continue at your own risk" filedescriptors may also be 0! i only closed the device, if the fd was>0. fixed. added checks for mmx-extension of gcc got some problems with the cvs-update on another machine. hope it's ok now added a newline at the end of the file still some cvs-merge errors, probably they are gone now... added FTGL support for text-rendering under macOS i have made FTGL the default (not knowing whether it will break) under windos GLTT is still the default (although FTGL should work too) under linux the configure script decides what to take, but prefers FTGL rather than GLTT added a new object for extruded text (FTGL only) you can now set the domain of the source TODO: make the argument vector settable too (not much use without) include <iostream> rather than <iostream.h> [part_velocity] sets the velocity into a specified direction [part_veldomain] sets the velocity into a specified domain (like sphere, triangle or whatever) [part_vertex] adds one particle each rendering cycle at a certain position -- useful for moving particle sources (but alas, only one particle at a time) added [part_vertex], [part_velocity], [part_veldomain] added checks for FTGL help for the resp. object just render a video onto a geo a an example for how to use [pix_buffer], [pix_buffer_read], [pix_buffer_write] a an example for how to use [pix_tIIR] a started (in April) primer for Gem, however, far from being finished updated reference-patches. [part_veldomain] is replaced by [part_velocity] [part_veldomain] has been replaced by [part_velocity] [part_velcone] and [part_velsphere] are obsoleted by [part_velocity] added an "obsolete"-warning added settable domains+arguments added support for grey and yuv images. (not tested) added FTGL-2.04 -- support model.cpp_ seems to have been committed to the CVS. it was just a backup for work... the particle-rate is now float instead of integer (let's you emit 0.5 particles per rendering cycle) general entry for pix-bugs fixed several bugs in the color-conversion routines use the color-conversion from GemPixUtil; pix_grey and pix_yuv are childs of pix_rgba (reduces code dramatically!) added a newline at end of file ;-) removed functions for YUV/YUValtivec/Grey images (why do we have classes ?) added an include for <assert.h>, this might break compilation on other platforms, if so, please ifdef it to __linux__ removed functions for YUV/YUValtivec/Grey images (why do we have classes ?) removed empty functions for YUV/YUValtivec/Grey images (why do we have classes ?) dos2unix (changed linefeed/carriage-returns) added checks before delete[] (just in case) added greyscale-support addedsome checks before destroy[]. hope this will fix the bug removed the create/destroy/blank routines for the buffer if m_interlace was 0, this used to crash my machine with a floating-point exception. thus "if(interleave<=0)interleave=1;" fixed the crash however, this is very slow. we'll have to write a faster routine ourselves added postrender() hack, to prevent the "source-object" to be drawn hopefully fixed a bit on installing the examples for linux added a comment removed debugging output removed debugging output made a list of undocumented objects updated some reference patches result of merge new help patches version is now 0.888-pre1 new indentation packed the code a little bit (to make it more readable to me) made some inlets made the heightMess do something arguments specify the grid-size made some inlets made the heightMess do something added methods for YUV added methods for YUV/grey added methods for RGBA/grey added methods for grey added inlet for size of dots removed overwrites of the virtual functions for yuv/grey-colourspace. Why do we have inheritance ? fixed the bug with filledLines added a comment, saying what this object is for added methods for YUV grabbed texture-size defaults to the gemwin-size BUG: the first incoming frame is captured automatically --> FIXED (nothing is done, if no frame is specified) added comments about ranges of the internal variables. also changed the range from 0..255 to 0.f..1.f added checks for control-range added some inlets changed ranges from 0..255 to 0.0..1.0 added some inlets cleaned up the code a bit. default behaviour is no ON (used to be OFF) cleaned up the code a bit. default behaviour is no ON (used to be OFF) RGBAcomparision is now done on luminance instead of each channel separately i think i've found a bug in GREY new indentation ? allow arguments and an intelligent gain handler code cleanup to make merge easier added methods for Grey made the inlet do something intelligent: set the range (normalized 0..1) allow arguments and an intelligent gain handler result of merge fixed the "toupper()" BUG: include <ctype.h> documented a lot ;-) fixed install on linux more bugs to come added some documentation fixed setBlack() and setWhite() for YUV added Gray-support added Gray-support (RGBA & Gray are in processImage, while YUV has a separate function) added YUV-Support (like grayscale: the difference image is written on the luma-channel; colour is discarded) added Gray-support split the processImage into processRGBAImage and processGrayImage added an inlet for the size of the metaimage added Gray-support YUV-support is still under construction (but now does something (stupid) instead of nothing) updated documentation added YUV support split processImage into processRGBAImage and processGrayImage removed the processRGBAImage-call (anyhow, it didn't do anything) filmQT4L procuces RGB if RGBA is not supported by the codec removed #include <OpenGL/gl.h> (it is included in GemBase anyhow) ifdef'ed the #include <OpenGL/glu.h> to __APPLE__ (else <GL/gl.h> put quite everything (but constructor/destructor) between one huge #ifdef. the rest should be filled with defaults by the film-class. on close() the filehandler is set to NULL (caused to crash on multiple closes) put quite everything (but constructor/destructor) between one huge #ifdef. the rest should be filled with defaults by the film-class. GL_RGBA is not defined at openGL<1.2 (i think); furthermore the include-check for CPPExtern.h is already made in CPPExtern.h added fromRGB16() to retrieve data from 16bit encoded RGB. added fromYV12(Y,U,V) which does not assume that the U and V planes are directly after the Y plane. this is handy if we have independent planes added some thoughts about requested/forced colourspaces. i think, that film* should at least return colour-spaces that could be handled by Gem/openGL itself. (it doesnt make sense for [pix_film] to handle IMG_FMT_RGB15) so the cs-request should really be a force and not just a suggestion rewrote the getFrame() to: a) get frames of different colour-spaces b) decode YUV/Gray really fast (directly access the decoding Y/U/V-planes) encapsuled everything but constructor/destructor within one big "#ifdef HAVE_QT4L" (using inheritance) made getFrame() return frames of all three colour-spaces. BAD: we only get RGB-images from the qt4l-library. this means: video(YUV)->buffer(RGB)->gem_image(RGBA/YUV/Gray). but what should i do ? encapsuled everything but constructor/destructor within one big "#ifdef HAVE_QT4L" (using inheritance) made getFrame() return frames of all three colour-spaces. fixed several bugs rewrote a bit in the #ifdef clauses and put them all together. it is really ugly on linux pix_filmNEW is now the default a bit of code-cleanup requestColor() now sets "m_wantedFormat" open() now returns false (to make the skip to next api work) added colour-change on the fly put quite everything (but constructor/destructor) between one huge #ifdef. the rest should be filled with defaults by the film-class. made the openStream sleep a bit. still opening streamed-files sometimes crashes added support for more colour-formats open() now utilizes setCsizeByFormat() some bug fixing added "colorspace"-method "post" becomes "DEBUG" which can be turned off. some reordering of decode-libs: libmpeg3 is better in decoding mpegs then libavifile (less crashes) added "colorspace"-method moved the pix_film things up a bit, because i had a hard time making [pix_film] an alias for [pix_filmNEW] made [pix_movieNEW] (for now; of course this should [pix_movie] LATER a friend of [pix_texture] this will let us use the sophisticated texturing mechanism of [pix_texture] for [pix_movie] added "upsidedown" to imageStruct; this is meant for texture-flipping of the image (most movies are loaded the "wrong" way) upsidedown defaults to true on __APPLE__; false elsewhere using "upsidedown" from imageStruct; added this flag to setTexCoords() to indicate whether the texture-coordinates should be flipped or not. the __APPLE__ define for flipping is therefore obsoleted. using "upsidedown" from imageStruct; using "upsidedown" from imageStruct; added YUV support this is how [pix_movie] looks like if derived from [pix_filmNEW] (and [pix_texture]) reported bugfix of filmQT4L removed the [pix_rgba], since now [pix_film] should output only RGBA/YUV/Grayscale now using [pix_buf] instead of re-loading the image for each blending-change now you see something after the first coordinate-change... commented out namespace std for windows (vc6 complained about this) my vc6-dos-compiler doesn't like <iostream> because it makes it like against some "new" io-libs while it rather needs "old" ones which are requested by <iostream.h> removed "m_film", since it is not needed <unistd.h> is only included if HAVE_LIBAVIPLAY is defined for libavifile we used to include <avifile/avifile.h>; since the headers for libavifile-0.7 is located in /usr/include/avifile-0.7/, we now search for the correct path in the "configure" and include only "avifile.h" made filmAVI work. a first step for pix_filmNEW to run on windows... first attempt of a port from [pix_filmQT] to the film*-class. however i have quicktime4windows not installed, so i had no way to compile this yet.... search for libavifile-headers rather than assuming they are in /usr/include/avifile disabled warnings on vc6 updated my vc6-projects vc6 doesn't like statical arrays with variable sizes disabled pete's "static"-hack with NO_HACK (default on windows) added filmQT changed "#define DEBUG" to "#define debug" new bugs (pix_buffer, libavifile, remote display) fixed a bug when searching for the linker/compiler flags for libavifile: not "avifile-config" is used if available temporarily added some variables (like m_auto) these should be gone when i really have access to a machine where quicktime4win is installed set the upside-down flag to true. (hope this is alright) added "fsaa" to the WindowHints-class fixed bugs in pix_buffer_read/pix_buffer_write when no or unknown pix_buffers were specifified further changes, but far from working yet added the self-registering code, so that MarkEx will stay available fixed another possible bug (check for NULL-images) added some comment to the render() for neither FTGL/GLTT support (while trying to find out added "--with-aviplay" as an alias for "--with-avi" fixed some BUGS... made it work under windows; however it likes crashing sometimes and i haven't found out yet why... made it compile with ffmpeg-0.4.8 (however i wasn't yet able to load a movie...) added no-warn-flags for vc6 changed (bool) to (int), since vc6 doesn't like assigning a boolean-value to an integer. fixed linefeed changed processRGB_Altivec to processRGBA_Altivec removed the check, whether the first atom of the text-list is a symbol. now we can print numbers too. added scaling for the mouse-coordinates. this allows output that is independant of the window-size added curve3d and tube swapped 2 variables in class-definition (apple's gcc likes it better this way) removed unused code put the auto_register code in a separate define whichcan be switched off by defining NO_AUTO_REGISTER_CLASS before including the CPPExtern.h defined NO_AUTO_REGISTER_CLASS to not register this class with pd on __linux__, __APPLE__ and NT include "config.h" only on linux note: this should be changed LATER to include "config.h" on all platforms. config.h should then include platform-specific config-files (that can be generated with configure or whatever) removed unneeded variables updated documentation added curve3d and tube added MIN and MAX functions made XWin be aware of resizes. however, it doesn't really work yet :-( added an alias for "colorspace"-message removed the auto-registering changed "post()" to "debug()" which is a define to either "post()" or nothing, so we can do some debug-information on demand help-patches for [curve3d] and [tube] this is my macOS-projectmaker-project set notowned=false when color-conversion is done (which is the truth) added HELPSYMBOL to set an alternate helpsymbol (it crashes, when more than 1 help-symbols are specified and we already do it here) added support for HELPSYMBOL_BASE (which is defined as "Gem/" in GemBase.h) added support for HELPSYMBOL_BASE (which is defined as "Gem/") -- used in CPPExtern.h initialized newfilm/newimage to 0 made config.h system-independent. the system-bound configurations are stored in separate files (like Base/configLinux.h) which are included by config.h (depending on __linux__ and friends) made config.h system-independent. this the config.h.in for linux-autoconf has moved to configLinux.h.in added "--disable-new" for "old" behaviour with pix_film/pix_video and friends under linux this the config.h.in for linux-autoconf has moved to configLinux.h.in include "Base/config.h" for all OS (since the config.h will separate the sysdependent things) get more information the last outlet (data) is still the last outlet (but now there are 6 instead of 4). this shouldn't break anything, since who can do something with it ? include "Base/config.h" for all OS (since the config.h will separate the sysdependent things) exclude "GemBase.h" and use "m_pd.h" instead (shouldn't be needed either, if it wasn't for post()) setting newfilm to 0 (shouldn't do anything since we are on linux (?)) used setBlack() instead of memset(); use reallocate() disabled the auto_registering if __NEW__ is not defined set the help-symbol to "pix_video" (instead of "pix_videoNEW") moved the disabling of auto-register to pix_video.h set the help-symbol to "pix_movie" (instead of "pix_movieNEW") disabled the auto_registering if __NEW__ is defined rewrote the "colorspace"-callback include "Base/config.h" for all OS (since the config.h will separate the sysdependent things) moved the disabling of auto-register to pix_video.h set the helpsymbol to "pix_film" instead of "pix_filmNEW" uncommented the use of canvas_getfilename since it makes problems with internet-streams (as far as i remember) disabled the auto_registering if __NEW__ is not defined exclude "Base/config.h" since it is already included by film.h fixed a stupid bug, that was blacking the "last" buffer each cycle. added object-arguments as initial parameters take care of colorspace-changes using setCsizeByFormat instead of doing it by hand setCsizeByFormat now also sets the "type" (pixel alignment) on __APPLE__ the default is yuv, everywhere else it is rgba using setCsizeByFormat() instead of doing this by hand also there is now a m_reqFormat defining the colourspace (but only GL_RGBA is supported....) support for different colour-spaces (use "colorspace"-message) support for different colour-spaces changed the configure a bit to make more intelligent searches in the search-paths, for both libs and headers we now search /usr/local/src and /usr/local/include instead of whole /usr/local (at least for headers) removed offsetMess() and swapMess(); should be handled by the parent-class use fromBGR() rather than doing it by hand. this enables other color-spaces too! added a colorspace-message if we HAVE_DIRECTSHOW, we disable the AUTO_REGISTER stuff, if(!) DO_AUTO_REGISTER_CLASS is not defined. pix_videoDS will define the DO_AUTO..., to register itself changed the m_colorspace, to not only be 0 (for YUV) and 1(for RGBA) but rather GL_YCBCR422_GEM and GL_RGBA removed swapMess() and dimenMess() - should be handled by the parent-class bugfix: configure wouldn't add "-l" flags for "standard" libraries found with AC_FIND_LIB updated bugfixes ("create" on remote X-servers works for most but my machine) another bugfix (pix_videoDS) reported rebuilt it with another version of "aclocal" rebuilt configure to add the "-l" flags another bug fixed for NT configLinux.h should be generated for all machines by the configure-script. don't know, how this came into the CVS once in a while i compile under NT... split __NEW__ in __VIDEO_NEW and __FILM_NEW added some braces to make execution order clearer made it a bit more stable on windos a bit of support for colourspace-switching colourspace-switching with GL_* rather than 0 (YUV) and 1 (RGBA) use __NEW_FILM rather than __NEW__ support for colour-space switching upside down added dummy messages for "enumerate" and "dialog" (used by pix_videoDS) "dialog" with no args is now the same as "setup" ("setup" message should be deprecated) added dummy messages for "enumerate" and "dialog" (used by pix_videoDS, in the future probably by others...) filed YUV-bug of pix_kaleidoscope my new windows-project BUGFIX: grayscale was missing the bugfix for Grey-support needs a header-file too... bugfix: one additional inlet was created for [primTri], [curve] and [polygon] by GemShape. i have added a new creator for GemShape (without arguments) that will not create the "size"-inlet - this is for objects that want only the "draw" and "width" message reviewed help-patches new help-patch from cyrille dos2unix -- linebreaks updated documentation added checks for the pd-version (because of reference install-paths) bugfix (videoV4L crashed when you tried to open a non-existant device) use integer calcs rather than floating point fixed typos updated doc my project-builder project has been updated updated documentation glu-objects (like sphere) have a bit of a problem with non-power-of-2 textures better checks for pd-version re-added inlets for height, size and mode transformed and scaled the newWave so that it behaves exactly like a square of the same dimension when there is no motion. added an include for windows removed a debug message added newImage-variable since it was needed use a movie rather than an image for demonstrating this updated docs updated documentation several BUG-fixes (esp: put the pivot-point of rubber/ripple into the center; made the sizeMess-work; use normalized coordinates for rubber (now independent of the window-dimensions) i think, all objects are documented (somehow: some help-patches are rather crude) YUV-support fixed (the image looks "different" (squashed) than RGBA, but at least it is functional and not crashing any more), Grayscale works great hmm: probably new linebreaks updated fixed bugs updated my vc-project added support for rectangle textures even with sphereMaps ("texture 1") re-removed them (somehow they reappeared in the CVS; guess it has to do with sourceforge's server-sync) ifdef'ed everything to be __APPLE__ only fixed types to keep compiler silent hopefully fixed fromUYVY since it did mix up U and V when decoding dv-streams exported the CVector class, since vc complaint about this (probably this will break future things...) ifdef'ed the inclusion of m_pd.h to make sure it is not included double (breaks with old version of pd) added some support for devices not handled by devfs; added upside-down flag decode RGBA by hand (since otherwise we would decode YUV->RGB->RGBA, all in software...) post "video4linux" and remove post of "fish" ;-) put QT-loader *after* the AVI-loader, because QT sometimes crashes when having loaded AVI's and tries to load QT's afterwards. so AVI is now handled by the built in (for win) routines, and QuickTime is only used if it fails updated my project all docu is done; but we need to compile it under windows with FTGL support (i failed...) moved the defines for the platforms (e.g: whether macos uses FTGL) into the appropriate config*.h files updated my vc-project it now compiles with FTGL as ben has pointed out, there seems to be a bug when rendering to non-square windows: everything get's distorted (say: on 640x480 a square would be a rectangle with a 4/3-aspect); since i think this might break a lot of things (?), i have left the old default behaviour but you can turn off compatibility with "compat 0" BUGFIX: used to crash when no model was loaded and trying to apply a texture. (at least if NDEBUG was not defined) added and disabled Guenter at the splash-screen (again) - added ErrorHandling, so Gem doesn't crash if the display-/window-context has vanished before we try to destroy the gemwindow - removed checking for ModLines on remote displays; this used to crash; unfortunately this disables fullscreen on remote displays more friendly output of the "create window" message added a newline at end of file now the gemwindow can be created even if we only have DirectColor (instead of TrueColor); however i haven't bothered yet with setting up the Palettes. fixed a type that prevented this from being compiled added better support for firewire under linux (thanks to franz hildgen) fixed a bug with filename-searching updated my project hopefully got the reference-types allright ("relative" instead of "absolute") added path-searching for fonts. so you can add your systems favorite font-path and get a lot of other TrueType-faces added a new justification "BASE" that does no justification at all; usefull, if you want a constant base-line. pix_depot and friends are now in pix_buffer et al. added the HELPSYMBOL-BASE "Gem/", so the help-patches will work, even though this object inherits from CPPExtern and not from GemBase ifdef'd the altivec code with __VEC__ changed the configure-process GemLibs can now be specified with "--gemlibs" additional search-paths for ALL libraries can be specified with "--libdir" small bug-fix: a unity gain of "1" was translated to 255>>8 (which is less than 1 and led to fading feed-backs) tried to fix rectangle textures tried to fix rectangle textures set the "newfilm"-flag when a new film is loaded force auto-registering when appropriate the mac-spefific button-offset has been moved to GemWinCreateMac randF is only an alias for randomF; there shouldn't be a separate help-file got the mapping from mac-buttons to gem-buttons right (hopefully) changed the helpsymbol added a method "width" for the linewidth (compatibility with other geos) new help-patches (e.g: working examples) removed the "maximum number of control points" restriction (so now, we can have 1000s of control points in polygon/curve) added grayscale support updated help-patches updated help-patches removed all traces of [pix_texture2] removed dependencies on GLUT [teapot] is now "built-in" cleaned GLUT-removal code a bit rolled the GLU code directly into the objects to enable rectangle textures fixed bugs fixed rectangle textureing updated help-patches removed to check in as "binary" again readded help-patches removed "done" message added auto-(re)loading of fonts when rendering is started fixed some cleanup-issues fixed initial materials (ambient, specular, diffuse, shininess, emission) added, since needed for debugging each gemhead now has an instance of the GemCache that is only reset when rendering is started (instead of being destroyed and re-created) hopefully this fixes some memory bug. added a reset()-method to re-set an instance without having to destroy and re-create it (see [gemhead]) removed the "drawing all groups" messages removed another post() initialize GemCache on creation, so we can dynamically create [gemhead]s and they will start rendering without having to restart the engine added a magic-number to the GemCache to ensure that the cache is valid added includes for NT compiled it and therefore updated my project files type change to get rid of compiler-warning under macOS accidentially the button-state was reported to be "1" when released (like on press); fixed removed unneeded variables to shut down macOS-gcc removed unneeded variables to silence macOS-gcc version is now ReleaseCandidate-1 compiled under windoze and thus updated my project-file updated documentation new help-patch removed "text"-pd files (should be binary) added as binary readded as binary re-added as binary (pd-files shouldn't be text) fixed line-breaks finally removed the YUV-stuff (should have been done for ages) i guess we should release rather sooner than later included the Base/config.h so we get the HAVE_DIRECTSHOW define when appropriate set the help-symbol of the directshow-version of pix_video to [pix_videoDS] (since the api is so different) fixed some includes (?) moved implementation code from header-file to cpp-file removed some verbosity fixed the help-patches (mainly removed the "auto 1" message from [pix_buf] as it was absolutely non-sense all the CPPExtern-objects now no their own class-name, which is stored in m_objectname error-messages report this name removed the non-mac specific part in the fullscreen-message (no one really cares if the fullscreen-flag is bool or int) the processGray tried to convert the image from RGB to RGBA in-place. i have no idea, how this code made it into the [pix_alpha] added second argument (number of slices) the re-creation of the sphere is now triggered by a change in state->texture too added 2nd argument for number of slices fixed the draw-type an example for texture-feedbacking added british alias-name added US-english alias-name ("gray") removed verbose output removed the genList from the new()-function and put it into render() (this was default on macOS already) added switchable support for FSAA under linux with nvidia-cards removed unused variable G fixed the bug that kept cs-converters from working on apple added some more verbosity on errors (so we know, where the error occured) fixed the alpha-channel fixed documentation removed unused variables removed the [pix_buf] entirely from the example-patches. i must have been somewhere else when i added them added setPixModified() where apropriate so that the pull-mechanism works moved utility functions (namely: powerOfTwo()) into Base/GemFuncUtil added some comments structured the location of functions a bit: GemPixObj does not hold any utility functions GemFuncUtil holds general untilities (like powerOfTwo()) GemPixUtil holds utility functions for image manipulation (only!) GemPixPete (NEW) holds functions from pete's_plugins (specialzied PixUtil-functions) moved utility functions (namely: powerOfTwo()) into Base/GemFuncUtil moved utility functions (namely: powerOfTwo()) into Base/GemFuncUtil added the [pix_texture2]-alias again. why has it vanished ? fixed the default viewpoint-values to the correct ones "0 0 4" instead of "0 0 -4" moved some fsaa-code from apple-only to openGL-dependent code (read: changed ifdefs) removed some more verbosity... defined SHIFT_Y1,... for YUV-shifting used SHIFT_Y1,... (from GemPixPete.h) as my images were green ;-) added bugs concerning win32-movie-loading seems to work now with RGBA/YUV/Grayscale initialized orgPixBlock to NULL (should fix crashes when no upstream image is present) quick hack to make Grey/YUV work use normal memory-allocation (no alignment tricks as these produces weird output) LATER make sure that memory is aligned correctly added run-time checks for GL_APPLE_ycbcr_422 (YUV-texturing) use run-time checks for YUV-texturing made YUV work correctly the RGBA2LUMA conversion was wrong another step towards a working version. now yuv/rgba/grey almost(!) behave the same setting the csize/format/type of the image more intelligently made YUV work on PCs (and used better namings (like "U" instead of "Red") using setCsizeByFormat rather than setting csize/format/type directly don't initialize to small images (64x64) any more as this is only for memory fragmentation removed the commented out the processGray-function (and not only it's body); use setCsizeByFormat use setCsizeByFormat use reallocate() instead of clear() and allocate() use setCsizeByFormat use reallocate() instead of clear() and allocate() removed unused variable replaced hardcoded shift-amount by SHIFT_Y. use setCsizeByFormat use reallocate() instead of clear() and allocate() fixed stupid typo use setCsizeByFormat use reallocate() instead of clear() and allocate() and complicated checks made YUV and Grey work made Grey work type fixing to shut down the compiler shut up macOS-gcc compiler warnings using ch*-constants rather than hardcoded positions of channel-data (e.g: chRed instead of 0) added U8 as another type made defines for shifts on mac/non-mac (although they don't do much now) added additional error-checks and a sanity-check on close(); should run now on QT-6.5 (as (un)stable as with QT-6.4) don't copy "pdata" with copyImage2Struct; pdata should always(!) be a pointer to some data that is owned by the image itself! this fixes the abiding crashes when deleting pix_-objects fixed YUV on big-endians added range-checking for the offset when doing smoothed & filled processing; this should fix the crashes at very high offsets on all colorspaces added a "yuv"-message to turn off direct YUV-texturing even if it is available (because our G4 with a geforce4MX says that it can do GL_ycbcr but actually the images look weird noted that pix_lumaoffset might crash with high offset-factors fixed some issues with compiling Gem on windows with nvidia's openGL-headers (have to include glext.h) changed the shifts to SHIFT_* to make it cross-platform changed from 0.888pre2 to 0.90 fixed "installation" for on windows changed version from 0.888 to 0.90 now the "release" compiles updated for 0.90 "not using client-texture" is now a debug-message (instead of post()) fixed a bug that did not texture rectangle images correctly texture mode can now be set via a "texture"-mess (spherical or not) texturing of line/point enabled (buggy output ?) the Release-binary under windoze used to crash when a pix_video was created; defining "DEBUG" on the correct place gets rid of this. (indeed an ugly workaround) reorganized a bit... ----------------------------------------------------------------------- hooks/post-receive -- Gem - Graphics environment for multimedia _______________________________________________ pd-gem-CVS mailing list pd-gem-CVS@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/pd-gem-cvs