Home | Back

GLUT Change Log
================

NOTE:  This document is potentially incomplete.

GLUT 3.1 -> 3.2
----------------

  o  Complete Unix-style nroff man pages for all the GLUT library
     routines.  See README.man for details.

  o  Implement the MESA_SWAP_HACK.  If you set the MESA_SWAP_HACK
     enviornment variable and you are using Mesa, GLUT will attempt
     to simply do a glXSwapBuffers to repair damage to a double
     buffered GLUT window that appears to have been last displayed
     using a glutSwapBuffers call.  This can *greatly* improve the
     redraw performance of GLUT programs running under Mesa when
     they are damaged (for example by popup menus that are not in
     the overlays).  See lib/glut/glut_mesa.c  This idea was suggested
     by Brian Paul on the Mesa mailing list.

     Note that some poorly behaved (probably buggy, actually) GLUT
     programs may no redraw correctly with MESA_SWAP_HACK.  The
     optimization relies upon your display callback being idempotent
     and that the window is not otherwise rendered to by other
     callback routines.

  o  Port Brian Paul's various Mesa 2.0 demo programs to use
     GLUT: bounce, gamma, gears, isosurf, offset, reflect,
     spin, tess_demo, texobj, & winpos.  The reflect program
     is definitely worth checking out.

  o  Include GLUT examples presented by Tom McReynolds and
     David Blythe at the "Advanced OpenGL Rendering" course at
     SIGGRAPH '96.

  o  Add progs/examples/simple.c showing how to simply draw
     a single triangle with OpenGL.  How long would an equivalent
     Direct3D immediate program be?  Much longer...

  o  Screen door transparency example using polygon stippling
     based on code from Tim Hall.  progs/examples/screendoor.c

  o  Fix bug where glutRemoveMenuItem about recalculating
     the menu's width in pixel that would sometimes cause an
     X protocol error.

  o  Make sure that when a popup menu is finished that if the
     initiating window has an overlay, the overlay colormap
     is installed.  An overlay popup menu could otherwise knock
     out the window's overlay colormap.

  o  Fix an "off by one" error in the window number printed out
     in a fatal error message when a redisplay is generated on
     a window without a display callback.

  o  Make consistent puncutation in some fatal error messages.

  o  Some work to make example programs easier to build on
     Windows NT and 95 (probably not complete; report bugs
     please).

GLUT 3.0 -> 3.1
----------------

  o  The various shape rendering routines use single
     precision instead of double precision.

  o  The box shape rendering routines don't generate
     redundant normals.

  o  Make glutExtensionSupported more robust.

  o  Added test20 for glutExtensionSupported.

  o  Fixed Makefile in /usr/share/src/GLUT images.

  o  Fixed multisample complication bug in dinospin.c

  o  Fixed multiple includes in moth.c

  o  Add contribed steam.c and glutmech demos (cool!).

  o  Added fontdemo and evaltest in progs/examples.

  o  Fix an HP/UX compilation problem with libglut.a

  o  Fix GETTIMEOFDAY macro to take two args on a Sun.  Why does
     Sun claim to be SVR4 when they have a BSD (2 parameter
     gettimeoday)?

  o  Add an #ifdef to glut_win.c for working around Solaris
     2.4 and 2.5 bug in XmuLookupStandardColormap.

GLUT 2.3 -> 3.0
----------------

  o  Major version change so GLUT 3.0 does have API additions!

  o  Added yacme (yet another colormap editor) demo by Patrick Bouchaud
     (SGI Switzerland).

  o  Updated glut.h so that GLUT_XLIB_IMPLEMENTATION=5 for 3.0
     and GLUT_API_VERSION=3 for 3.0.

  o  Change all the sample programs to _not_ #include <GL/gl.h> or
     <GL/glu.h>.  From the spec:  "Because a very large window system
     software vendor (who will remain nameless) has an apparent
     inability to appreciate that OpenGL's API is independent of their
     window system API, portable ANSI C GLUT programs should not
     directly include <GL/gl.h> or <GL/glu.h>.  Instead, ANSI C GLUT
     programs should rely on <GL/glut.h> to include the necessary
     OpenGL and GLU related header files."  Sigh.

  o  New Helvetica fonts: 10, 12, and 18 point versions.
     GLUT_BITMAP_HELVETICA_10, GLUT_BITMAP_HELVETICA_12, and
     GLUT_BITMAP_HELVETICA_18.

  o  Implement glutFullScreen.

  o  Implement glutSetCursor.

  o  Implement glutGetModifiers.

  o  Implement glutBitmapWidth and glutStrokeWidth.

  o  Reimplement test2 to use glutGet(GLUT_ELAPSED_TIME) instead of
     the less than portable gettimeofday.

  o  Ripped out all the support for compiling earlier GLUT API
     versions.  It got very ugly and hard to maintain.

  o  Added GLUT_LUMINANCE to GLUT verison 3 API and documented it, but
     it is not implemented as part of GLUT 3.0; should be in 3.1.

  o  Added OpenVMS support to GLUT library with much help from Andy Vesper.

  o  The display mode mask parameter for glutInitDisplayMode and the
     milliseconds parameter for glutTimerFunc are now of type unsigned
     int (previously unsigned long).

  o  glutMenuStatus obsoletes glutMenuState.  The callback registered
     by glutMenuStatus is triggered just like glutMenuState, except
     that an additional two parameters pass the X and Y location of the
     mouse when the menu status changes.  This is to remedy problems a
     GLUT program using glutPassiveMotion might have re-sync'ing with
     the mouse position after menu use.

  o  In previous releases of GLUT, various user input events were
     discarded when menus were in use.  Now, these callbacks are
     generated.  If the effect of the events should be performed after
     the menu is released, it is up to the GLUT program to delay the
     action.

  o  Careful to ignore Enter/LeaveNotify events that come
     from the pop-up window pointer grab and ungrab (ie,
     Enter/LeaveNotify events not marked NotifyNormal for their mode.
     Eliminates "spurious" entry events.

  o  Fixed bug with new work (like subwindow creation is a display callback)
     not being placed on the work queue and getting lost.

  o  Fix memory corruption bug in caching of values returned by
     glXChooseVisual; no longer cache.

  o  Fixed bug with glutCopyColormap sometimes not accurately
     re-establishing the WM_COLORMAP_WINDOWS property.

  o  Fixed bug in deliver of enter and left callbacks.  Before "virtual"
     enter and leave (X protocol teminology) were being reported.  The
     semantic for the entry callback is it is entering and leaving
     a GLUT windows real estate that counts; not passing through
     the window hiearchy.

  o  Add XFlush to __glutFinishMenu in case menu call back will want to
     read from a terminal window.  In this case, the X server may still
     be grabbed since the XUngrabPointer hasn't been flushed out of
     Xlib's buffers to the X server for processing (deadlock!).

  o  Don't allow modification of menus while in use.  Spec ammended to
     say "It is illegal to create or destroy menus or change, add, or
     remove menu items while a menu (and any cascaded sub-menus) are in
     use (i.e., popped up)." Allowing modifications of menus while in
     use was dubious, the semantics would be unclear, and even more
     unfortunate, might vary from system to system.

  o  Spec warns that you should not call routines that require a current
     window or menu when there is not a current window or menu (like
     before you create one or if you destroy the current one).

  o  Fixed bug where glutSpecialFunc callback would only operate if
     a glutKeyboardFunc callback was also registered.

  o  Fixed bug in color index visual selection so that largest number
     of bits of color resolution is preferred to smallest.

  o  Fixed bug where if a double buffer X window was used to emulate
     a single buffer window, GLUT should set the OpenGL read mode
     to GL_FRONT (it was only changing the draw mode).

  o  Fixed an obscure ordering bug with simultaneous glutPushWindow and
     glutPopWindow calls on different subwindows of the same window.

  o  Now, it is a fatal error to receive an expose event with no
     display callback registered (it was stupid to allow mapped windows
     that made no attempt to redisplay).  The spec has been changed to
     say NULL can not be passed to glutDisplayFunc; previously, it
     said this would disable the callback.  Stupid.  What was I thinking?

  o  GLUT spec is indexed now and generally improved.

  o  Support for IRIX 6.2's N32 and N64 object style to take advantage
     of new MIPS processor instruction sets.  See lib/glut.n32 and
     lib/glut.n64 (IRIX 6.x only).

  o  Fortran bindings also available for N32 and N64.  See lib/fglut.n32
     and lib/fglut.n64.

  o  Tests support the new API version 3 interfaces.

  o  Contributed "bigtest" test program.

  o  Added splatlogo example demonstrating pixel path functionality.

  o  Added zoomdino program demonstrating use overlay rubber-banding.

  o  Contributed demo programs: hanoi, hanoi2, gears, moth, text3d,
     and noof.

  o  Contributed SGI stereo program: stereo-plane

  o  Being out of colors in the default colormap without overlay planes
     will no longer lead to a fatal error when allocating the color gray
     for a pop-up menu.  Instead, GLUT finds the closest matching color
     automatically.  Help from Brian Paul to code this.

  o  Eliminated Glut64.mk from all Makefile.sgi's

GLUT 2.2 -> 2.3
----------------

  o  Fix obscure problem with how X Input extension button press and
     release events get selected; it seems GLUT should also be
     selecting for the DeviceButtonPressGrab event class since this
     ensures that a release will be delivered to the same window as the
     press; that was my intent for how buttons presses and releases
     should work.  I just didn't know you needed this obscure event
     class to be selected.  Such a grab is the default for pointer
     events.

  o  If windows are indirect, make sure that a glFinish is properly
     done for all windows that have been made current to recently (ie,
     last main loop iteration).  This helps ensure windows don't get
     ahead of themselves.  Previously, GLUT simply did a glFinish to
     whatever window GLUT happened to be made current to last at the
     end of each main loop iteration.

  o  Fix what can be considered a bug where calling glutPostRedisplay
     within a display callback would not leave the redisplay set; this
     applies other routines that set work state needing to be done.
     Essentially, clear the workMask before the work callbacks are done
     instead of after.  The new test15 tests this.

  o  To support other OpenGL higher-level library that might want to
     be able to call glXMakeCurrent, GLUT should not try to track
     make current state and short circuit glXMakeCurrent calls when
     GLUT believes the window/context are already current.  glXMakeCurrent
     should implement the same "short circuit" case, but can appropriately
     track the changed state from multiple libraries.

  o  The glutRemoveMenuItem routine was totally busted previous to 2.3.
     Now it is fixed and test14 is a regression test for it.

  o  Add Greg Humphreys's hanoi program to progs/contrib

  o  Examples demonstrating Open Inventor used with GLUT found in
     progs/inventor

  o  New progs/examples programs:  abgr, triselect, texenv, stenciltst,
     stars, fogtst

GLUT 2.1 -> GLUT 2.2
---------------------

  o  Fixed bug in glutGet(GLUT_WINDOW_X) and glutGet(GLUT_WINDOW_Y).

  o  Added Philip Winston's human kinetics demo in progs/demos/walker
     (COOL!).

  o  Improved test suite.

  o  Work around SunOS 4.x's non-ANSI compliant realloc.

  o  Better handle delivery of reshape callback.

  o  Pre-build stroke font C files to avoid having to rely on lex/yacc
     working correctly.  Too many Linux systems missing flex/bison or
     having verisons that resulted in undefined symbols while building
     strokegen.

  o  Better portability, mostly avoiding compiler warnings.

  o  64-bit clean (as far as I know!).

  o  Compiles right for systems like DEC Alpha that support GLX 1.1 but
     do not support GLX_SGIS_multisample extension.

  o  At Brian Paul's request for better Mesa support, changed colormap
     allocation to work if RGBA window has PsueoColor visual.

  o  Fix Drew Bliss reported bug in colormap allocation (resulted in
     GLUT programs not correctly sharing colormaps).

  o  Interesting bug involving the fact that XFlush may actually
     read X protocol into Xlib's internal buffers so if you call
     XFlush, then select on the X socket, you might find nothing to
     read, but when in fact there is data in the Xlib internal buffers
     that should be processed.  Using ISDN (I guess it has weird
     network properties) showed this bug in a GLUT test13.  The fix is
     to use XPending after XFlush.

  o  Improve X Input documentation references in the GLUT spec.

  o  Widen margins so spec kills less trees.

  o  Clarify GLUT spec to explain glutGet values returned by
     GLUT_WINDOW_X, GLUT_WINDOW_Y, GLUT_WINDOW_WIDTH, and
     GLUT_WINDOW_HEIGHT as being returned in pixels.

  o  Make sure consistently using reshape (and not resize) within GLUT
     spec.

  o  GLUT spec should document that glutExtensionSupported,
     glutCreateWindow, and glutCreateSubwindow return int.

GLUT 2.0 -> GLUT 2.1
---------------------

  o  XCOMM comments in Imakefiles results in unusable Makefiles
     for systems that pre-date an X11R5-based imake.

  o  The rings.c and worms.c examples use "float" returning
     versions of trigonometric functions (ie, cosf, sinf, etc.).  While
     ANSI C reserves these names for their logical purpose, many
     systems do not implement these functions.

  o  Calls to __glutWarning in glut_cindex in places lack
     variable %d argument.

  o  Intended comments in lib/Imakefile cause avoiable
     warnings for some C pre-processors.

  o  Old (pre-X11R6) versions of Xlib did not have XK_Page_Up
     and XK_Page_Down keysym defs in /usr/include/X11/keysymdef.h
     These are aliases to XK_Prior and XK_Next.

  o  The X11R5 to_wfont derived strokegen program for converting
     PEX fonts to C data structures was full of memory access
     bugs.  On most systems, they were not fatal.  That's
     what I get for trying to steal PEX code I gues. :-)

  o  Various Purify-detected memory access bugs within the
     GLUT library have been fixed.

  o  Better Imakefile handling of lex usage in lib/glut/Imakefile

  o  Better document colormap management routines.

  o  Correctly document the prototype for glutCreateSubWindow

  o  Add documentation usage hint explaining glutGetWindow
     and glutGetMenu can be used to determine what window
     or menu a callback is for.

  o  Fix bug in glutDetachMenu

  o  Fix bug in delivery of visibility callback for subwindows
     of an unmapped top-level window.

  o  scube.c should use more likely to be optimized
     glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA) instead of
     glBlendFunc(GL_ONE_MINUS_SRC_ALPHA, GL_SRC_ALPHA)

  o  Supression of various compiler detected warnings through
     the GLUT distribution.

  o  Use of the "polygon offset" extension if available in origami.c

  o  SGI mkmkfiles.sgi script doesn't make Fortran by default now;
     see instructions in README.fortran

  o  Improvements to the test programs.

  o  Avoid X Input extension calls when extension is present but
     no devices are available.

GLUT 1.0 -> GLUT 2.0
---------------------

  o  API extensions; see README.glut2

  o  Bug fixes.

  o  More demos.

  o  Update documentation.

- Mark