docs/README.WIN32

    Mesa/Readme.win32

    Last Updated: Sun, Dec 06 1998, 08:49 EST - tjump@spgs.com

    Note: While this build set supports generation of a 3Dfx specific
          DLL using Mesa, David Bucciarelli's original build files
          are the "supported" method. -Ted

*** What's New

- Build environment change: The Glide SDK is no longer assumed to be in
  the global INCLUDE/LIB environment vars, it is required that you set the
  value 'GLIDE2X' as either an environment variable pointing to your Glide
  SDK install directory or that you configure that as a build option to
  nmake.exe when building fxmesagl32.  Examples:

    nmake /f nmake.mak GLIDE2X=g:\sdk\glide2x fxmesagl32

          <or>

    nmake /f nmake.mak GLIDE2X=g:\sdk\glide2x allfx

          <or>

    nmake /f nmake.mak GLIDE2X=g:\sdk\glide2x progs.3dfx.demos

  The DevStudio workspace files for 3Dfx OpenGL require the definition of
  GLIDE2SDK as an environment variable pointing to where your copy of the
  Glide SDK has been installed. Adding this to your AUTOEXEC.BAT would do
  so (change the directories to match):

       SET GLIDE2SDK=G:\SDK\GLIDE2X

- Static build/link of all non-3Dfx targets works now, cool. Required some
  nontrivial chicanery in the GLUT headers though to allow it to use the
  mesa 'wgl' functions statically linked instead of always importing the
  sytem wgl functions from GDI32. Mostly this was an extension of an
  existing method used in the headers, just taking into account more
  functions.

- Handled an undocumented change in how Microsoft NMAKE for DevStudio 6
  defines _NMAKE_VER so that the makefiles should work properly.

- Microsoft Compiler/Win32 specific build tuneups in the header files
  providing for dramatically faster compilation times of mesa itself and of
  code using Mesa. Primarily via the removal of any call sequence which
  would invoke WINDOWS.H (which triggers ~6-10K lines of code inclusion).

  Supporting this, a macro name change was made in the gl.h/glu.h/glut.h
  header files which should be client-code transparent. The change was to
  specify function prototypes with the (newly created) macro GLAPI instead
  of WINGDIAPI to ease conflicts with multiple definitions of WINGDIAPI
  (needs to be one thing to access Win32 API calls, another to define
  export funcs when building a Mesa/GLU/GLUT DLL which themselves need to
  import functions from Win32, etc.).

  Similarly, call sequence definition usage of APIENTRY, CALLBACK, and
  WINAPI were renamed to prevent collision/need for windows.h.

  Lastly, these changes were incorporated into the GLUT source to
  facilitate same compile-time performance increases.

  All of these changes should be transparent to code USING
  mesa/mesaglu/glut, and in most cases did not require changes in their
  source files within the Mesa/GLUT code base (but there were a few).

- 'wgl' function prototypes and supporting data types declared in gl/gl.h
  to provide for the usage of these from client Windows code without
  requiring WINDOWS.H to be included. If you're using non-trivial functions
  (such as any that pass structures around) you will need to include
  WINDOWS.H *before* gl/gl.h, however in that case you needed them anyway.

  One benefit of this is that glut-based code on Windows may now use
  wglGetProcAddress to get the function pointers for all supported OpenGL
  extensions, preventing the need for directly importing them into
  code. The 3Dfx/Demos/glbpaltx.c has been modified to use this mechanism
  by way of example.

  This code has also been copied into gl/glut.h to support glut programs on
  Windows that do not use Mesa.

- new command line build target: all.debug

  Builds all DLL files and test programs in release, then debug builds,
  emplacing the requried DLL files in the EXE directories for ready test
  runs.

- Expanded MESA_WGL_FX options. When setting it to instruct fxMesa to
  render windowed you may provide it a clue to your system's display, or
  instruct it to try to detect the pixel format. This information is used
  in constructing the windowed-rendering hack to try for optimal
  performance, such as it is. You specify this by adding one of the
  following tags in the envvar prefixed by a colon.

    noconv    - does not do any pixel format conversion
                This is synonymous to '16bgr'.
    16bgr     - uses a 16-bit RGB sequenced dib section and does no pixel
                format conversion as this is the internal format of the
                3Dfx hardware when used by fxMesa.
                This is synonymous to 'noconv'.
    16rgb     - uses a 16-bit RGB sequenced dib section and does pixel
                format conversion each frame from the 3Dfx internal format
                to 5:6:5 RGB.
    15rgb     - uses a 16-bit RGB sequenced dib section and does pixel
                format conversion each frame from the 3Dfx internal format
                to 5:5:5 RGB.
    15bgr     - uses a 16-bit RGB sequenced dib section and does pixel
                format conversion each frame from the 3Dfx internal format
                to 5:5:5 RGB.
    24rgb     - uses a 24-bit RGB sequenced dib section and does pixel
                format conversion each frame from the 3Dfx internal format
                to 8:8:8 RGB. Also suitable for a 32-bpp display.
    24bgr     - uses a 24-bit BGR sequenced dib section and does pixel
                format conversion each frame from the 3Dfx internal format
                to 8:8:8 RGB. Also suitable for a 32-bpp display.
    query     - Queries Windows for the displays pixel format and attempts
                to adjust accordingly. This query is done by using
                DirectDraw to allocate a "Primary" surface and then
                checking what that surface's pixel format is (as the Win32
                API for getting the pixel format from a window device or
                screen device context is useless as far as I have been able
                to determine).  This does *NOT* mean that fxMesa is now
                requiring DDraw, however it will use it if found. This
                prevents system incompatabilities with WinNT 4SP2 and
                earlier systems.

    Example, put this in your AUTOEXEC.BAT to have fxMesa detect a best
    case match on context creation:

         set MESA_WGL_FX=windowed:query

    Additionally, if you wish to find out *what* it detected, you may add a
    ":logged" to this which will cause it to create/append to a file named
    FXMESAGL32.LOG whenever a context is created, e.g.:

         set MESA_WGL_FX=windowed:query:logged

    *** WARNING: Some display drivers/windows version combinations do not
        support all of the possible DIB section formats that can be
        requested, and thanks to Microsoft there is no way of detecting
        (that I have yet discovered) that the requested DIB section will
        not work and when it is utilized the application crashes.

  Eventually I hope to incorporate support for DirectDraw Overlay surfaces
  to facilitate less data crunch overhead when dealing with the in-window
  hack. For example: if a primary display support overlays of BGR565
  sequence then the data could be read from the 3Dfx framebuffer into a
  DDraw "System Memory" or "Non-Local Video Memory" surface - both of which
  actually reside in on-board DRAM (the latter in AGP space) which could
  then be blitted/flipped in a much faster fashion than the current DIB
  section handling.

  The DIB section handling will always be the default, as that's much more
  compatible since it utilizes GDI features sets that have more longevity
  to them.

*** Legalese

These build files are provided as-is and are submitted to be included with
the "Mesa 3-D Graphics Library" package as (currently) maintained by Brian
Paul. These project build files are free software; you can redistribute it
and/or modify it under the terms of the GNU Library General Public License
as published by the Free Software Foundation; either version 2 of the
License, or (at your option) any later version.

These project files are distributed in the hope that they will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Library
General Public License for more details.

You should have received a copy of the GNU Library General Public License
along with this library; if not, write to the Free Software Foundation,
Inc., 675 Mass Ave, Cambridge, MA 02139, USA.

*** Maintenance Responsiblity and Technical Support

While these files are now part of the Mesa core distribution please do NOT
contact Mr. Paul for help with them if you encounter problems as he can't
help you (currently).  I will, however, attempt my straightforward best in
assisting anyone with using these files on their system.  I can NOT
guarantee instant responses owing to other responsiblities, but I do try
dang hard to answer any mail w/in 24 hours.  I may be contacted at the
above email address for the forseeable future.

-Ted
mailto://tjump@spgs.com
http://www.i21.com/~tjump

*** General Information

These build files facilitate convenient building of many variants of Mesa,
both as static link libraries (including mesaglu) and as dynamic link
libraries that in some cases may be used as "drop-in" replacements for
OpenGL32.DLL on both Windows95 and Windows NT.

The construction of the Win32 command-line build files and projects has
been something of a pet project of mine, and is based upon my own
"standard" Win32 build environment as supplied by the "nmake.mif" file.
They have been tested under Windows95 OSR2, Windows NT 4.0SP3, and Windows
NT 5.0 beta 1.  The libraries that they generated have been tested (via the
demo programs) in a *limited* fashion on the above three systems, including
the 3Dfx versions.

The reason I went with command-line build environment instead of the more
convenient IDE-based project files is for two reasons: 1. These appear to
have some amount of portability between versions (the nmake syntax hasn't
changed much since Microsoft C 7.0) while the IDE project files seem to
change drastically each version. and 2. These are readable with any ascii
editor and such are better self-documentation of the file relationships for
more people such that it will facilitate supporting other Win32 compilers.

While these files only deal with building for x86 targeted code it *should*
be possible to add the necessary logic to them to build for the other MSVC
supported CPU targets, I simply have no hardware to test them on nor the
alternative compilers to build with.

*** Prerequisites for use

1. You must have a 32-bit Microsoft compiler installed. I have tested
this with Visual C 5.0 (SP3) and Visual C 4.2, but with minor
(possibly no) modification to the nmake.mak and nmake.mif files this
sequence should work on Visual C 2.0 also. The workspace files
(mesalib.dsw and mesademos-*.dsw) and their included project files
(*.dsp) are specific to the DevStudio IDE - I have made no attempt at
building a VC4 IDE project set as I do not use that any more.  Note
that the VC workspace files NO LONGER use NORE are dependant upon the
nmake.mak and nmake.mif files for construction of definition (*.DEF)
and resource (*.RC) files.

*** Visual C 4.x Users Warning ****

Note that early editions of VC4 do NOT have header files current enough
for use building this code base. If you are using VC4 you will either need
to get an update to version 4.2 *or* you may download the Platform SDK
directly from Microsoft's web site (www.microsoft.com) and update your
build environment that way.

*** Visual C 4.x Users Warning ****

2. You must have the PATH, INCLUDE, and LIB environment variables set
properly. With VC5 you can easily get this by executing the VCVARS32.BAT
file that was created for you upon installation. It is found in the
DevStudio\VC\BIN directory, wherever you installed DevStudio. VC4 provides
a similar batch file in it's BIN directory also.

3. (optional) If you're going to build for 3Dfx/Voodoo you will need to
have previously installed the Glide SDK version 2.3 or later, if I
recall. This may be retrieved from www.3dfx.com for no money and some
download time. ;-) These build files assume that you have the Glide SDK
added to the respective environment variables (LIB and INCLUDE).

4. (optional) If you're going to build for S3/Virge you will need the S3
Developers Toolkit which may be downloaded from www.s3.com for the price of
registering on-line and some time. NOTE: I can build the s3mesa.dll file to
completion, however the compilation of s3mesa.c currently generates a large
amount of compiler warnings and between that and the fact that I can not at
all test it I can make no claims to it's ability to execute.  Again, like
the 3Dfx version before this, these build files assume you have the S3Dtk H
and LIB files in the path of their respective environment variables.
Note 2: As of Mesa3.0beta6 I have build files, both command-line and IDE,
which should be able to build the s3mesa code base if it weren't for updates
being required in the S3 DD code support (Mesa-3.0/src/s3 directory).

I advise putting any include and lib files for secondary toolkits (Glide,
S3Tk, whatever) in their respective environment variables *before* the
Microsoft-assigned default values.

*** FAQ: Frequenty Asked Questions and Other Important Information ***

- When running the 3Dfx demos under Windows NT, they crash on exit, what's
  up?

  This is apparently a problem in Glide itself. The workaround is to go to
  your C:\WINNT\SYSTEM32 directory and rename the file FXOEM2X.DLL to
  FXOEM2X.DL_ to prevent Glide from loading and initializing it upon
  startup.  This is known to be an issue with cards that do not have "TV
  out" and is known to cause crashes on Diamond Monster II 8M and 3Dfx
  Reference boards, all using 3Dfx Reference Drivers version 2.53. Other
  hardware/driver combinations will also likely exhibit this behavior.

- I'm having a problem building Mesa for static library linking.

  This was caused by some incomplete testing on my part, and a fix is now
  available in the form of an add-on to the base Mesa 3.0 release.  The
  file to get is:

       via FTP download from: iris.ssec.wisc.edu
         you want to go here: /pub/Mesa/patches_to_3.0/
        you want to get file: Mesa-3.0-w32-static-fixes.tar.gz

  This required a minor addition to INCLUDE/GL for a clean solution, the
  file "include/gl/mesa_wgl.h" is automatically included by
  "include/gl/gl.h" when a Win32 non-DLL build is in progress to provide
  prototypes for the various wgl functions.

  The only remaining hitch in this setup is that the 3Dfx build is not yet
  running as a static build, because of problems with conflicts in
  existance of the various GDI functions like ChoosePixelFormat,
  etc. *sigh*

  Anyway, the "allstatic" target now works as expected and builds all
  book/sample/demos programs to boot. ;^)

- How do I get fxMesa to render in a window on the desktop instead of only
  full-screen?

  Use the Microsoft Windows fxMesa-in-a-window hack!

  Seriously, if you want fxMesaGL to render using the 3Dfx Voodoo1 or
  Voodoo2 hardware into a window on the desktop then all you need to do is
  set the MESA_WGL_FX environment variable to anything other than
  "fullscreen" and it will render into a window.  If you wish to go
  fullscreen then you only need to NOT have the environment variable, or
  have it set to "fullscreen".  You may also switch at runtime between
  fullscreen-mode and windowed by pressing ALT-ENTER on the keyboard
  (unless the application using Mesa does something with those keystrokes,
  of course).

  As of 8/13/98 this should be running a LOT better for more people as a
  low-compatability item was cleaned up which prevented it from working on
  many (most?) display drivers under Windows 9x.

- I have my 3Dfx card hooked to it's own monitor and I want the output to
  stay on even if I switch to another program, is this possible?

  If the Glide environment variable SST_DUALHEAD is set to '1' then fxMesa
  will never disable the Voodoo output on a Voodoo1 or Voodoo2 display
  regardless of whether the fxMesa application is "current" or not. This
  works regardless of whether it's rendering using the window hack
  mentioned above or not.

- I want to run the Mesa demos on my Intel740 card using it's own OpenGL
  acceleration, how do I do this?

  Build GLUT standalone for use with system OpenGL and GLU drivers!

  The Command-line project supports building all test/demo programs against
  these drivers also! This allows you full use of GLUT on Windows using
  hardware accelerated OpenGL. Wheee! This includes the "3dfx/demos"
  directory of which only two programs will not run on "standard"
  opengl. Note that there are a few of the sample programs which will NOT
  work without Mesa as they directly call into Mesa instead of using the
  extension mechanism.

*** Included programs that exhibit unfortunate or bad behavior

- demos/bounce - doesn't run on high-colors screens?  It's requesting an
  INDEX display from GLUT and that fails on my true-color desktop. Changing
  this to _RGB let's the program work, but it doesn't display
  properly. This is probably just an idiosyncracy of my machine though, as
  if I test the program using GLUT for System OpenGL on my Intel740 OpenGL
  accelerated machine it's just hunky-dory.

- demos/glutfx - runs, but crashes on exit (but not on my Intel740 machine)

- demos/texobj - runs, but crashes on exit if ESC is pressed. Exits cleanly
  if the Close box on the window frame is pressed with the mouse. Go figure.

- book/aaindex - doesn't run, can't get pixel format, because it wants an
  INDEX display maybe (but is okay on my Intel740 machine)?

- most of the book/* demos don't respond to ESC being pressed.

- 3dfx/demos/* - all demos run, however they all crash on exit. I've traced
  this so far as to determine the call it's happening with. The crash comes
  from within Glide during the processing of the grGlideShutdown() call, as
  in invalid memory reference exception. I'm wondering if this is because
  of some state or processing not being completed before the call. Dunno,
  but putting grSstIdle() in just before grGlideShutdown() does NOT fix the
  problem.

- 3dfx/demos/tunnel2 - does not run on my system even with SLI mode
  disabled. Hmmmm, maybe I need to disconnect my Voodoo2 cards?

*** Important Notes and Changing Default values

- The optimizer settings have been manually reworked in both command line
  and DevStudio IDE files to hopefully prevent possible irrational code on
  the part of the code generator.  Formerly, it was configured for "/Ox",
  now it is configured for safer handling at a slight potential performance
  cost. This may not be required for Visual Studio 6 but I can't test that
  (yet).

- These files build with the code targeted for Pentium processors and
  8-byte structure padding.

- The IDE-built programs seem to be "happier" in that the command line
  build of the 3Dfx demo "fire" will grenade on exit (?). Otherwise pretty
  much everything may be built with either interface.

- The currently configured Mesa version is 3.1, and MesaDemos version is
  the same. To change this permanently you will need to edit NMAKE.MAK and
  change the lines that look like this (they start o/a line 116):

    # Currently, Mesa is at rev 3.1 ...
    #
    !IF "$(MESAVER)" == ""
    MESAVER=3.1
    !ENDIF

    # used in building all of the resource files for the Mesa DLLs
    #
    !IF "$(MESAFILEVER)" == ""
    MESAFILEVER=3,1,0,0
    !ENDIF

- Currently the build files are configured to be used from a Win32
  directory that is included inside the main Mesa-3.1 heirarchy.

- The build files are smart enough to find the files for the core lib, glu,
  glut, and the various demo programs if they are unpacked in the current
  Mesa-3.1 heirarchy, like this:

    \Mesa-3.1
    \Mesa-3.1\src
    \Mesa-3.1\src-glu
    \Mesa-3.1\src-glut
    \Mesa-3.1\Win32
    \Mesa-3.1\samples
    \Mesa-3.1\demos
    \Mesa-3.1\book
    \Mesa-3.1\3Dfx\demos

    ... should work.  This arose because my initial build tests for the
    demo files were done before MesaDemos 2.6 had been released.

- To enable use of MMX instructions by the VC5 compiler you may add the
  "USE_MMX=1" option to the NMAKE command line, or edit the default in the
  NMAKE.MAK file.  This does appear to have some affect on the performance
  on the library and does not seem to harm it in any way *but* I have done
  *no* verification of this. I have an MMX processor so I figured what the
  heck. This option is only available with VC5 when building from the
  command line.

  This is probably required if you are going to modify the code to include
  inline MMX instructions though.

- With the exception of the static link libraries generated by this file
  set (mesagl.lib, mesaglu.lib, mesaglut.lib) all DLLs and executables are
  built against the "Multithreaded DLL" runtime - this means that they
  require MSVCRT.DLL or MSVCRTD.DLL in the path to execute.

  ** CHANGED 8/11/98 ***

  Note also that the demos are all built aginst the "OpenGL32, GLU32, and
  GLUT32" and as such they are fairly agnostic wrt: building against Mesa
  for CPU-rendering, Mesa-for-3Dfx, Mesa-for-S3, or System OpenGL.

  If you want to build them for use on your system and your display card
  provides full OpenGL acceleration (Permedia, Intel740, Intergraph,
  whatever) then you only need to build GLUT prior to building any of the
  demo programs. For convenience, the GLUT project is included in each of
  the demo projects Workspace files for the DevStudio IDE builds BUT it is
  not automatically built - you still need to build it first manually.

  Note that if you have GLUT already installed on your system (gl/glut.h in
  yoru INCLUDE path, glut32.lib/glut32d.lib in your LIB path, and the DLL
  in your PATH) then you do NOT need to build GLUT prior to the test
  programs.

- The 3Dfx build of Mesa has primarily been tested with Quake 2 and it runs
  (mostly) fine on my PC (take that for what you want it)...

  ** CHANGED  8/11/98 ***

  There is still something going on that causes Glide to crash on shutdown,
  when I run fxMesa under Windows NT, however it does not appear to occur
  under Windows 9x on either Voodoo1 or Voodoo2 cards. *sigh*

- I can not test the S3 build as I have no machines available with Virge
  based display cards.

- The multithreaded test code is *not* built as it requires pthreads and I
  have as of yet spent not time trying to get that running. The latest word
  that I saw WRT threading support on win32 was that they are intending to
  support it natively within Win32 - so I'm waiting it out until they get
  it done.

- Similarly, the 'xdemos' are not currently built because I haven't gotten
  around to building the client libs for native win32 and getting it all
  setup for use.

*** Output Files

All final output files (DLL/LIB) are placed in the Mesa-3.1/lib directory,
with the exception of the fxMesaGL32 build which is placed in
Mesa-3./lib/FX and the executable images which are placed in their source
directories.

To be able to execute the various test programs, you will need to copy the
requisite DLL files into the same directory as the EXE files. Note that
most of the 3Dfx/demos/* programs WILL run with the non-FX build of Mesa -
just very slowly. The two programs which are hard-linked with the FX build
and will not run without it are "glbpaltx" which uses "gl3DfxSetPaletteEXT"
directly instead of via the extensions mechanism and "tunnel2" which uses
"fxMesaSelectCurrentBoard" API for selecting between multiple 3Dfx cards
installed in one system. Likewise, "paltex" directly uses the
"glColorTableEXT" extension and thus may not run on anything except
Mesa. If these applications used the proper extension mechanism they could
then be used on more than "just" fxMesa to good effect (for example, the
rest of the "3Dfx/demos" run just peachy on the Intel740 card in my test
machine) under WinNT.

Because I'm anal about my computer and it's organization, and I like to
prevent collision between builds, each of the subprojects has their own
intermediate file directory inside .\win32\release (for example, when
building mesagl.lib all of it's intermediate files will be found in
.\win32\release\lib.mesagl).  This makes it very easy to cleanup as you
only need to remove .\win32\release.

*** Okay, Enough, how do I build with this stuff already Ted!

Okay, no major calamity here. The basic way to use the project file is to
call it via NMAKE from the command line. The format is:

    nmake[.exe] /f nmake.mak [options] [target]

The most likely [options] values you will use may be any combination of the
following:

    DEBUG=1 or DEBUG=0
    USE_MMX=1 or USE_MMX=0
    USE_CRTDLL=1 or USE_CRTDLL=0

    Note that all three of these options are OFF by default.

The [target] includes but is not limited to the following (for full details
please peruse the NMAKE.MAK and NMAKE.MIF files - but be warned that
NMAKE.MIF is rather large and sometimes hard to follow):

    --- convenience targets ---

    all                 - builds everything
    libfiles            - builds all linking library files
    progs               - builds all executable images

    --- library files, static and dynamic ---

    mesagl              - static lib build of Mesa core.
    mesaglu             - static lib build of MesaGLU core.
    mesaglut            - static lib build of Mesa GLUT core.

    mesagl32            - dynamic lib build of Mesa core.

    mesaglu32           - dynamic lib build of GLU core, generates
                          GLU32.DLL and/or GLU32d.DLL.

    mesaglut32          - dynamic lib build of GLUT core, generates
                          GLUT32.DLL and/or GLUT32d.dll.

    --- hardware accelerated mesa builds ---

    fxmesagl32          - builds Mesa for use on top of the 3Dfx
                          Glide runtime libs

    s3mesagl32          - builds mesa for use on top of the S3
                          'S3Tk' runtime libs.

    --- executable images ---

    progs.book          - builds all programs in \book directory
    progs.demos         - builds all programs in \demos directory
    progs.samples       - builds all programs in \samples directory

        These targets generate all of the programs in their respective
        directories and link the executables against OpenGL32.DLL,
        GLU32.DLL, and GLUT32.DLL (or their debug equivalents).

    progs.3dfx.demos    - builds all programs in \3dfx\demos directory

        This target generates the 3Dfx/Demo executables, linking them
        against GLUT32.DLL, GLU32.DLL, OPENGL32.DLL and are thus NOT
        hard-bound to using Mesa per-se as you can simply NOT build the
        Mesa core and GLU libraries.

   --- Microsoft/SGI OpenGL-based GLUT and Demo program builds ----

   *** IMPORTANT SAFETY TIP: If you're going to build these variants of
       GLUT then DO NOT build any other target libraries in this package
       first, OR from the command line run the "nmake /f nmake.mak clean"
       command first!  This is because generation of the GLUT for SGI
       OpenGL target libraries conflicts in naming with the static build
       libraries of Mesa and it's supporting GLUT build.

   Currently, you may build GLUT as either GLUT32.DLL or GLUT.DLL for
   use running against either Microsoft or SGI OpenGL for Window,
   respectively.  This allows for the general use of GLUT 3.7 on Windows
   systems with fully compliant OpenGL.

   You can build the GLUT DLL files either with the command line by
   issuing either of these commands:

        nmake /f nmake.mak glut.sysgl

        <or>

        nmake /f nmake.mak glut.sgigl

   OR by using the DevStudio MesaLib Worksapce build the GLUT_SGIGL or
   GLUT_SYSGL projects within the DevStudio IDE.

   Unfortunately, the only way to build the test programs against this
   build of GLUT is via the command line, and I will NOT be making
   duplicate demo program projects for the IDE as it's just not worth it,
   sorry.

   To build the test programs against either MS or SGI OpenGL, you do so
   via either of these two commands:

        nmake /f nmake.mak progs.sysgl

        <or>

        nmake /f nmake.mak progs.sgigl

   To use the GLUT-for-system-OpenGL in your own programs, you need to do
   three things by way of preparation, after building GLUT of course:

         1. Copy include\gl\glut.h to somewhere in your %INCLUDE% path, one
            likely candidate location would be in your
            "DevStudio\VC\INCLUDE\GL" directory.

         2. Copy the linking libraries to somewhere in your %LIB% path, one
            likely candidate location would be in your "DevStudio\VC\LIB"
            directory. The linking libraries you need to copy are as
            follows:

                .\Release\GLUT32.LIB
                .\Release\GLUT.LIB
                .\Debug\GLUT32.LIB
                .\Debug\GLUT.LIB

        3. Copy the runtime libraries to somewhere in your %PATH%, one
           likely candidate location would be in WINDOWS\SYSTEM. the files
           that you should copy are as follows:

                .\Release\GLUT32.DLL
                .\Release\GLUT32.PDB
                .\Release\GLUT.DLL
                .\Release\GLUT.PDB
                .\Debug\GLUT32d.DLL
                .\Debug\GLUT32d.PDB
                .\Debug\GLUTd.DLL
                .\Debug\GLUTd.PDB

Some examples are in order ...

    ... build all dynamic-link libs using MSVCRT.DLL for C runtime:

        nmake /f nmake.mak USE_CRTDLL=1 alldynamic

    ... To build all library variants and all test and demonstration
        programs with the default settings you do this:

        nmake /f nmake.mak all

    ... to build all static link libs and nothing else you do this:

        nmake /f nmake.mak allstatic

    ... to build all non-accelerated dynamic link libs you do this:

        nmake /f nmake.mak alldynamic

    ... to build all 3Dfx targeted dynamic link libs you do this:

        nmake /f nmake.mak allaccel

    ... to build all S3 Virge targetd dynamic link libs you do this:

        nmake /f nmake.mak alls3

    ... to build all libraries, static and dynamic, in all versions
        you do this:

        nmake /f nmake.mak libfiles

    ... to subsequently build all demo and test programs you do this:

        nmake /f nmake.mak progs

    ... to cleanup all intermediate files you do this:

        nmake /f clean

You get the picture. (I hope) ;^)  You may also specify specify
single targets in a convenient fashion. The rule is simple, any of the
above named lib files, static or dynamic, may be built by providing it's
name on the command line as the target. Examples:

    ... to build only Mesa as OpenGL32.DLL ...

        nmake /f nmake.mak opengl32

    ... to build only Mesa on top of the 3Dfx Glide API ...

        nmake /f nmake.mak fxMesaGL32
              <or>
        nmake /f nmake.mak fxMesaGL

    ... to build only Mesa on top of the S3 Toolkit ...

        nmake /f nmake.mak s3MesaGL32
              <or>
        nmake /f nmake.mak s3mesaGL

*** Revision history for ./win32 project files

1/18/98 - initial cut submitted and included with core mesa
2/5/98  - fixed internal dependency within nmake.mif upon there being
          a $(DEVDIR) variable to make some temporary batch files
          dependant upon (thanks to Keven T. McDonnell for finding
          that there was this particular bug). I also updated the
          build files for 2.6beta6.
2/8/98  - added DevStudio workspace and project files for all lib
          files and some test programs. Updated readme.win32.
6/25/98 - initial revision for Mesa 3.0, does not include IDE files,
          not everything is running. *sigh*
7/20/98 - Mesa 3.0beta6 rev of all build files, all libs built and
          minimally tested, all demo programs built and minimally
          tested to within limits of my PC. ;^) Eveything looks
          MUCH better now ...
7/30/98 - Minor updates/edits based upon feedback from
          Eero Pajarre <epajarre@koti.tpo.fi>. These updates include a fix
          to the Mesa-on-3Dfx build such that Quake-II now runs almost
          properly on my system. It runs, just *very* slowly and with *no*
          textures. Hmmm. Doesn't make any difference whether Quake is set
          to use 8-bit textures or not.
8/13/98 - Lots of build cleanups, minor bug fixes in fxwgl.c, and
          compatability fix in fxapi.c for in-window rendering using 3Dfx
          hardware.
8/26/98 - Final revisions for Mesa 3 release checked
9/22/98 - Fixed static builds for all but fxMesaGL32 and s3MesaGL32 targets
9/29/98 - Reorganized FAQ information and added Added faq entry about Glide
          bug under NT (crash on exit) and a workaround.
11/21/98 - Updated files for Mesa 3.1 beta 1
           Updated fxMesa window-hack code
           Updated fxMesa resolution support to handle 1600x1200 & 1280x1024