OpenGL + Mesa 3D + MinGW

Question

I program C++ applications on (Ubuntu) Linux and compile them to 2 operating systems: natively to Linux by using "g++" (GNU C++ compiler) and cross-compile them to Windows by using "i386-mingw32-g++" (MinGW C++ cross-compiler).

Now, I am trying to cross-compile "OpenGL" applications (from Linux to Windows) - for that I need some OpenGL library files.
As an implementation of OpenGL, I use "Mesa 3D" open source library. Fortunately, Ubuntu offers "precompiled" mesa3d libraries (libgl1-mesa-dev and libglu1-mesa-dev) for Linux, but unfortunately, there are not any precompiled libraries for using with MinGW cross-compiler (thus Windows versions of them) - so I "MUST" manually compile them "from source" (which can be downloaded from http://www.mesa3d.org homepage).

But I have no idea to do that. Within the mesa3d source code, there are some documents how to build the libraries for some specific platforms - specifically there is "README.MINGW32" readme file.
But just on the first lines there is some command

mingw32-make -f Makefile.mgw [OPTIONS...]

which I cannot run because I don't have any "mingw32-make" program installed. By other softwares, I used to build them with the "classic"

./configure -> make -> make install

procedure, but this is not working on mesa3d libraries.
For example, I have also tried

./configure --host=i386-mingw32

and it has been configured ok but after typing

make

it threw some error

error: #error No PIPE_SUBSYSTEM_WINDOWS_xxx subsystem defined.

and now I am LOST :-( .

So does anybody know how to build the mesa3d libraries for using them with mingw32?

P.S.: And of course, I DON'T WANT to use Microsoft Visual Studio NEVER MORE :-) . And also I would like to do all the required things to build up OpenGL environment from Linux and using mingw32 for cross-compiling. I hope it is possible.

P.S.2: On http://www.mesa3d.org/systems.html page there is written that using MinGW with Mesa 3D is "deprecated". I am on the right way to use it just that way?

Answer 1

There is no need to compile Mesa3D for MinGW. MinGW includes the GL library; it's just not called libGL.a nor libGLU.a; the libraries are instead called libopengl32.a and libglu32.a. Sadly, this is what Microsoft decided to call them under Windows on Visual Studio, so both Windows and GNU/Linux release of MinGW decided to include the libraries named as above.

So, when cross compiling for Windows, just change:

-lGL -lGLU

into

-lopengl32 -lglu32

---

Mesa3D on GNU/Linux is actually a name for a libre implementation of OpenGL. This is because OpenGL is a trademark, and SGI's policy did not allow for anyone to use the name without paying a hefty sum of money. Despite this, SGI provided the author, Brian Paul, a copy of the testing suite.

Mesa3D is on most platforms with native OpenGL implementation known as a software "sanity check" library with emulation of numerous high-end-hardware-only features. But on GNU/Linux, much more important is something called DRI or direct rendering interface. This is the real deal, the hardware acceleration and all. This library is also called Mesa3D; in fact, provisioning for DRI is a part of Mesa3D as well as of X11 and the kernel.

Mesa3D provides libre OpenGL-compatible headers on GNU/Linux and other free platforms.

Summary: Mesa3D is the name for the GNU/Linux implementation of OpenGL since that name is trademarked. This library includes an excellent, featureful, correct and relatively fast software implementation. It also includes direct rendering interface for hardware acceleration.