Re: [Mesa3d-users] using osmesa together with nvidia GL implementation

[Alex F. <ale...@gm...>]

Thanks Viral,
I will look into this and I'm sure your comments will be most helpful.
Very much appreciate your time.
Cheers,
Alex


On Wed, Feb 17, 2010 at 4:47 PM, Viral Patel <Vir...@an...> wrote:
> It was a while ago that I built Mesa for this purpose, here are some
> hints based
> on my notes and what I recall, there are several ways to skin a cat so
> you may be
> able to do the same in other ways.
>
> This is what I did to build things:
>
> ....................
> make distclean
>
> Edit configs/autoconf.in
>    change "GL_LIB = GL" to "GL_LIB = MyMesaGL"
>    change "GLU_LIB = GLU" to "GLU_LIB = MyMesaGLU"
>    change "GLUT_LIB = GLUT" to "GLUT_LIB = MyMesaGLUT"
>    change "OSMESA_LIB = @OSMESA_LIB@" to "OSMESA_LIB = My@OSMESA_LIB@"
>    Add "CFLAGS += -DUSE_MGL_NAMESPACE" under the existing CFLAGS def.
>    Add "CXXFLAGS += -DUSE_MGL_NAMESPACE" under the existing CXXFLAGS
> def.
>
> ./configure --prefix=<SOME DIRECTORY> --disable-driglx-direct
>                --enable-gl-osmesa=yes --disable-glw --with-driver=xlib
>                --with-osmesa-bits=32
> gmake
> ....................
>
> This will give you a set of libraries such that MyMesaGLU will be linked
>
> against MyMesaGL and the same for GLUT.
>
> Typically Mesa didn't seem particularly setup to be used as an alternate
>
> linkable gl implementation and the default configurations are meant to
> build
> a replacements for the system gl implementation. So you may need to go
> through the headers and check that all symbols are being mangled.
>
> Good luck.
>
> Viral
>
> -----Original Message-----
> From: Alex Flint [mailto:ale...@gm...]
> Sent: 17 February 2010 15:54
> To: Viral Patel
> Cc: mes...@li...
> Subject: Re: [Mesa3d-users] using osmesa together with nvidia GL
> implementation
>
> Thanks for your help Viral.
>
> I am now having difficulty compiling with USE_MGL_NAMESPACE
>
> $ env CFLAGS=-DUSE_MGL_NAMESPACE ./configure
> $ make
> ...
> glxcmds.c:1577: error: 'glXGetCurrentDisplayEXT' aliased to undefined
> symbol 'glXGetCurrentDisplay'
> glxcmds.c:1728: error: 'glXQueryContextInfoEXT' aliased to undefined
> symbol 'glXQueryContext'
> glxcmds.c:2174: error: 'glXGetFBConfigAttribSGIX' aliased to undefined
> symbol 'glXGetFBConfigAttrib'
> glxcmds.c:2178: error: 'glXChooseFBConfigSGIX' aliased to undefined
> symbol 'glXChooseFBConfig'
> glxcmds.c:2183: error: 'glXGetVisualFromFBConfigSGIX' aliased to
> undefined symbol 'glXGetVisualFromFBConfig'
> glxcmds.c:3087: error: 'mglXGetProcAddress' aliased to undefined
> symbol 'glXGetProcAddressARB'
>
> The docs say that I should add "CFLAGS += -DUSE_MGL_NAMESPACE" to the
> relevant config. It seems that configure has added this to
> configs/autoconf:
> $ grep USE_MGL_NAMESPACE configs/*
> configs/autoconf:CFLAGS =  -DUSE_MGL_NAMESPACE -Wall
> -Wmissing-prototypes -std=c99 -ffast-math -fno-strict-aliasing \
> configs/current:CFLAGS =  -DUSE_MGL_NAMESPACE -Wall
> -Wmissing-prototypes -std=c99 -ffast-math -fno-strict-aliasing \
>
> Is this sufficient or should I add it manually to other configs?
>
> Thanks again in advance,
> Alex
>
>
> On Wed, Feb 17, 2010 at 1:24 PM, Viral Patel <Vir...@an...>
> wrote:
>> Hi Alex,
>>
>> You can link in both Mesa and Non-Mesa gl implementations, the key is
> to
>> use
>> mesa in mangled mode, such that, when calling into mesa all gl calls
>> are:
>>
>> mgl*
>>
>> but calling into non-mesa the calls are,
>>
>> gl*
>>
>> This either means you need to write all your code twice or preferably
>> use function
>> pointers/virtual interfaces to reference each implementation.
>>
>> You will also need to be careful to use mgl calls only when using with
>> an OSMesa gl
>> context.
>>
>> What you are currently doing is linking in 2 different libraries with
>> the conflicting
>> symbols, simplistically the linker is using some functions from mesa
> and
>> some from
>> nvogl.
>>
>> I would look at the mangled_* headers in mesa, and the mesa doc has a
>> section on
>> mangling the api entry points.
>>
>> Hope this helps.
>>
>> Viral Patel
>>
>> Ansys Inc,
>>
>>
>> -----Original Message-----
>> From: Alex Flint [mailto:ale...@gm...]
>> Sent: 17 February 2010 12:27
>> To: mes...@li...
>> Subject: [Mesa3d-users] using osmesa together with nvidia GL
>> implementation
>>
>> Hi there,
>>
>> My program is happily linking to the nvidia OpenGL implementation for
>> its normal onscreen rendering. I would like within a small part of
>> this program to use the offscreen rendering capabilities of OSMesa. If
>> I link to both /usr/lib/libGL.so and /usr/lib/libOSMesa.so then I get
>> strange results: either my offscreen-rendered images are blank or GL
>> crashes, depending on which order I do the linking.
>>
>> My first question is: is this even supposed to work? If so I will
>> investigate further and post much more detailed info about what's
>> going wrong.
>>
>> If not, is there any other way that I can get both 3D acceleration for
>> onscreen rendering and (un-accelerated) offscreen rendering working in
>> the same binary?
>>
>> Thanks in advance,
>> Alex
>>
>>
> ------------------------------------------------------------------------
>> ------
>> SOLARIS 10 is the OS for Data Centers - provides features such as
>> DTrace,
>> Predictive Self Healing and Award Winning ZFS. Get Solaris 10 NOW
>> http://p.sf.net/sfu/solaris-dev2dev
>> _______________________________________________
>> Mesa3d-users mailing list
>> Mes...@li...
>> https://lists.sourceforge.net/lists/listinfo/mesa3d-users
>>
>