Facing the CUDA-12 + GLib issue macro clash issue

I am working on MATLAB - Jetson, i am running a MATLAB example, Fog Recitfication on NVIDIA Jetson, i will share the code so that you can understand the issue,

function fog_rectification() %#codegen

% coder.updateBuildInfo('addNonBuildFiles','mw_noinline_fix.h'); 
% coder.updateBuildInfo('addIncludePaths','.');                  
% coder.updateBuildInfo('addCompileFlags','-include mw_noinline_fix.h');
coder.gpu.kernelfun;
hwobj = jetson();
width = 600;
height = 404;
videoFileName = 'foggy.mp4';
vobj = VideoReader(hwobj,videoFileName,'Width',width,'Height',height);
dispObj = imageDisplay(hwobj);

while vobj.hasFrame
    input = vobj.readFrame;
    
    % Run fog rectification on the frame
    outrectifiedImage = fog_rectification_algorithm (input,width,height);
    reshapedImg = cat(3, outrectifiedImage(:,:,1).', ...
        outrectifiedImage(:,:,2)',outrectifiedImage(:,:,3)');
    image(dispObj,reshapedImg);
end
end

and here is the MATLAB code for data generation

hwobj = jetson('169.254.172.219','hr','0000'); 
envCfg = coder.gpuEnvConfig('jetson'); 
envCfg.BasicCodegen = 1; 
envCfg.Quiet = 1; envCfg.HardwareObject = hwobj; 
coder.checkGpuInstall(envCfg); cfg = coder.gpuConfig('exe');
cfg.Hardware = coder.hardware('NVIDIA Jetson');
cfg.Hardware.BuildDir = '~/remoteBuildDir2'; 
cfg.GenerateExampleMain = 'GenerateCodeAndCompile'; 
codegen('-config ',cfg,'fog_rectification','-report');

the error i am facing is..

Error executing command "touch -c /home/hr/remoteBuildDir2/MATLAB_ws/R2024a/C/Users/Syed_Jameel/OneDrive/Documents/MATLAB/Examples/R2024a/nvidia/FogRectificationForVideoOnNVIDIAJetsonExample/codegen/exe/fog_rectification/*.*;make -f fog_rectification_rtw.mk -j4 all MATLAB_WORKSPACE="/home/hr/remoteBuildDir2/MATLAB_ws/R2024a" LINUX_TARGET_LIBS_MACRO="" -C /home/hr/remoteBuildDir2/MATLAB_ws/R2024a/C/Users/Syed_Jameel/OneDrive/Documents/MATLAB/Examples/R2024a/nvidia/FogRectificationForVideoOnNVIDIAJetsonExample/codegen/exe/fog_rectification". Details

STDERR: In file included from /usr/lib/aarch64-linux-gnu/glib-2.0/include/glibconfig.h:9,

from /usr/include/glib-2.0/glib/gtypes.h:32,

from /usr/include/glib-2.0/glib/galloca.h:32,

from /usr/include/glib-2.0/glib.h:30,

from /usr/include/gstreamer-1.0/gst/gst.h:27,

from /home/hr/remoteBuildDir2/MATLAB_ws/R2024a/C/ProgramData/MATLAB/SupportPackages/R2024a/toolbox/target/supportpackages/nvidia/sources/server/frameReader.h:5,

from /home/hr/remoteBuildDir2/MATLAB_ws/R2024a/C/Users/Syed_Jameel/OneDrive/Documents/MATLAB/Examples/R2024a/nvidia/FogRectificationForVideoOnNVIDIAJetsonExample/codegen/exe/fog_rectification/fog_rectification.cu:23:

/usr/include/glib-2.0/glib/gmacros.h:249:28: error: missing ')' after "__has_attribute"

249 | #if g_macro__has_attribute(__noinline__)

| ^~~~~~~~~~~~

In file included from /usr/local/cuda/bin/../targets/aarch64-linux/include/device_types.h:59,

from /usr/local/cuda/bin/../targets/aarch64-linux/include/builtin_types.h:56,

from /usr/local/cuda/bin/../targets/aarch64-linux/include/cuda_runtime.h:90,

from <command-line>:/usr/local/cuda/bin/../targets/aarch64-linux/include/crt/host_defines.h:83:23: error: missing binary operator before token "( 83 | __attribute__((noinline))

[Ramakrishna Mandalapu]

Hi Syed,

Can you provide the Jetson board type and the MATLAB version used?

Thanks,

Ramakrishna

[Isha]

Hello Syed,

This is a known issue: both GLib and CUDA use the identifier __noinline__ (and related attribute checks like __has_attribute) in their headers. Wrap your includes so CUDA’s __noinline__ doesn’t pollute GLib headers:

 // In a custom header, e.g. mw_noinline_fix.h 
     
    #ifdef __CUDACC__ 
      #pragma push_macro("__noinline__") 
      #undef __noinline__ 
    #endif 
     
    #include <glib.h> 
    #include <gst/gst.h> 
     
    #ifdef __CUDACC__ 
      #pragma pop_macro("__noinline__") 
      #endif

Then, in MATLAB’s build info:

      coder.updateBuildInfo('addIncludePaths','.'); 
      coder.updateBuildInfo('addNonBuildFiles','mw_noinline_fix.h'); 
      coder.updateBuildInfo('addCompileFlags','-include mw_noinline_fix.h'); 

This ensures GLib headers are parsed cleanly, and CUDA still gets __noinline__ where it needs it.

You can also refer to the link below:

https://stackoverflow.com/questions/70301375/noinline-macro-conflict-between-glib-and-cuda?newreg=9dfeef18bee7466e956fe3c0c28dc627

Hope this helps.