MSVC parser stack overflow

[kar-dim]

When building the latest stable CImg (v3.7.5) and pre-release 3.7.6, I get compiling errors when compiling with MSVC:

2>C:\...path_here...\libs\CImg\CImg.h(27348): fatal error C1026: parser stack overflow, program too complex
2>C:\...path_here...\libs\CImg\CImg.h(27348): note: the template instantiation context (the oldest one first) is
2>C:\...path_here...\libs\CImg\CImg.h(12751): note: while compiling class template 'cimg_library::CImg'
2>C:\...path_here...\libs\CImg\CImg.h(20316): note: see reference to class template instantiation 'cimg_library::CImg<unsigned int>' being compiled
2>C:\...path_here...\libs\CImg\CImg.h(46327): note: see reference to class template instantiation 'cimg_library::CImg<float>' being compiled
2>C:\...path_here...\libs\CImg\CImg.h(32985): note: while compiling class template member function 'double cimg_library::CImg<float>::_eval(cimg_library::CImg<float> *const ,const char *const ,const double,const double,const double,const double,cimg_library::CImgList<float> *const ) const'
2>C:\...path_here...\libs\CImg\CImg.h(32884): note: see the first reference to 'cimg_library::CImg<float>::_eval' in 'cimg_library::CImg<float>::eval'
2>C:\...path_here...\libs\CImg\CImg.h(71012): note: see the first reference to 'cimg_library::CImg<float>::eval' in 'cimg_library::cimg::eval'
2>C:\...path_here...\libs\CImg\CImg.h(17857): note: while compiling class 'cimg_library::CImg<float>::_cimg_math_parser'
2>C:\...path_here...\libs\CImg\CImg.h(27309): note: while compiling class template member function 'double cimg_library::CImg<float>::_cimg_math_parser::mp_debug(cimg_library::CImg<float>::_cimg_math_parser &)'
2>C:\...path_here...\libs\CImg\CImg.h(21513): note: see the first reference to 'cimg_library::CImg<float>::_cimg_math_parser::mp_debug' in 'cimg_library::CImg<float>::_cimg_math_parser::compile'
2>C:\...path_here...\libs\CImg\CImg.h(18014): note: see the first reference to 'cimg_library::CImg<float>::_cimg_math_parser::compile' in 'cimg_library::CImg<float>::_cimg_math_parser::_cimg_math_parser'
2>C:\...path_here...\libs\CImg\CImg.h(32991): note: see the first reference to 'cimg_library::CImg<float>::_cimg_math_parser::_cimg_math_parser' in 'cimg_library::CImg<float>::_eval'
2>C:\...path_here...\libs\CImg\CImg.h(32884): note: see the first reference to 'cimg_library::CImg<float>::_eval' in 'cimg_library::CImg<float>::eval'
2>C:\...path_here...\libs\CImg\CImg.h(71012): note: see the first reference to 'cimg_library::CImg<float>::eval' in 'cimg_library::cimg::eval'

System: Windows 11
Compiler: MSVC, Visual Studio 2022 (v143)
C++ flag: /std:c++20

[dtschump]

That looks like a limitation of the MSVC compiler.
Claude suggests:

In the Visual Studio project properties:
Project Properties → C/C++ → Command Line → Additional Options:
/Zm300
Or up to /Zm500 if necessary. This flag increases the memory allocated to the compiler for templates.

Not sure if this can be useful or not. As I'm mainly working on Linux, that's an issue I can hardly reproduce.

[kar-dim]

Hmm this did not fix it. I guess I will switch to clang if I want to upgrade CImg, so for now CImg is not compatible with MSVC (with the above configuration), just wanted to share anyway. You can close the issue, thank you!