e-mattes

(Both CopperSpice and the test program are compiled with -stdlib=libc++ present/absent.)

I found a workaround/solution: Pass RTLD_GLOBAL to dlopen().

I don't really like using RTLD_GLOBAL for loading a shared library that uses CopperSpice as the CopperSpice shared libraries dynamically export a lot of symbols (e.g., those of libtiff) which...

BTW, the documentation should mention that Linux executables (not only shared libraries!) need to dynamically export symbols to make CopperSpice work. For now, I use a linker script to export...

I'm using Visual C++ toolchain version 143 (Visual Studio 2022), -MD (or -MDd for debug builds), Windows 10. Here's the definition of MemoryBarrier in C:/Program Files (x86)/Windows Kits/10/Include/10.0.17134.0/um/winnt.h: #define MemoryBarrier...

Here's a direct comparison of the message boxes from diamond-1.5.0-osx11.5-x64.dmg (running with x86_64 emulation) and the arm64 version I built myself. I do not know whether the problem is due...

(The arm64 version had the focus, that's why there are some differences in color.)

If I replace the CS dylib and so files in the app unpacked from diamond-1.5.0-osx11.5-x64.dmg with files I built myself (for x86_64), the windows look OK. If I build Diamond...

I tried again: | CopperSpice | diamond | Result | ----------------- | ------------ | ---------- | dmg | dmg | ok | dmg | mine | ok | mine |...

I prefer Unix Makefiles over Ninja because CMAKE_VERBOSE_MAKEFILE doesn't work with ninja. I'll try building with ninja next week. All of my own apps show the same behavior, that's why...