Targeting LLVM is an easy and obvious way, but let's mind the disadvantages.
1. You start to depend on a LLVM project existence, plans, roadmaps, and ecosystem. Your resources, as a small independent grass–root publisher, are nothing compared to Apple ones. Sooner or later the effort of staying in sync with newer LLVM versions will become unbearable and your project will die. 2. Yeah yeah, you can target both LLVM and GCC. Bamm! Now you have 2 hardwired dependencies instead of one. 3. Depending on "C/C++", instead of "GCC/LLVM", is completely different story. ISO/IEC standards cannot disappear, cannot pivot, cannot change beyond any recognition, cannot switch to prohibitive licensing. 4. Think about compilers for specialized embedded hardware. IAR, Keil, etc. The possibility to reach exotic platforms is a big deal.
