Needed for https://reviews.llvm.org/D39218
Details
Diff Detail
- Repository
- rL LLVM
Event Timeline
I don't necessarily object to this accessor.
But for our generic non-emscripten toolchain support that we are adding right now, maybe we should make it have whatever defaults we are building now with any OS (not just explicitly "unknown")?
D39218 uses the path we want want if the OS is 'unknown'. But shouldn't we make it so that any 'wasm-*-*' triple behaves the same way? i.e. someone who wants to put their own OS name in there but wants to use that same layout (or for that matter, maybe we want to adopt that in emscripten too).
Hmm.. i'm not sure I understand why one might want to put a non-"unkown" OS string in there. What would that mean exactly? And if you did want to that it seems reasonable that at the very least you compile your own version of libclangrt, no?
If you wanted to come up with your own set of syscalls or lower-layer stuff, You'd basically be making up your own OS, which you could name. But libclangrt.builtins in particular doesn't really depend on the OS, it's mostly just architecture-specific or processor-specific stuff. So you could reuse the clang support for that.
Anyway I didn't mean to hold up this little CL over that; just an observation.
I see what you mean, but currently the libclant_rt.* libraries are segmented by "OS", so we are simply following that (perhaps broken) pattern for now.