the attached patch lets @encode(wchar_t) not crash (with test case). The returned value matches g++, which returns "i" for both i386 and x86_64 archs.
I committed this part, thanks!
The patch also makes one of lits error messages less confusing.
I didn't commit this part; it looks reasonable to me, but Daniel should weigh in.
Daniel, what do you think of the lit patch? Would've saved me some 5
minutes back then.
Sure, looks fine. I would phrase it:
lit.fatal('No site specific configuration available! You may need to run 'make test' in your Clang build directory.')