[llvm] [MCJIT] Fix frem.ll test failure with LLVM_ENABLE_RPMALLOC on Windows (PR #200319)

[Lang Hames via llvm-commits]

lhames wrote:

Ok.

A couple of related questions have just occurred to me: 

1. What happens if you drop the `fflush` from the testcase and do nothing else? I'd have expected the output to be flushed when `lli` exited anyway? Do we need the `fflush` in the testcase at all?

If dropping `fflush` entirely works then we should probably just do that, and address the fact that `fflush` resolves incorrectly as a separate issue.

2. Is this MCJIT specific? What happens if you use the `-jit-kind=orc` option in the `frem.ll` testcase?

Since my suggestion above doesn't work, I think the best thing to do is inject `fflush` and any other required symbols into the JIT symbol tables. MCJIT and ORC both support this, but via different APIs.

https://github.com/llvm/llvm-project/pull/200319