For some reason we hadn't seen such cases in the wild which makes
me think that clang and rustc don't generate these. In the bug which
reproduces it only occurs with LTO so my guess is that some LTO pass
is creating this alias + gep.
See: https://github.com/emscripten-core/emscripten/issues/8731
Of the 3 new tests added here the first 2 already passed, and
test/MC/WebAssembly/offset.s would previously crash.
Can we also check if llvm-mc can correctly parse the result of this test? That will possibly break the current codebase.