[PATCH] D50998: [LLD] [COFF] Check the instructions in ARM MOV32T relocations

[Reid Kleckner via Phabricator via llvm-commits]

rnk added a comment.

In https://reviews.llvm.org/D50998#1212923, @mstorsjo wrote:

> Does @efriedma have any opinion about this one? It's slightly unconventional to check the instruction contents in the linker, but OTOH it's at least a simple and safe place to check that nothing really did break these pairs. Unless there's some easy way to add a check in LLVM (and I guess the correct solution there is to keep them together as a pseudo instruction longer, as suggested in the other review), do you think this would be worthwhile?


Not @efreidma, but that doesn't stop me from opining... I think it's a reasonable safety check. I don't think we have any relaxations in the COFF linker, but they are important elsewhere (turn GOT load into `leaq sym(%rip)` for ELF), and that requires looking at instructions.


https://reviews.llvm.org/D50998