[PATCH] D28898: [ELF] - Allow emulation to be different from input objects target.

[George Rimar via llvm-commits]

>> +  // If we have at least one archive, -m is ignored and we require
>> +  // to have at least one ELF input file.
>> +  bool HasArchive = llvm::find_if(Files, [](InputFile *F) {
>> +                      return F->kind() == InputFile::ArchiveKind;
>> +                    }) != Files.end();
>> +  if (HasArchive) {
>> +    error("target unknown: at least one ELF file required");
>
>I don't think this is correct. You should wait until the end of symbol
>resolution. Having an archive doesn't guarantee that a member is
>fetched.
>
>Have you found out *why* the kernel build system uses a mismatched -m?
>

I asked next question to Dmitry Goloving who tried booting LLD linked kernel:
(I think he configured it differently and uses few workarounds, that is why kernel links fine for him already)

me:
>> I have a question also. You added -m elf_i386 to workaround emulation conflict issue in LLD, do you know
>> does output produced by BFD boot fine after that change ?
he:
>Doesn't seem to affect BFD at all.

So I think that is like this patch assumes just a mistype or something. Which is accepted both by bfd and gold.
I think we still want to follow that behavior, right ?

>Cheers,
>Rafael