r212871 - Require an x86 registered target for this test by splitting it out

[Artyom Skrobov]

>>> Should we just remove it for the time being?
>>
>> Is there a way to put it in an x86 specific directory?
>
> I think Reid tried to fix this in r213180 by passing the --target flag
> to this test.
>
> But it's still failing further down, e.g.
> http://lab.llvm.org:8011/builders/clang-ppc64-elf-linux2/builds/15641/steps/check-all/logs/Clang%3A%3Acl-x86-flags.c
>
> We could sprinkle more --target options in there, but it's starting to
> look weird when we're passing both a --target and /arch: flag, for
> example. Maybe we can make lit support REQUIRES: x86-default-target?

Adding Alp to this discussion, as last year, he expressed his strong reservations against target-triple-specific tests:

> There's already a tendency amongst some contributors to avoid thinking
> about portability using XFAIL, but at least with XFAIL we have a chance
> to detect those even if it takes a few months or years for somebody to
> notice. With REQUIRES, those tests won't even get run.
>
> [...]
>
> There are just too many tests switched off today with REQUIRES because
> the author was too lazy to make their feature portable, or because at a
> later point someone broke the test and found it easier to disable it
> than fix the problem.
>
> [...]
>
> PS. I'd like to remove the target_triple feature from XFAIL as well at
> some point. It has almost no legitimate users.