[llvm] r182092 - [PowerPC] Merge/rename PPC fixup types

[David Fang]

>> Judgment call for the core-team, and potential reviewers:
>> Would a far-from-complete PPCMachObjectWriter be accepted on trunk, with the
>> caveat that it is highly experimental?  This would allow cross-compilation
>> to PIC mach-o/PPC objects.  I think I could extract out this part of my
>> branch into a patch for review.  I need to take the time to write some test
>> cases for it.
>
> I think so. The writer itself can be tested from any host, so if you
> can split it and add tests I don't see why it should not be accepted.
>
> Note that the tests should use llvm-mc to to produce a macho .o from a
> .s if at all possible.

I would ideally like to close the loop and take llvm/clang-produced asm 
and feed that back into llvm-mc, if the PPCAsmParser supports it.  Ulrich, 
does the PPCAsmParser parse Apple-asm syntax, like 'r' in register names?
I hand-tested the hello-printf.s produced for PPC/mach-o and it griped 
about invalid operands, and unknown '.machine' directives.

So far, I've only been comparing clang->.o (integrated) vs. 
clang->.s->/usr/bin/as->.o.

Fang

-- 
David Fang
http://www.csl.cornell.edu/~fang/