r177329 - This code works around what appears to be a bug in another part of clang.

[reed kotler]

On 03/19/2013 08:07 AM, Rafael Espíndola wrote:
>> Okay. thanks for the help.
>>
>> I will take care of this today.
> Fixed in r177402. Please put a bit more effort into testing and
> understanding the problems next time. This thread took more effort
> than the bug it is about :-(
>
>> Reed
>>
>>
> Cheers,
> Rafael
I appreciate all the help.

I understand what you want in the future as to how to handle these 
situations in the code when they arise and I will do my best to adhere 
to it.

I would like to then add an llvm_unreachable in our code to make sure 
that I don't again see something unexpected.

So I'm not expecting to see a Function Declaration and then a Global 
Value that is not a function when attributes are being propagated.

I will file a bug against this whole file TargetLowering.cpp. There is 
similar cleanup needed everywhere for several issues that I've found 
while adding the Mips attributes.

Reed