[llvm-commits] [llvm-gcc-4.2] r53113 - /llvm-gcc-4.2/trunk/gcc/config/darwin.h

[Chris Lattner]

On Jul 7, 2008, at 1:30 PM, Evan Cheng wrote:

>
> On Jul 7, 2008, at 1:12 PM, Chris Lattner wrote:
>
>>
>> On Jul 7, 2008, at 10:48 AM, Eric Christopher wrote:
>>
>>>
>>> On Jul 7, 2008, at 10:19 AM, Evan Cheng wrote:
>>>
>>>> It would be nice if someone with configure foo can help with this.
>>>> Perhaps Eric?
>>>
>>> Sure. There are a few ways you can go about this:
>>
>> What fix is required here?  I don't understand the problem.
>>
>> If a user invokes 'llvm-gcc -mllvm -foo x.o' and the linker doesn't  
>> support llvm LTO options, why does it matter if llvm-gcc or ld  
>> rejects them?
>
> Users with old linker may expect the -mllvm options are passed to  
> cc1 only. It could be non-user-friendly if the linker than outputs  
> an error message.  Like Eric said, perhaps it's better to make it  
> explicit the options are meant for the linker (e.g. -mllvm-lto -foo)?

Ok, I can see that.  I'm fine with -mllvm-lto, but I don't think we  
should get more complex than that.  This is really just a "compiler  
hacker feature".

-Chris