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

[Evan Cheng]

On Jul 7, 2008, at 2:55 PM, Devang Patel wrote:

>
> On Jul 7, 2008, at 1:59 PM, Chris Lattner wrote:
>
>>
>> On Jul 7, 2008, at 1:55 PM, Eric Christopher wrote:
>>
>>>>
>>>> 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".
>>>
>>> Can't just use -Wl, then?
> (BTW, this should work without the patch that started this thread)
>
> Again, if the linker does not support -mllvm then -Wl,-mllvm will fail
> anyway ;).
>
>> Yes, that is fine with me!  What do you think Devang/Evan?
>
> I'm fine with current state :)
>
> I do not prefer separate new -mllvm option for the linker.

The issue is when user call llvm-gcc -mllvm -foo x.c which would pass  
the option to both the compiler proper and the linker. It seems like  
an undesirable behavior.

Evan

>
>
> This is a debugging aid for llvm hacker that helps him investigate
> code generation/optimization bugs. If the user does not know what he
> is debugging (lto or no-lto) then all bets are off!
>
> If the user wants llvm code generator to print useful info to find out
> what went wrong, when LTO is ON, then you really want the linker to
> accept this command or reject. You do not want to let compiler driver
> eat this silently and not ask LTO to produce any output, which may
> mislead the user.
>
> -
> Devang
>
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits