[llvm-commits] [llvm-gcc-4.2] r103976 - /llvm-gcc-4.2/trunk/gcc/config/i386/i386.h

[Bill Wendling]

On May 18, 2010, at 2:47 PM, Duncan Sands wrote:

> Hi Bill,
> 
>>> Presumably due to how gcc is setting up the
>>> various flags.  Can you please add some comments.
>>> 
>> 
>> That would involve documenting all of the effects that the "flag_frame_pointer" variable has. That's beyond the scope of this patch. And I think that the code is clear enough (certainly enough for GCC-derived code).
> 
> your patch did have me scratching my head, maybe because I hadn't
> seen your later patches at that point.
> 
It's hard to defend the use of 'magic numbers' in this case. :-) But GCC is already using them all over for this flag. I'm hoping that someone who's interested in understanding why "3" is important here will look at the rest of the code. flag_frame_pointer is mercifully used in only a few places. I can add a few simple comments to help this along, though.

-bw