Re: [RFC PATCH V2] MIPS: fix build with binutils 2.24.51+

[Ralf Baechle <ralf@linux-mips.org>]

On Tue, Aug 19, 2014 at 06:27:12PM +0200, Manuel Lauss wrote:

> With binutils snapshots since 29.07.2014 I get the following build failure:
> {standard input}: Warning: .gnu_attribute 4,3 requires `softfloat'
>   LD      arch/mips/alchemy/common/built-in.o
> mipsel-softfloat-linux-gnu-ld: Warning: arch/mips/alchemy/common/built-in.o
>  uses -msoft-float (set by arch/mips/alchemy/common/prom.o),
>  arch/mips/alchemy/common/sleeper.o uses -mhard-float
> 
> Extend cflags with a soft-float directive for the assembler, and add
> hardfloat directives to assembler files dealing with FPU
> registers to compensate.

I had a discussion about this with Maciej.  He suggested that this
behavious of binutils should be taken a look at but also that we rather
should remove the -msoft-float option from the kernel and I support his
view.

I did add -msoft-float in 6218cf4410cfce7bc7e89834e73525b124625d4c
[[MIPS] Always pass -msoft-float.] in 2006 because back then there was a
wave of bug reports from people attempting to use hard fp in the kernel
which of course did result in FPR corruption.  Adding -msoft-float made
sure that floating point operations would result in a link error because
the kernel does not supply a soft-fp library.

But maybe there are other methods to achieve the same - such as

#define float diediedie
#define double goboom

  Ralf