[gccsdk] Building UnixLib

Christian Ludlam christian at recoil.org
Sun Feb 11 23:44:59 GMT 2007


On 11 Feb David J. Ruck wrote:

> On 11 Feb 2007 Peter Naulls <peter at chocky.org> wrote:
> > In message <9747a6b34e.druck at druck.freeuk.net>
> >           "David J. Ruck" <druck at druck.org.uk> wrote:
> >>     [ {SOFTFLOAT}={FALSE}
> > 
> > Perhaps.  If memory serves, both objasm and 'as' complain about
> > undefined symbols.  The symbol in question is only defined in 'as'
> > because it was added some time ago during experiments with GCC +
> > softfloat.  I don't think it would be unreasonable for the symbol to be
> > defined in the makefile or source under this combination.
> 
> The makefile is has ASflags containing
> 
>    -predefine "SOFTFLOAT SETA 0" 
>    
> which should be ok, and {FALSE} should already be known. Its not something
> I've used in objasm myself, but it seems to match the syntax given in the
> DDE manual, so I'm not sure what else I'm missing.

It's objecting to the braces: only builtin symbols are enclosed in them, the
line should read

 [ SOFTFLOAT = {FALSE}
 
and since {FALSE} is a bool, you probably need to use SETL in the makefile.

ARM's RVDS assembler uses the builtin {FPU} which returns the name of the FPU
system as a string (specified on the command line). It returns any of

 none
 vfpv2
 softvfp
 fpa        } obsolete
 softfpa    }

I don't know if recent ObjAsms support this too.

-- 
Christian Ludlam
christian at recoil.org



More information about the gcc mailing list