Subject: Re: Build failure with MKSOFTFLOAT=yes
To: Masao Uebayashi <uebayasi@pultek.co.jp>
From: Nick Hudson <skrll@netbsd.org>
List: current-users
Date: 01/17/2004 18:34:32
On Saturday 17 January 2004 5:00 pm, Masao Uebayashi wrote:
> Mike,
>
> > -G_LIB2FUNCS_1=_muldi3 _negdi2 _lshrdi3 _ashldi3 _ashrdi3 _ffsdi2 _clz
> > _cmpdi2 _ucmpdi2 _floatd idf _floatdisf _fixunsdfsi _fixunssfsi
> > _fixunsdfdi _fixdfdi _fixunssfdi _fixsfdi _fixxfdi _fixu nsxfdi
> > +G_LIB2FUNCS_1=_muldi3 _negdi2 _lshrdi3 _ashldi3 _ashrdi3 _ffsdi2 _clz
> > _cmpdi2 _ucmpdi2 _floatd idf _floatdisf _fixunssfsi _fixunsdfdi _fixdfdi
> > _fixunssfdi _fixsfdi _fixxfdi _fixunsxfdi G_LIB2FUNCS_2=_floatdixf
> > _fixunsxfsi _fixtfdi _fixunstfdi _floatditf _clear_cache _trampoline
>
> Thanks, this is what I'm figuring out...
>
> I hope this can be approached more easily after incorpolating Nick's
> libgcc change.
My changes to libgcc doesn't change how G_LIB2FUNCS_1 is generated. Everything
in libgcc3/arch/*.mk is pulled by mknative-gcc from a Makefile generated by
configuring gcc. If the right fix is to re-order these functions then this
should be done in gnu/dist/gcc/gcc/Makefile.in where LIB2FUNCS_1 is defined.
I briefly looked at gnu/dist/gcc/gcc/libgcc2.c which is the source for these
functions and some do use the others, e.g. _fixdfdi uses _fixunsdfdi so I can
see why the order is important.
Nick