GCC Bugzilla Bug 14352

Most of the ARM target makefile fragments reference arm/lib1funcs.asm, but only
t-arm-elf references all the symbols that need to be defined to get the new
floating point code:
LIB1ASMFUNCS = _udivsi3 _divsi3 _umodsi3 _modsi3 _dvmd_tls _bb_init_func \
        _call_via_rX _interwork_call_via_rX \
        _negdf2 _addsubdf3 _muldivdf3 _cmpdf2 _unorddf2 _fixdfsi _fixunsdfsi \
        _truncdfsf2 _negsf2 _addsubsf3 _muldivsf3 _cmpsf2 _unordsf2 \
        _fixsfsi _fixunssfsi

t-linux just says this:

LIB1ASMFUNCS = _udivsi3 _divsi3 _umodsi3 _modsi3 _dvmd_lnx

This generates undefined references in libgcc_s.so to, for instance, the functions
defined by L_addsubdf3.

If I copy LIB1ASMFUNCS from t-arm-elf to t-linux that solves the problem (for
this particular target at least, and keeping _dvmd_lnx...)

Subject: Re:  New: New FP code not compiled for 
 non-arm-elf targets

> t-linux just says this:
> 
> LIB1ASMFUNCS = _udivsi3 _divsi3 _umodsi3 _modsi3 _dvmd_lnx
> 
> This generates undefined references in libgcc_s.so to, for instance, the functions
> defined by L_addsubdf3.
> 
> If I copy LIB1ASMFUNCS from t-arm-elf to t-linux that solves the problem (for
> this particular target at least, and keeping _dvmd_lnx...)

t-linux is not a soft-float target, so shouldn't need any of the FP 
support functions.

R.

Subject: Re:  New FP code not compiled for non-arm-elf targets

But this prevents using it with -msoft-float; which is often
advantageous.  Hmm, I'm not sure how best to do it, but I'd like to
make this new requirement clearer - especially now that we have
--with-float=soft.

I just ran into this again.  Richard, I know the soft-float routines are not
generally needed, but would there be any harm in building them anyway?

An alternative would be to detect --with-float=soft in t-linux, and adjust
LIB1ASMFUNCS.  I think that can be done.

Confirmed.