Re: How do I keep newlib from building a bunch of variants i don't need? (MORE)

[Rick Mann <rmann at latencyzero dot com>]

And now my build has stopped, complaining about conflicts:

arm-elf-gcc -B/Users/rmann/Desktop/PXA320Work/newlib-1.14-arm-elf/arm- 
elf/fpu/newlib/ -isystem /Users/rmann/Desktop/PXA320Work/newlib-1.14- 
arm-elf/arm-elf/fpu/newlib/targ-include -isystem /Users/rmann/Desktop/ 
PXA320Work/newlib-1.14.0/newlib/libc/include  -mhard-float -DPACKAGE= 
\"newlib\" -DVERSION=\"1.14.0\"  -I. -I../../../../../../ 
newlib-1.14.0/newlib/libc/stdlib  -O2 -DABORT_PROVIDED - 
DHAVE_GETTIMEOFDAY -DARM_RDI_MONITOR -fno-builtin      -O2 -g -O2   - 
mcpu=xscale -march=armv5te  -mhard-float -c ../../../../../../ 
newlib-1.14.0/newlib/libc/stdlib/__adjust.c
../../../../../../newlib-1.14.0/newlib/libc/stdlib/__adjust.c:1:  
warning: switch -mcpu=xscale conflicts with -march= switch
/var/tmp//ccdlVO44.s: Assembler messages:
/var/tmp//ccdlVO44.s:25: Error: selected processor does not support  
`stfe f4,[sp,#-12]!'
/var/tmp//ccdlVO44.s:30: Error: selected processor does not support  
`ldfd f0,.L17'
/var/tmp//ccdlVO44.s:31: Error: selected processor does not support  
`ldfd f1,.L17+8'
/var/tmp//ccdlVO44.s:37: Error: selected processor does not support  
`mvfeqd f4,f0'
/var/tmp//ccdlVO44.s:38: Error: selected processor does not support  
`mvfned f4,f1'
/var/tmp//ccdlVO44.s:45: Error: selected processor does not support  
`mvfd f0,f4'
/var/tmp//ccdlVO44.s:46: Error: selected processor does not support  
`ldfe f4,[sp],#12'
/var/tmp//ccdlVO44.s:55: Error: selected processor does not support  
`mvfltd f4,#0'
/var/tmp//ccdlVO44.s:61: Error: selected processor does not support  
`ldfd f4,[r1,#0]'
/var/tmp//ccdlVO44.s:66: Error: selected processor does not support  
`mnfned f4,f4'
/var/tmp//ccdlVO44.s:77: Error: selected processor does not support  
`mufd f4,f4,f0'
/var/tmp//ccdlVO44.s:86: Error: selected processor does not support  
`dvfd f4,f4,f0'


OF COURSE this won't work. But I don't understand why it's even  
trying to build hard-float version of things.

Also, there are examples that show using "--with-float=soft", but a  
grep of the whole distro looking for "with-float" returned no  
results. I can't find any documentation for things like "with-float"  
and "with-newlib". Where should I look?

TIA,
Rick


On Aug 30, 2007, at 4:19 PM, Rick Mann wrote:

> I really just want a newlib that targets -mcpu=xscale - 
> march=armv5te, without anything else. No thumb, no interwork, no  
> big-endian (as superior as I think Big Endian is), etc.
>
> I tried this:
>
> $ ../newlib-1.14.0/configure --prefix=/usr/local/arm-elf -- 
> target=arm-elf --with-float=soft
>
> and adding "-mcpu=xscale -march=armv5te" to the CFLAGS_FOR_TARGET  
> in the Makefile, but it appears to be creating a lot of variants.
>
>
> TIA,
>
> Rick