This is the mail archive of the
newlib@sources.redhat.com
mailing list for the newlib project.
Re: building newlib for i386-rtemself
- From: Joel Sherrill <joel dot sherrill at OARcorp dot com>
- To: Schwarz Robert <robert2 dot schwarz at erlf dot siemens dot de>
- Cc: newlib at sources dot redhat dot com
- Date: Tue, 01 Jun 2004 10:29:38 -0500
- Subject: Re: building newlib for i386-rtemself
- Organization: OAR Corporation
- References: <512E099F0DE23F4C8A8C39FB1B881FCBA0E6DE@erlf307a.erlf.siemens.de>
Schwarz Robert wrote:
Hello,
I got trouble building newlib-1.9.0 for i386-rtemself-targets. I'm using the
standard configure like
../newlib-1.9.0/configure --target=$TARGET --prefix=$PREFIX |& tee
configure.log
and it works fine but when making it, it stops at this point:
f_exp.c:34: error: impossible constraint in `asm'
gmake[6]: *** [f_exp.o] Error 1
gmake[6]:
I think the trouble is about soft-float. But our Target has a hardware
floating-point-unit.
I got the feeling that he's doing it fine for all the other TArgets(seems
like he's building for athlon,m486,k6 too) only for i386 he's making
trouble.
Somebody got an Idea?
When you build gcc/newlib, you build multilib verions of libraries
since the toolset supports multiple CPU models within the family.
newlib 1.9.0 is old so that probably means that the gcc you are using
is of similar vintage. some gcc versions had a lot of soft fp bugs.
RTEMS including the soft-float multilib option for CPUs like the
i386ex pushed at those corners. You should be able to
configure with --disable-multilib or edit a file like
gcc/config/i386/t-XXX and eliminate the multilibs you do not want.
Or better.. just update to newer stuff and probably get pre-built
binaries :)
--joel