This is the mail archive of the libc-alpha@sources.redhat.com mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Building arm toolchain with soft floating point (was: Re: non local labels in some handcoded assembly)


Dan:

Yes, this will be my last message on this in this forum.
And yes the howto contained some more recent (and desperate attempts) to 
eliminate the fp code.
I have just rerolled the howto  to an ealier version.

My recent tootlchain attempts which ultimately failed because of glibc.
consisted in hacking  the gcc spec file  to always use -msoft-float
for compilation and to add -without-fp to the glibc configuration.
This resulted in a bunch of undefined basic math/arithm  symbols during make.
This is described in another posting to this list which I was not able to send 
all morning. ;-(
I'll have a look at you goole finding next. 
Thanks,

Robert


On Tuesday 26 August 2003 11:30, Dan Kegel wrote:
> robert wrote:
> > I managed to build  a cross compilation a while abck with as described in
> >
> > http://www.muth.org/Robert/Patch/
> >
> > The problem with this toolchain, that glibc contained assembly code
> > for a totally obsolete arm fp coprocessor.
> > My recent efforts are to get rid of this code and have the floating point
> > stuff all in software.
>
> Hmm.  Looking at how you built the bootstrap gcc:
>
>    ./configure --prefix=$cwd/../usr --target=arm-linux
> --program-prefix=arm- \ --disable-threads -with-cpu=strongarm110
> -enable-languages=c --disable-shared --without-fpu make
>    make CFLAGS+=-msoft-float
>
> That looks a bit fishy.  Why the initial make folowed by a make
> CFLAGS+=-msoft-float?
>
> Comparing with what a couple other folks have done,
> http://gcc.gnu.org/ml/gcc/2003-03/msg00452.html
> http://sources.redhat.com/ml/crossgcc/2002-02/msg00128.html
> makes me wonder if you shouldn't be adding the
> --with-softfloat-support=internal option when you build gcc.  (I wouldn't
> know; I'm just a google junkie.)
>
> This may be getting a bit offtopic for the libc-alpha list...
> maybe we should move this thread to the gcc or crossgcc list?
>
> - Dan


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]