This is the mail archive of the
mailing list for the glibc project.
Re: still building soft-fp on ppclinux
On Sat, Oct 19, 2002 at 02:29:12PM -0400, Jack Howarth wrote:
> Forget about my suggestion of merging the common files
> together. That will never work because the two different
> libm-test-ulps can never be merged. I guess that leaves us
> stuck with hacking the build so that the appearance of
> paths sysdeps/powerpc/fpu and sysdeps/powerpc/soft-fp in
> config-sysdirs of config.make depends on the setting of
This is so odd, because when I tested on my tree (which couldn't have
been more than a week old), the fpu directory got picked up first when
configuring a traditional ppc build, and the soft-fp stuff only got
picked up with --without-fp or with the new
And yes, despite how odd it sounds, I did test on both fpu and
non-fpu, and I did verify what objects got built and what didn't. I
even debugged the math libraries with gdb at the source level to make
sure the assembly wasn't odd.
Be that as it may... This just needs to be fixed. I've updated my
tree only to notice that my patches have been removed, so I guess I'll
have to work from yesterday's tree.
BTW, I have an old ppclinux box. Perhaps it's some weird interaction
mondavi:~$ cat /etc/issue
Yellow Dog Linux release 2.0 (Pomona)
Kernel 2.4.8-0.4gsmp on a ppc (\l)
mondavi:~$ uname -a
Linux mondavi.quesejoda.com 2.4.8-0.4gsmp #1 SMP Mon Feb 18 10:36:49 CST 2002 ppc unknown
More news later.