Martin Blapp wrote:
> If we add this patch, we should also add the bits in libm:
>
> #ifndef HAVE_SQRTL
> long double
> sqrtl(long double x)
> {
> return sqrt((double) x);
> }
> #endif
[..]
> or we just compile stubs.c in and the gcc provided bits.
> But I guess we would also not to link against libm then.
>
> root@fuchur:/usr/src# diff -ruN gnu/lib/libstdc++/Makefile.orig
> gnu/lib/libstdc++/Makefile
> --- gnu/lib/libstdc++/Makefile.orig Thu May 30 23:35:38 2002
> +++ gnu/lib/libstdc++/Makefile Thu May 30 23:32:19 2002
> @@ -32,7 +32,7 @@
> valarray-inst.cc ext-inst.cc
>
> # C parts of math
> -SRCS+= nan.c signbit.c signbitf.c signbitl.c # stubs.c
> +SRCS+= nan.c signbit.c signbitf.c signbitl.c stubs.c
>
> # Embedded copy of libsupc++
> SRCS+= del_op.cc del_opnt.cc del_opv.cc del_opvnt.cc \
>
> But now it's working ...
You know, I think I'd rather activate the stubs.c file than patch libm.
If libm is going to provide real "long double" functions, then it should
provide real ones, not faked lower precision ones.
stubs.c fills in the blanks for the parts that libm doesn't provide,
according to config.h. We can even add ifdefs to the config.h file if we
have some arches that provide the long double versions and others that do
not.
To be sure we're all on the same page, simply uncommenting stubs.c solves
your problem with no libm changes, right?
Cheers,
-Peter
--
Peter Wemm - [EMAIL PROTECTED]; [EMAIL PROTECTED]; [EMAIL PROTECTED]
"All of this is for nothing if we don't go to the stars" - JMS/B5
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message
