glibc-2.1 on PowerPC cannot disable floating-point

[David Edelsohn <dje@watson.ibm.com>]

	I still am learning about the structure of the glibc source
distribution and how the configure process, build process and directory
structure interact.  I did not notice the following problem until this
week: Unlike other architectures the PowerPC configuration does not have a
separate "fpu" directory for FPU functionality.  Because of this merged
structure, the glibc configuration option to disable floating-point has no
effect.  The PowerPC assembly functions explicitly manipulate FPU
registers which fail if the PowerPC FPU unit is not present or disabled,
so compiling with software floating point emulation does not work around
the problem (and in fact fails because GCC inlined assembly manipulates a
forbidden register).

	Was there a reason that the PowerPC port did not utilize the "fpu"
structure of other ports (e.g., alpha, arm, i386, m68k, and sparc)?  Can
the floating-point functionality be moved into an "fpu" subdirectory so
that it follows the structure of other ports and can be disabled?

Thanks, David
===============================================================================
David Edelsohn                                      T.J. Watson Research Center
dje@watson.ibm.com                                  P.O. Box 218
+1 914 945 4364 (TL 862)                            Yorktown Heights, NY 10598