FD_ZERO() on Linux with egcs-2.92.18

[Frank Pilhofer <fp at informatik dot uni-frankfurt dot de>]

  Hi,

 the FD_ZERO() macro does not compile on Linux 2.0.31 using the current
egcs snapshot. Here's a test program:

#include <sys/time.h>

int main ()
{
  fd_set set;
  FD_ZERO (&set);
}


 The FD_ZERO() macro resolves to

  __asm__ __volatile__("cld ; rep ; stosl" :"=m"
    (*(__kernel_fd_set *) (  &set  )) :"a" (0),
    "c" ((1024 / (8 * sizeof(unsigned long)) ) ),
    "D" ((__kernel_fd_set *) (  &set  )) :"cx","di")  ;

 and gcc then reports:

egcs-bug-2.cc: In function `int main()':
egcs-bug-2.cc:6: Invalid `asm' statement:
egcs-bug-2.cc:6: fixed or forbidden register 2 (cx) was spilled for class CREG.

 My egcs version is

rose:~/uni/mico/test > g++ -v               
Reading specs from
/local/pkg/egcs/19981101/lib/gcc-lib/i586-pc-linux-gnulibc1/egcs-2.92.18/specs
gcc version egcs-2.92.18 19981101 (gcc2 ss-980609 experimental)

 I read about the new register allocation routines. Does that mean that
Linux will need "fixincludes" until the Linux kernels are fixed?

	Frank

-- 
 + Frank Pilhofer                        fp@informatik.uni-frankfurt.de  +
 |                                      http://www.uni-frankfurt.de/~fp/ |
 +---- Life would be a very great deal less weird without you.  - DA ----+