Re: [cygnus.egcs.bugs] Re: config.guess fails on HP-UX

[Ben Elliston <bje at cygnus dot com>]

>   > This change looks safe enough to make, so I've applied it to the official
>   > version of config.guess.  Thanks for the report,

> This is just to determine if the compiler works, right?  I just want to make
> sure we're not losing important flags which could change the behavior of
> certain tests.

config.guess contains a heredoc which is used solely on HP-UX to determine
the CPU type and kernel revision being used:

             sed 's/^              //' << EOF >$dummy.c
              #include <stdlib.h>
              #include <unistd.h>

              int main ()
              {
              #if defined(_SC_KERNEL_BITS)
                  long bits = sysconf(_SC_KERNEL_BITS);
              #endif
                  long cpu  = sysconf (_SC_CPU_VERSION);

                  switch (cpu)
                {
                case CPU_PA_RISC1_0: puts ("hppa1.0"); break;
                case CPU_PA_RISC1_1: puts ("hppa1.1"); break;
                case CPU_PA_RISC2_0:
              #if defined(_SC_KERNEL_BITS)
                    switch (bits)
                        {
                        case 64: puts ("hppa2.0w"); break;
                        case 32: puts ("hppa2.0n"); break;
                        default: puts ("hppa2.0"); break;
                        } break;
              #else  /* !defined(_SC_KERNEL_BITS) */
                    puts ("hppa2.0"); break;
              #endif
                default: puts ("hppa1.0"); break;
                }
                  exit (0);
              }
EOF

It does not affect how other test fragments (like, say, AC_C_CONST) are
compiled.

Ben