Weird configure errors on HPUX 11.00

[Robert A Nesius <rnesius at ichips dot intel dot com>]

Platform: HPUX 11.00
Target: hppaw2.0w-hp-hpux11.00

Problem: During the 'gmake bootstrap' build fails because LC_ALL
is not defined.  LC_ALL is not defined because locale.h is
not being included.  locale.h is not being included because configure
doesn't think the system has it.  But, the system does have it.

Here's what confdefs.h, conftest.c, and conftest.out look like
just after configure has created conftest.c and compiled it with
output redirected into conftest.out.
===========<excerpted from configure output> =========================
# configure was hacked to dump the aforementioned files before deleting
# them when errors occurred.

checking for locale.h...
HEADER locale.h
confdefs.h is

#define HAVE_LONG_DOUBLE 1
#define HAVE_VOLATILE 1
#define TIME_WITH_SYS_TIME 1
#define HAVE_STDBOOL_H 1
#define STRING_WITH_STRINGS 1
#define HAVE_SYS_WAIT_H 1
#define HAVE_LIMITS_H 1
#define HAVE_STDDEF_H 1
#define HAVE_STRING_H 1
#define HAVE_STRINGS_H 1
#define HAVE_TIME_H 1
#define HAVE_FCNTL_H 1
#define HAVE_UNISTD_H 1
#define HAVE_STAB_H 1
#define HAVE_SYS_FILE_H 1
#define HAVE_SYS_TIME_H 1
#define HAVE_SYS_RESOURCE_H 1
#define HAVE_SYS_PARAM_H 1
#define HAVE_SYS_TIMES_H 1
#define HAVE_SYS_STAT_H 1
#define HAVE_MALLOC_H 1
#define HAVE_LANGINFO_H 1
#define HAVE_STRINGIZE 1
#define HAVE_INTTYPES_H 1
#define ENUM_BITFIELDS_ARE_UNSIGNED 1
#define HAVE_STRTOUL 1
#define HAVE_BSEARCH 1
#define HAVE_PUTENV 1
#define HAVE_POPEN 1
#define HAVE_BCOPY 1
#define HAVE_STRCHR 1
#define HAVE_STRRCHR 1
#define HAVE_KILL 1
#define HAVE_GETRLIMIT 1
#define HAVE_SETRLIMIT 1
#define HAVE_SYSCONF 1
#define HAVE_ISASCII 1
#define HAVE_GETTIMEOFDAY 1
#define HAVE_PUTC_UNLOCKED 1
#define HAVE_GETRUSAGE 1
#define HAVE_ICONV 1
#define HAVE_NL_LANGINFO 1
#define HAVE_LSTAT 1
#define GETGROUPS_T gid_t
#define HAVE_VPRINTF 1
#define HAVE_STRSTR 1
#define HAVE_PRINTF_PTR 1
#define HAVE_GETPAGESIZE 1
#define HAVE_MMAP_ANON 1
#define HAVE_MMAP_FILE 1
#define HAVE_DECL_BCOPY 1
#define HAVE_DECL_GETENV 0
#define HAVE_DECL_ATOL 0
#define HAVE_DECL_SBRK 1
#define HAVE_DECL_ABORT 1
#define HAVE_DECL_ATOF 0
#define HAVE_DECL_GETCWD 1
#define HAVE_DECL_GETWD 1
#define HAVE_DECL_STRSIGNAL 1
#define HAVE_DECL_PUTC_UNLOCKED 0
#define HAVE_DECL_FPUTS_UNLOCKED 0
#define HAVE_DECL_STRSTR 1
#define HAVE_DECL_ENVIRON 1
#define HAVE_DECL_MALLOC 1
#define HAVE_DECL_REALLOC 1
#define HAVE_DECL_CALLOC 1
#define HAVE_DECL_FREE 1
#define HAVE_DECL_BASENAME 0
#define HAVE_DECL_GETOPT 1
#define HAVE_DECL_GETRLIMIT 1
#define HAVE_DECL_SETRLIMIT 1
#define HAVE_DECL_GETRUSAGE 1
#define PACKAGE "gcc"
#define VERSION "3.0"
#define HAVE_ALLOCA_H 1
#define HAVE_ALLOCA 1
#define HAVE_LIMITS_H 1
conftest.c is
#line 4724 "configure"
#include "confdefs.h"
#include <locale.h>
conftest.out is
In file included from /usr/include/locale.h:157,
                 from configure:4725:
/opt/gcc/lib/gcc-lib/hppa2.0n-hp-hpux11.00/2.95.3/include/stdio.h:30: warning: `
__va__list' redefined
/usr/include/locale.h:26: warning: this is the location of the previous definiti
on
no
==========================

I'm using the pre-built gcc 2.95.3 from utah.cs.edu..  Interestingly,
the gcc 2.95.2 doesn't cause these warnings to be generated and
the configure succeeds (as does the build).

So - moral of the story is use the 2.95.2 gcc?  (Let me guess,
that's been said already?)  It seems configure is considering any
output emitted as a warning, etc.. as a failure mode?

-Rob


-- 
#include <sig.h>
------------------------------------------------------------------
Robert Nesius             rnesius@ichips.intel.com    503.712.2181
DPG Engineering Computing SW Applications Team