Re: Failing of test 47: semantics.at:129 on HP-UX 10.20

[Akim Demaille]

>>>>> "Peter" == Peter Ohlerich <address@hidden> writes:

Peter> Okay, here they are (I stopped the script as I saw that it went
Peter> into a loop and always wrote the same things into the logs)....

Thanks!

Peter, could you send some form of ID of this C compiler?

People (autoconf@),

we are in trouble: this compiler has the `bug' which Pavel once
reported:

configure:2866: cc -c -Ae -O +Onolimit +DAportable  conftest.c >&5
cc: "configure", line 2885: error 1502: Array size must be a constant expression
.
configure:2869: $? = 1
configure: failed program was:
#line 2854 "configure"
#include "confdefs.h"
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#if STDC_HEADERS
# include <stdlib.h>
# include <stddef.h>
#else
# if HAVE_STDLIB_H
#  include <stdlib.h>
# endif
#endif
#if HAVE_STRING_H
# if !STDC_HEADERS && HAVE_MEMORY_H
#  include <memory.h>
# endif
# include <string.h>
#else
# if HAVE_STRINGS_H
#  include <strings.h>
# endif
#endif
#if HAVE_INTTYPES_H
# include <inttypes.h>
#endif
#if HAVE_UNISTD_H
# include <unistd.h>
#endif
int
main ()
{
int _array_ [1 - 2 * !((sizeof (unsigned char)) >= 0)]
  ;
  return 0;
}



It is not too bad a piece of news.  First of all, this code is
triggered only for cross-compilers, and this compiler certainly
isn't.  It's only the test suite that forces the cross-compiling
strategy to be tried even on regular cases.

What shall we do?

It's very easy to set a guard against such compilers.  But once we
detected the compiler cannot evaluate such expressions, what do we do?