Re: verify fails on ia64-hpux11.23 in libunistring-0.9.3

[Gary V. Vaughan]

Hi Paul,

Thanks for following up.

On Sun, May 01, 2011 at 10:52:48AM -0700, Paul Eggert wrote:
> On 05/01/11 10:15, Gary V. Vaughan wrote:
> > $ cat ,t.c
> > #include <stdint.h>
> > #include "intprops.h"
> > int main (int argc, char **argv) {
> >   printf("%lu\n", (unsigned long) ULONG_MAX);
> >   printf("%lu\n", (unsigned long) UINT_FAST8_MAX);
> >   return 0;
> > }
> > $ cc -I. -I../lib -o ,t ,t.c && ./,t
> > 4294967295
> > 4294967295
> > $ cc -I. -I../lib +DD64 -o ,t ,t.c && ./,t 
> > 18446744073709551615
> > 4294967295
> > 
> > I have no idea why or how that is happening though :(
> 
> Surely the '#include "intprops.h"' is unnecessary in the
> above example, since 'main' uses only symbols that are
> defined by stdint.h.  However, it may be that '#include
> <stdio.h>' is needed for proper operation.

Oh yes, sorry.  I was editing the same test file and didn't
re-check the dependencies.

> Could you
> try that test again, without intprops.h, but with
> stdio.h?

No difference; the outputs are identical with the change of
header file:

  $ pwd
  /opt/build/libunistring-0.9.3/tests
  $ cat ,t.c
  #include <stdint.h>
  #include <stdio.h>
  int main (int argc, char **argv) {
    printf("%lu\n", (unsigned long) ULONG_MAX);
    printf("%lu\n", (unsigned long) UINT_FAST8_MAX);
    return 0;
  }
  $ cc -I. -I../lib -o ,t ,t.c && ./,t
  4294967295
  4294967295
  $ cc -I. -I../lib +DD64 -o ,t ,t.c && ./,t 
  18446744073709551615
  4294967295

And using only the system headers:

  $ cc -o ,t ,t.c && ./,t 
  4294967295
  4294967295
  $ cc +DD64 -o ,t ,t.c && ./,t 
  18446744073709551615
  4294967295

> Which include file in HP-UX defines UINT_FAST8_MAX?

  $ find /usr/include -name '*.h' -exec fgrep -l UINT_FAST8_MAX {} \;
  /usr/include/stdint.h
  $ grep UINT_FAST8_MAX /usr/include/stdint.h
  #  define       UINT_FAST8_MAX          UINT_MAX
  $ find /usr/include -name '*.h' -exec fgrep -l UINT_MAX {} \; 
  /usr/include/limits.h
  /usr/include/stdint.h
  /usr/include/wchar.h
  $ grep UINT_MAX `find /usr/include -name '*.h' -exec fgrep -l UINT_MAX {} \;`
  /usr/include/limits.h:#    define UINT_MAX      4294967295U  /* max value of 
an unsigned integer */
  /usr/include/limits.h:#    define UINT_MAX      037777777777  /* max value of 
an unsigned integer */
  /usr/include/limits.h:#    define USI_MAX       UINT_MAX        /* max 
decimal value of an unsigned 
  /usr/include/stdint.h:#  define UINT_LEAST32_MAX        UINT_MAX
  /usr/include/stdint.h:#  define UINT_FAST8_MAX          UINT_MAX
  /usr/include/stdint.h:#  define UINT_FAST16_MAX         UINT_MAX
  /usr/include/stdint.h:#  define UINT_FAST32_MAX         UINT_MAX
  /usr/include/stdint.h:#  define SIG_ATOMIC_MAX  UINT_MAX
  /usr/include/stdint.h:#    define       WCHAR_MAX       UINT_MAX
  /usr/include/stdint.h:#    define       WINT_MAX        UINT_MAX
  /usr/include/wchar.h:#  include <limits.h>       /* UINT_MAX               */
  /usr/include/wchar.h:#               define WCHAR_MAX  UINT_MAX  /* max value 
of an unsigned integer */
  /usr/include/wchar.h:#               define WCHAR_MAX  (wchar_t)UINT_MAX  /* 
max value of an unsigned integer */

> Perhaps
> that include file is clashing with gnulib's stdint.h.  For
> example, it could be that <limits.h> (which intprops.h includes)
> redefines UINT_FAST8_MAX to the wrong value.

That seems to me to be what is happening.  Is this something that gnulib
stdint.h can fix?  It looks like a stupid mistake on the part of the
HPUX headers.  If you concur, I'll also report it back upstream so that
later compiler patch levels might be fixed in due course.

> Also, which include file in HP-UX defines uint_fast8_t,
> and what does it define it to?

  $ find /usr/include -name '*.h' -exec fgrep -l uint_fast8_t {} \;
  /usr/include/sys/_inttypes.h
  $ grep uint_fast8_t /usr/include/sys/_inttypes.h
  typedef unsigned int uint_fast8_t; 
  $ fgrep -l sys/_inttypes.h /usr/include/*.h
  /usr/include/crt0.h
  /usr/include/curses.h
  /usr/include/dial.h
  /usr/include/grp.h
  /usr/include/inttypes.h
  /usr/include/lc_core.h
  /usr/include/localedef.h
  /usr/include/model.h
  /usr/include/stddef.h
  /usr/include/stdint.h
  /usr/include/string.h
  /usr/include/term.h
  /usr/include/time.h
  /usr/include/utmp.h
  /usr/include/utmps.h
  /usr/include/utmpx.h
  /usr/include/values.h

Hmm... so the root of the problem is that with the 64bit ABI, ULONG_MAX
is promoted to 2**64-1, where UINT_FAST8_MAX is always 2**32-1 to match
unsigned int declaration of uint_fast8_t - irrespective of ABI.

And yet, ia64*-hpux11.31 shows the same results as above, save that I
need to add '#include <limits.h>' to get a ULONG_MAX definition, and
that the original bug I reported at the start of this thread isn't
triggered here!

Cheers,
-- 
Gary V. Vaughan (gary AT gnu DOT org)

pgptQdhHcSgps.pgp
Description: PGP signature