Re: m4-1.4.9b on HP-UX 11

[Gary V. Vaughan]

On 2 Jun 2007, at 02:31, Bruno Haible wrote:
> Gary V. Vaughan reported in
>   http://article.gmane.org/gmane.comp.lib.gnulib.bugs/10446
> that on HP-UX 11.23:
>
> > ia64-hp-hpux11.23: 2 gnulib test failures
> > hppa2.0w-hp-hpux11.23: 2 gnulib test failures
> >
> >   gmake  check-TESTS
> >   gmake[4]: Entering directory `/opt/build/m4-1.4.9b/tests'
> >   test-lseek.c:76: assertion failed
> >   ./test-lseek.sh[14]: 11912 Abort(coredump)
> >   FAIL: test-lseek.sh
> >   test-vasprintf-posix.c:1238: assertion failed
> >   /opt/fsw/bash30/bin/bash: line 1: 11917 ABORT instruction        
> > (core dumped) EXEEXT='' EXEEXT='' EXEEXT='' srcdir='.' EXEEXT=''  
> > srcdir='.' EXEEXT='' srcdir='.' ${dir}$tst
> >   FAIL: test-vasprintf-posix
>
> The test-lseek.sh failure was already fixed on 2007-05-30.

ACK.

> The test-vasprintf-posix failure I cannot reproduce with CC="cc - 
> Ae" on
> the machine td176.testdrive.hp.com (an ia64-hp-hpux11.23 machine). Can
> you investigate what is the result of printf %Lf NaN on your machine?

It seems to be a result of the flags we're using.  All our *-hpux11.23
builds use:

  CC=cc CFLAGS="-z +O2 +Onofltacc +Olit=all +Oentrysched  
+Odataprefetch \
  +Onolimit" LDFLAGS="-Wl,+nodefaultrpath"


> 1) With the native printf, like this:
> ==================== foo.c ===================
> #include <stdio.h>
> int main ()
> {
>   static long double zero = 0.0L;
>   char result[100];
>   sprintf (result, "%Lf", zero / zero);
>   printf ("%s\n", result);
>   return 0;
> }

$ cc -z +O2 +Onofltacc +Olit=all +Oentrysched +Odataprefetch +Onolimit \
> foo.c -Wl,+nodefaultrpath -o foo
$ ./foo
1.000000
$ cc -Ae foo.c -o foo
$ ./foo
-nan

It seems +Onofltacc is to blame on hpux11.23, but we've made it a  
standard
flag for compilations on hpux in order for other FLOSS projects to  
build,
correctly; and, in fact, on hpux11.11, hpux11.00 and hpux10.20  
+Onofltacc
is passed to the compiler, and it doesn't cause the test to fail.   
From the
manual page:

  default
    Allows contractions, such as fused
    multiply-add (FMA), but disallows any
    other floating point optimization that
    can result in numerical differences.

  limited
    Like default, but also allows floating
    point optimizations which may affect the
    generation and propagation of
    infinities, NaNs, and the sign of zero.

  relaxed
    In addition to the optimizations allowed
    by limited, permits optimizations, such
    as reordering of expressions, even if
    parenthesized, that may affect rounding
    error.  This is the same as +Onofltacc.

Perhaps that just means there is a compiler bug with +Onofltacc on  
hpux11.23?

> 2) With the gnulib vasprintf, by inserting an fprintf(stderr,...)  
> statement
> near test-vasprintf-posix.c:1238 ?

With CC=cc CFLAGS="-z +O2 +Onofltacc +Olit=all +Oentrysched  
+Odataprefetch \
  +Onolimit" LDFLAGS="-Wl,+nodefaultrpath":

  1.000000 33
  test-vasprintf-posix.c:1239: assertion failed
  /opt/fsw/bash30/bin/bash: line 1: 12683 ABORT instruction        
(core dumped) EXEEXT='' EXEEXT='' EXEEXT='' srcdir='.' EXEEXT=''  
srcdir='.' EXEEXT='' srcdir='.' ${dir}$tst
  FAIL: test-vasprintf-posix

With CC=cc CFLAGS="-z +O2 +Olit=all +Oentrysched +Odataprefetch  
+Onolimit" \
LDFLAGS="-Wl,+nodefaultrpath":

  -nan 33
  -nan 33
  PASS: test-vasprintf-posix

Do you think the test is at fault here?  Or is it an hpux compiler  
bug?  If
the latter, can it be worked around?

Cheers,
        Gary
--
  ())_.              Email me: address@hidden
  ( '/           Read my blog: http://blog.azazil.net
  / )=         ...and my book: http://sources.redhat.com/autobook
`(_~)_ Join my AGLOCO Network: http://www.agloco.com/r/BBBS7912

PGP.sig
Description: This is a digitally signed message part