OS/2 Shareware BBS: 10 Tools/10-Tools.zip/perl560.zip/hints/aix.sh

Text File | 2000-03-05 | 12KB | 385 lines

# hints/aix.sh
# AIX 3.x.x hints thanks to Wayne Scott <[e-mail]>
# AIX 4.1 hints thanks to Christopher Chan-Nui <[e-mail]>.
# AIX 4.1 pthreading by Christopher Chan-Nui <[e-mail]> and
#      Jarkko Hietaniemi <[e-mail]>.
# AIX 4.3.x LP64 build by Steven Hirsch <[e-mail]>
# Merged on Mon Feb  6 10:22:35 EST 1995 by
#   Andy Dougherty  <[e-mail]>

#
# Contact [e-mail] for any of the following:
#
#    - AIX 43x and above support
#    - gcc + threads support
#    - socks support
#
# Apr 99 changes:
#
#    - use nm in AIX 43x and above
#    - gcc + threads now builds
#    [(added support for socks) Jul 99 SOCKS support rewritten]
#
# Notes:
#
#    - shared libperl support is tricky. if ever libperl.a ends up
#      in /usr/local/lib/* it can override any subsequent builds of
#      that same perl release. to make sure you know where the shared
#      libperl.a is coming from do a 'dump -Hv perl' and check all the
#      library search paths in the loader header.
#
#      it would be nice to warn the user if a libperl.a exists that is
#      going to override the current build, but that would be complex.
#
#      better yet, a solid fix for this situation should be developed.
#

# Configure finds setrgid and setruid, but they're useless.  The man
# pages state:
#    setrgid: The EPERM error code is always returned.
#    setruid: The EPERM error code is always returned. Processes cannot
#          reset only their real user IDs.
d_setrgid='undef'
d_setruid='undef'

alignbytes=8

case "$usemymalloc" in
'')  usemymalloc='n' ;;
esac

# Intuiting the existence of system calls under AIX is difficult,
# at best; the safest technique is to find them empirically.

# AIX 4.3.* and above default to using nm for symbol extraction
case "$osvers" in
   3.*|4.1.*|4.2.*)
      usenm='undef'
      ;;
   *)
      usenm='true'
      ;;
esac

so="a"
# AIX itself uses .o (libc.o) but we prefer compatibility
# with the rest of the world and with rest of the scripting
# languages (Tcl, Python) and related systems (SWIG).
# Stephanie Beals <[e-mail]>
dlext="so"

# Take possible hint from the environment.  If 32-bit is set in the
# environment, we can override it later.  If set for 64, the
# 'sizeof' test sees a native 64-bit architecture and never looks back.
case "$OBJECT_MODE" in
32)
    cat >&4 <<EOF

You have OBJECT_MODE=32 set in the environment.
I take this as a hint you do not want to
build for a 64-bit address space. You will be
given the opportunity to change this later.
EOF
    ;;
64)
    cat >&4 <<EOF

You have OBJECT_MODE=64 set in the environment.
This forces a full 64-bit build.  If that is
not what you intended, please terminate this
program, unset it and restart.
EOF
    ;;
*)  ;;
esac

# Trying to set this breaks the POSIX.c compilation

# Make setsockopt work correctly.  See man page.
# ccflags='-D_BSD=44'

# uname -m output is too specific and not appropriate here
case "$archname" in
'') archname="$osname" ;;
esac

cc=${cc:-cc}

case "$osvers" in
3*) d_fchmod=undef
    ccflags="$ccflags -D_ALL_SOURCE"
    ;;
*)  # These hints at least work for 4.x, possibly other systems too.
    ccflags="$ccflags -D_ALL_SOURCE -D_ANSI_C_SOURCE -D_POSIX_SOURCE"
    case "$cc" in
     *gcc*) ;;
     *) ccflags="$ccflags -qmaxmem=16384" ;;
    esac
    nm_opt='-B'
    ;;
esac

# These functions don't work like Perl expects them to.
d_setregid='undef'
d_setreuid='undef'

# Changes for dynamic linking by Wayne Scott <[e-mail]>
#
# Tell perl which symbols to export for dynamic linking.
case "$cc" in
*gcc*) ccdlflags='-Xlinker' ;;
esac
# the required -bE:$installarchlib/CORE/perl.exp is added by
# libperl.U (Configure) later.

case "$ldlibpthname" in
'') ldlibpthname=LIBPATH ;;
esac

# The first 3 options would not be needed if dynamic libs. could be linked
# with the compiler instead of ld.
# -bI:$(PERL_INC)/perl.exp  Read the exported symbols from the perl binary
# -bE:$(BASEEXT).exp        Export these symbols.  This file contains only one
#                symbol: boot_$(EXP)     can it be auto-generated?
case "$osvers" in
3*)
    lddlflags="$lddlflags -H512 -T512 -bhalt:4 -bM:SRE -bI:\$(PERL_INC)/perl.exp -bE:\$(BASEEXT).exp -e _nostart -lc"
    ;;
*)
    lddlflags="$lddlflags -bhalt:4 -bM:SRE -bI:\$(PERL_INC)/perl.exp -bE:\$(BASEEXT).exp -b noentry -lc"
    ;;
esac

# This script UU/usethreads.cbu will get 'called-back' by Configure
# after it has prompted the user for whether to use threads.
cat > UU/usethreads.cbu <<'EOCBU'
case "$usethreads" in
$define|true|[yY]*)
    ccflags="$ccflags -DNEED_PTHREAD_INIT"
    case "$cc" in
    gcc) ;;
    cc_r) ;;
    cc|xl[cC]_r)
        echo >&4 "Switching cc to cc_r because of POSIX threads."
        # xlc_r has been known to produce buggy code in AIX 4.3.2.
        # (e.g. pragma/overload core dumps)     Let's suspect xlC_r, too.
        # [e-mail]
        cc=cc_r
        ;;
    '')
        cc=cc_r
        ;;
    *)
        cat >&4 <<EOM
For pthreads you should use the AIX C compiler cc_r.
(now your compiler was set to '$cc')
Cannot continue, aborting.
EOM
        exit 1
        ;;
    esac

    # c_rify libswanted.
    set `echo X "$libswanted "| sed -e 's/ \([cC]\) / \1_r /g'`
    shift
    libswanted="$*"
    # c_rify lddlflags.
    set `echo X "$lddlflags "| sed -e 's/ \(-l[cC]\) / \1_r /g'`
    shift
    lddlflags="$*"

    # Insert pthreads to libswanted, before any libc or libC.
    set `echo X "$libswanted "| sed -e 's/ \([cC]\) / pthreads \1 /'`
    shift
    libswanted="$*"
    # Insert pthreads to lddlflags, before any libc or libC.
    set `echo X "$lddlflags " | sed -e 's/ \(-l[cC]\) / -lpthreads \1 /'`
    shift
    lddlflags="$*"

    ;;
esac
EOCBU

# This script UU/uselargefiles.cbu will get 'called-back' by Configure
# after it has prompted the user for whether to use large files.
cat > UU/uselargefiles.cbu <<'EOCBU'
case "$uselargefiles" in
''|$define|true|[yY]*)
    lfcflags="`getconf XBS5_ILP32_OFFBIG_CFLAGS 2>/dev/null`"
    lfldflags="`getconf XBS5_ILP32_OFFBIG_LDFLAGS 2>/dev/null`"
    # _Somehow_ in AIX 4.3.1.0 the above getconf call manages to
    # insert(?) *something* to $ldflags so that later (in Configure) evaluating
    # $ldflags causes a newline after the '-b64' (the result of the getconf).
    # (nothing strange shows up in $ldflags even in hexdump;
    #  so it may be something in the shell, instead?)
    # Try it out: just uncomment the below line and rerun Configure:
# echo >&4 "AIX 4.3.1.0 $lfldflags mystery" ; exit 1
    # Just don't ask me how AIX does it, I spent hours wondering.
    # Therefore the line re-evaluating lfldflags: it seems to fix
    # the whatever it was that AIX managed to break. --jhi
    lfldflags="`echo $lfldflags`"
    lflibs="`getconf XBS5_ILP32_OFFBIG_LIBS 2>/dev/null|sed -e 's@^-l@@' -e 's@ -l@ @g`"
    case "$lfcflags$lfldflags$lflibs" in
    '');;
    *) ccflags="$ccflags $lfcflags"
       ldflags="$ldflags $lfldflags"
       libswanted="$libswanted $lflibs"
       ;;
    esac
    lfcflags=''
    lfldflags=''
    lflibs=''
    ;;
esac
EOCBU

# This script UU/use64bitint.cbu will get 'called-back' by Configure
# after it has prompted the user for whether to use 64 bit integers.
cat > UU/use64bitint.cbu <<'EOCBU'
case "$use64bitint" in
$define|true|[yY]*)
        case "`oslevel`" in
        3.*|4.[012].*)
        cat >&4 <<EOM
AIX `oslevel` does not support 64-bit interfaces.
You should upgrade to at least AIX 4.3.
EOM
        exit 1
        ;;
        esac
        ;;
esac
EOCBU

cat > UU/use64bitall.cbu <<'EOCBU'
# This script UU/use64bitall.cbu will get 'called-back' by Configure
# after it has prompted the user for whether to be maximally 64-bitty.
case "$use64bitall" in
$define|true|[yY]*)
        case "`oslevel`" in
        3.*|4.[012].*)
        cat >&4 <<EOM
AIX `oslevel` does not support 64-bit interfaces.
You should upgrade to at least AIX 4.3.
EOM
        exit 1
        ;;
        esac
        echo " "
        echo "Checking the CPU width of your hardware..." >&4
        $cat >size.c <<EOCP
#include <stdio.h>
#include <sys/systemcfg.h>
int main (void)
{
  printf("%d\n",_system_configuration.width);
  return(0);
}
EOCP
        set size
        if eval $compile_ok; then
        lfcpuwidth=`./size`
        echo "You are running on $lfcpuwidth bit hardware."
        else
        dflt="32"
        echo " "
        echo "(I can't seem to compile the test program.  Guessing...)"
        rp="What is the width of your CPU (in bits)?"
        . ./myread
        lfcpuwidth="$ans"
        fi
        $rm -f size.c size
        case "$lfcpuwidth" in
        32*)
        cat >&4 <<EOM
Bzzzt! At present, you can only perform a
full 64-bit build on a 64-bit machine.
EOM
        exit 1
        ;;
        esac
        lfcflags="`getconf XBS5_LP64_OFF64_CFLAGS 2>/dev/null`"
        lfldflags="`getconf XBS5_LP64_OFF64_LDFLAGS 2>/dev/null`"
        # See jhi's comments above regarding this re-eval.  I've
        # seen similar weirdness in the form of:
        #
# 1506-173 (W) Option lm is not valid.  Enter xlc for list of valid options.
        #
        # error messages from 'cc -E' invocation. Again, the offending
        # string is simply not detectable by any means.  Since it doesn't
        # do any harm, I didn't pursue it. -- sh
        lfldflags="`echo $lfldflags`"
        lflibs="`getconf XBS5_LP64_OFF64_LIBS 2>/dev/null|sed -e 's@^-l@@' -e 's@ -l@ @g`"
        # -q32 and -b32 may have been set by uselargefiles or user.
            # Remove them.
        ccflags="`echo $ccflags | sed -e 's@-q32@@'`"
        ldflags="`echo $ldflags | sed -e 's@-b32@@'`"
        # Tell archiver to use large format.  Unless we remove 'ar'
        # from 'trylist', the Configure script will just reset it to 'ar'
        # immediately prior to writing config.sh.  This took me hours
        # to figure out.
        trylist="`echo $trylist | sed -e 's@^ar @@' -e 's@ ar @ @g' -e 's@ ar$@@'`"
        ar="ar -X64"
        nm_opt="-X64 $nm_opt"
        # Note: Placing the 'lfcflags' variable into the 'ldflags' string
        # is NOT a typo.  ldlflags is passed to the C compiler for final
        # linking, and it wants -q64 (-b64 is for ld only!).
        case "$lfcflags$lfldflags$lflibs" in
        '');;
        *) ccflags="$ccflags $lfcflags"
           ldflags="$ldflags $lfcflags"
           lddlflags="$lfldflags $lddlflags"
           libswanted="$libswanted $lflibs"
           ;;
        esac
        case "$ccflags" in
        *-DUSE_64_BIT_ALL*) ;;
              *) ccflags="$ccflags -DUSE_64_BIT_ALL";;
        esac
        case "$archname64" in
        ''|64*) archname64=64all ;;
        esac
        longsize="8"
        # Don't try backwards compatibility
        bincompat="$undef"
        d_bincompat5005="$undef"
        lfcflags=''
        lfldflags=''
        lflibs=''
        lfcpuwidth=''
        ;;
esac
EOCBU

cat > UU/uselongdouble.cbu <<'EOCBU'
# This script UU/uselongdouble.cbu will get 'called-back' by Configure
# after it has prompted the user for whether to use long doubles.
case "$uselongdouble" in
$define|true|[yY]*)
    ccflags="$ccflags -qlongdouble"
    # The explicit cc128, xlc128, xlC128 are not needed,
    # the -qlongdouble should do the trick. --jhi
    d_Gconvert='sprintf((b),"%.*llg",(n),(x))'
    ;;
esac
EOCBU

# If the C++ libraries, libC and libC_r, are available we will prefer them
# over the vanilla libc, because the libC contain loadAndInit() and
# terminateAndUnload() which work correctly with C++ statics while libc
# load() and unload() do not.  See ext/DynaLoader/dl_aix.xs.
# The C-to-C_r switch is done by usethreads.cbu, if needed.
if test -f /lib/libC.a -a X"`$cc -v 2>&1 | grep gcc`" = X; then
    # Cify libswanted.
    set `echo X "$libswanted "| sed -e 's/ c / C c /'`
    shift
    libswanted="$*"
    # Cify lddlflags.
    set `echo X "$lddlflags "| sed -e 's/ -lc / -lC -lc /'`
    shift
    lddlflags="$*"
fi

# EOF