Re: [PATCH] Detecting bad /dev/null's in Configure

[metaconfig.git] / U / modified / Head.U

?RCS: $Id: Head.U,v 3.0.1.9 1997/02/28 15:02:09 ram Exp $
?RCS:
?RCS: Copyright (c) 1991-1993, Raphael Manfredi
?RCS: 
?RCS: You may redistribute only under the terms of the Artistic Licence,
?RCS: as specified in the README file that comes with the distribution.
?RCS: You may reuse parts of this distribution only within the terms of
?RCS: that same Artistic Licence; a copy of which may be found at the root
?RCS: of the source tree for dist 3.0.
?RCS:
?RCS: $Log: Head.U,v $
?RCS: Revision 3.0.1.9  1997/02/28  15:02:09  ram
?RCS: patch61: make sure we unset CDPATH for shells that support this
?RCS: patch61: improved Korn shell detection and handling
?RCS:
?RCS: Revision 3.0.1.8  1995/07/25  13:40:02  ram
?RCS: patch56: added SVR4-ish /opt directories to path list (ADO)
?RCS: patch56: OS/2 platforms are using another path separator
?RCS:
?RCS: Revision 3.0.1.7  1995/03/21  08:46:15  ram
?RCS: patch52: definition of paths wrongly added spurious ':' chars
?RCS:
?RCS: Revision 3.0.1.6  1994/10/29  15:54:19  ram
?RCS: patch36: make sure ENV is unset before calling /bin/ksh
?RCS:
?RCS: Revision 3.0.1.5  1994/08/29  16:03:44  ram
?RCS: patch32: now sets PATH only using existing directories
?RCS:
?RCS: Revision 3.0.1.4  1994/06/20  06:54:28  ram
?RCS: patch30: now computes its invocation name into 'me'
?RCS: patch30: symbol me is made visible to all units read-only
?RCS:
?RCS: Revision 3.0.1.3  1993/12/15  08:15:07  ram
?RCS: patch15: added /sbin:/usr/sbin:/usr/libexec in PATH for BSD/386
?RCS:
?RCS: Revision 3.0.1.2  1993/11/10  17:32:35  ram
?RCS: patch14: ensure PATH is reset to '.' before testing for alias
?RCS:
?RCS: Revision 3.0.1.1  1993/08/27  14:38:07  ram
?RCS: patch7: not all 'test' programs support the -x option
?RCS:
?RCS: Revision 3.0  1993/08/18  12:04:58  ram
?RCS: Baseline for dist 3.0 netwide release.
?RCS:
?X: 
?X: This is the very first unit in the Configure script.  It is mostly just
?X: things to keep people from getting into a tizzy right off the bat.
?X: 
?MAKE:Head:
?MAKE:	-pick wipe $@ %<
?V:PATH p_ me
?T:argv Id p paths OS2_SHELL DJGPP
?T:inksh needksh avoidksh newsh changesh reason
?LINT:extern ENV CDPATH SHELL MACHTYPE _exe is_os2
?LINT:change ENV CDPATH _exe is_os2
#! /bin/sh
#
# If these # comments don't work, trim them. Don't worry about any other
# shell scripts, Configure will trim # comments from them for you.
#
# (If you are trying to port this package to a machine without sh,
# I would suggest you have a look at the prototypical config_h.SH file
# and edit it to reflect your system. Some packages may include samples
# of config.h for certain machines, so you might look for one of those.)
#
?X:
?X: NOTE THAT A CONFIGURE SCRIPT IS IN THE PUBLIC DOMAIN (whether or not
?X: the software which uses it is in the public domain).
?X:
# Yes, you may rip this off to use in other distribution packages. This
# script belongs to the public domain and cannot be copyrighted.
#
?X:
?X: WE ASK YOU NOT TO REMOVE OR ALTER THE FOLLOWING PARAGRAPH, PLEASE:
?X:
# (Note: this Configure script was generated automatically. Rather than
# working with this copy of Configure, you may wish to get metaconfig.
# The dist-3.0 package (which contains metaconfig) was posted in
# comp.sources.misc and is available on CPAN under authors/id/RAM so
# you may fetch it yourself from your nearest archive site.)
#
?X: OK, I'll not alter, but I'll add to it then ...
#
# Though this script was generated by metaconfig, it is OK to send
# patches against it. It's up to the Configure pumpkin to backport
# the patch to the metaunits if it is accepted.
# See Porting/pumpkin.pod for more information on metaconfig.
#
?X:
?X: NOTA BENE:
?X: If you develop you own version of metaconfig based on this work,
?X: you have to add some comments telling that the script was generated
?X: by your version, not mine: It credits your work.
?X:

# $Id: Head.U,v 3.0.1.9 1997/02/28 15:02:09 ram Exp $
#
# Generated on <DATE> [metaconfig <VERSION> PL<PATCHLEVEL>]
# (with additional metaconfig patches by <MAINTLOC>)

cat >c1$$ <<EOF
ARGGGHHHH!!!!!

SCO csh still thinks true is false.  Write to SCO today and tell them that next
year Configure ought to "rm /bin/csh" unless they fix their blasted shell. :-)

(Actually, Configure ought to just patch csh in place.  Hmm.  Hmmmmm.  All
we'd have to do is go in and swap the && and || tokens, wherever they are.)

[End of diatribe. We now return you to your regularly scheduled programming...]
EOF
cat >c2$$ <<EOF

OOPS!  You naughty creature!  You didn't run Configure with sh!
I will attempt to remedy the situation by running sh for you...
EOF

true || cat c1$$ c2$$
true || exec sh $0 $argv:q

(exit $?0) || cat c2$$
(exit $?0) || exec sh $0 $argv:q
rm -f c1$$ c2$$

if test -f /dev/cputype -a -f /dev/drivers -a -f /dev/osversion; then
	cat >&4 <<EOF
***
*** I'm sorry but this system looks like Plan 9 and Plan 9 doesn't do
*** Configure that well.  (Plan 9 is close to UNIX but not close enough.)
*** Please read the README.plan9 for further instructions.
*** Cannot continue, aborting.
***
EOF
	exit 1
fi

if test ! -c /dev/null ; then
	cat >&4 <<EOF
***
*** I'm sorry, but /dev/null appears to be a file rather than a device.
*** Please consult your operating sytem's notes for making a device
*** in /dev.
*** Cannot continue, aborting.
***
EOF
	exit 1
fi

: compute my invocation name
me=$0
case "$0" in
*/*)
	me=`echo $0 | sed -e 's!.*/\(.*\)!\1!' 2>/dev/null`
	test "$me" || me=$0
	;;
esac

?X:
?X: To be able to run under OS/2, we must detect that early enough to use
?X: the proper path separator, stored in $p_. It is : on UNIX and ; on
?X: DOSish systems such as OS/2.
?X:
: Proper separator for the PATH environment variable
p_=:
: On OS/2 this directory should exist if this is not floppy only system :-]
if test -d c:/. || ( uname -a | grep -i 'os\(/\|\)2' ) 2>&1 >/dev/null ; then
    if test -n "$OS2_SHELL"; then
		p_=\;
		PATH=`cmd /c "echo %PATH%" | tr '\\\\' / `
?X: That's a bug in ksh5.22
		OS2_SHELL=`cmd /c "echo %OS2_SHELL%" | tr '\\\\' / | tr '[A-Z]' '[a-z]'`
		is_os2=yes
	elif test -n "$DJGPP"; then
		case "X${MACHTYPE:-nonesuchmach}" in
		*cygwin) ;;
		*) p_=\; ;;
		esac
	fi
fi

?X:
?X: There are two schools of thoughts here. Some people correctly argue that
?X: the user has a better chance than we do of setting a reasonable PATH and
?X: others argue that Configure is the best place there is to set up a suitable
?X: PATH. Well, here we try to compromize by keeping the user's PATH and
?X: appending some directories which are known to work on some machine or the
?X: other. The rationale behind this being that a novice user might not have a
?X: proper environment variable set, and some directories like /etc (where
?X: chown is located on some BSD systems) may be missing--RAM.
?X:
?X: SVR4 adds an /opt directory for optional packages.  Some sites use
?X: various permutations on /opt as opposed to /usr or /usr/local.-- ADO
?X:
?X: We only add directories that are not already in the PATH of the
?X: user and the directories must exist also.
?X:
?X: 20040627, Merijn, HP-UX's ANSI compiler is in /opt/ansic/bin if present
?X:  and should be before /usr/ccs/bin, where the braindead bundled compiler
?X:  can be found. The .usr/ccs/bin compiler cannot be used to build perl.
?X:
: Proper PATH setting
paths='/bin /usr/bin /usr/local/bin /usr/ucb /usr/local /usr/lbin'
paths="$paths /opt/bin /opt/local/bin /opt/local /opt/lbin"
paths="$paths /usr/5bin /etc /usr/gnu/bin /usr/new /usr/new/bin /usr/nbin"
paths="$paths /opt/gnu/bin /opt/new /opt/new/bin /opt/nbin"
paths="$paths /sys5.3/bin /sys5.3/usr/bin /bsd4.3/bin /bsd4.3/usr/ucb"
paths="$paths /bsd4.3/usr/bin /usr/bsd /bsd43/bin /opt/ansic/bin /usr/ccs/bin"
paths="$paths /etc /usr/lib /usr/ucblib /lib /usr/ccs/lib"
paths="$paths /sbin /usr/sbin /usr/libexec"
paths="$paths /system/gnu_library/bin"

for p in $paths
do
	case "$p_$PATH$p_" in
	*$p_$p$p_*) ;;
	*) test -d $p && PATH=$PATH$p_$p ;;
	esac
done

PATH=.$p_$PATH
export PATH

: shall we be using ksh?
inksh=''
needksh=''
avoidksh=''
newsh=/bin/ksh
changesh=''
?X: Use (alias -x) and not (alias) since zsh and bash recognize the alias
?X: builtin but not the -x option which is typically ksh...
?X: We need to set up PATH before calling the "alias" built-in since some
?X: systems like HP-UX have a binary called /bin/alias.
if (PATH=.; alias -x) >/dev/null 2>&1; then
		inksh=true
fi
?X: On HP-UX, large Configure scripts may exercise a bug in /bin/sh, use ksh
if test -f /hp-ux -a -f /bin/ksh; then
	needksh='to avoid sh bug in "here document" expansion'
fi
?X: On AIX4, /bin/sh is really ksh and it causes problems, use sh
if test -d /usr/lpp -a -f /usr/bin/bsh -a -f /usr/bin/uname; then
	if test X`/usr/bin/uname -v` = X4; then
		avoidksh="to avoid AIX 4's /bin/sh"
		newsh=/usr/bin/bsh
	fi
fi
?X: On Digital UNIX, /bin/sh may start up buggy /bin/ksh, use sh
if test -f /osf_boot -a -f /usr/sbin/setld; then
	if test X`/usr/bin/uname -s` = XOSF1; then
	        avoidksh="to avoid Digital UNIX' ksh"
	        newsh=/bin/sh
	        unset BIN_SH # if this is 'xpg4' sh will start up ksh
	fi
fi
?X: If we are not in ksh and need it, then feed us back to it
case "$inksh/$needksh" in
/[a-z]*)
?X: Unset ENV to avoid any ~/.kshrc that could alias cd or whatever...
		ENV=''
		changesh=true
		reason="$needksh"
	;;
esac
?X: If we are in ksh and must avoid it, then feed us back to a new shell
case "$inksh/$avoidksh" in
true/[a-z]*)
	changesh=true
	reason="$avoidksh"
	;;
esac
?X: Warn them if they use ksh on other systems, which are those where
?X: we don't need ksh nor want to avoid it explicitely, yet are using it.
case "$inksh/$needksh-$avoidksh-" in
true/--)
		cat <<EOM
(I see you are using the Korn shell.  Some ksh's blow up on $me,
mainly on older exotic systems.  If yours does, try the Bourne shell instead.)
EOM
	;;
esac
case "$changesh" in
true)
	export newsh
	echo "(Feeding myself to $newsh $reason.)"
?X: Make sure they didn't say sh <Configure by checking whether $0 ends
?X: with Configure or not. If they did say sh <../../Configure, then too
?X: bad for them anyway, since we lost that path indication...
?X: Otherwise, execing $0 ensures we keep the full remote source dir
?X: indication for src.U.
	case "$0" in
	Configure|*/Configure) exec $newsh $0 "$@";;
	*) exec $newsh Configure "$@";;
	esac
	;;
esac

?X: Unset CDPATH to avoid surprises when using cd under some shells
?X: Can't unset it because that's not portable to very old shells.
?X: Can't set it to '' because then bash 2.02 won't do "cd UU"  --AD  6/98.
?X: Don't want to set it to '.' because then ksh prints out the
?X: name of the directory every time you cd to it. --AD 6/98
?X: In order to inflict the least harm, change it only if it's set.
: if needed set CDPATH to a harmless value that is not chatty
: avoid bash 2.02 problems with empty CDPATH.
case "$CDPATH" in
'')	;;
*)	case "$SHELL" in
	*bash*)	CDPATH='.' ;;
	*)		CDPATH='' ;;
	esac
	;;
esac
: Configure runs within the UU subdirectory
test -d UU || mkdir UU
?X: Use ./* to avoid any confirmation prompts from enhanced shells -- WED
cd UU && rm -f ./*