From owner-cvs-all  Thu Oct  8 22:07:52 1998
Return-Path: <owner-cvs-all>
Received: (from daemon@localhost)
          by hub.freebsd.org (8.8.8/8.8.8) id WAA22239
          for cvs-all-outgoing; Thu, 8 Oct 1998 22:07:52 -0700 (PDT)
          (envelope-from owner-cvs-all)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.15.68.22])
          by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id WAA22215
          for <committers@FreeBSD.ORG>; Thu, 8 Oct 1998 22:07:48 -0700 (PDT)
          (envelope-from bde@godzilla.zeta.org.au)
Received: (from bde@localhost)
	by godzilla.zeta.org.au (8.8.7/8.8.7) id PAA19770;
	Fri, 9 Oct 1998 15:07:39 +1000
Date: Fri, 9 Oct 1998 15:07:39 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199810090507.PAA19770@godzilla.zeta.org.au>
To: jkh@time.cdrom.com, wollman@khavrinen.lcs.mit.edu
Subject: Re: cvs commit: src/libexec/getty main.c
Cc: committers@FreeBSD.ORG
Sender: owner-cvs-all@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

>> System headers go first.  Header files should never include other
>> header files, except for certain system headers and as necessary to
>> meet formal Standards.
>
>Hmmm.  formal Standards.  When you figure out what those are,
>including the ones in the wake of our "cleanup" which have caused
>certain interfaces to no longer behave as described in their own man
>pages (in terms of what must be included as a prerequisite), kindly
>let me know. :-)

Which interaces are those?   Only a few of have been broken since I last
checked (on Aug 4).  They are libutil.h (surprise), kvm.h (ssize_t),
err.h (FILE * vs void *), resolv.h (redefining foo as __foo leaves no
prototype for foo), dialog.h (dialog_ftree() doesn't match reality),
and many kernel-only interfaces (the bugs are mostly in new man pages
that never matched realilty).

>> declaration of functions in heaer files.  This whole debacle
>> illustrates properties_read(), and non-Standard library functions
>> generally, should not take `FILE *' arguments.  It should be an fd, or
>
>A nice goal, but much less easy in practice.

Fairly easy for incremental changes.

Bruce

#!/bin/sh

# Check prototypes in man pages against reality.
# XXX scribbles in /tmp.

cd /usr/share/man
for i in `find . -name '*.gz' -links +0 | xargs zegrep -l '(\.Fn|#include)'`
do
	zcat $i |
	awk -v file=$i '
	BEGIN {
		print ".Dd"
		print ".Sh SYNOPSIS"
		ft = ""
		if (index(file, ".9.gz") != 0)
			print ".Fd #define KERNEL 1"
		printf "" > "/tmp/zFnlist"
	}
	/^\.Fd/ { print $0 }
	/^\.Ft/ { ft = $0 }
	/^(\.Fn.*"|\.Fn.* void)/ {
		print $2 > "/tmp/zFnlist"
		print ".Fd #undef " $2
		print ft
		ft = ""
		print $0
		# Print semicolon on a separate line to avoid too many args.
		print ";"
	}
	/^\.Fo/ {
		print $2 > "/tmp/zFnlist"
		print ".Fd #undef " $2
		print ft
		ft = ""
		print $0
		while (getline >= 0 && index($0, ".Fc") != 1)
			print $0
		print ".Fc"
		# Print semicolon on a separate line to avoid too many args.
		print ";"
	}
#	/^[ \t]*#include/ { print ".Fd " $1 " " $2 }
	END { print "int pedantic_pacificer;" }
	' |
	nroff -mandoc |
	sed -e 's/.//g' -e 's/^[ 	]*//' -e '/^$/d' -e '/^SYNOPSIS$/d' |
	tr '\012' '@' |
	sed -e 's/@@//g' |
	tr '@' '\012' > /tmp/z.c

	# Here -I /sys is for <scsi/something.h>, and -I /usr/include is
	# before that to avoid a bugfeature in gcc: warnings for `long long'
	# are apparently suppressed if long long is used in headers in
	# standard places, but not if it is used elsewhere.
	if cc -ansi -pedantic -Werror -I /tmp -I /usr/include -I /sys -S \
	    -o /dev/null /tmp/z.c >/dev/null 2>&1
	then
		expect=`sort < /tmp/zFnlist`
		sedcmd="s/.*\`\(.*\)'.*/\1/"
		got=`cc -Wredundant-decls -I /tmp -I /usr/include -I /sys -S \
		    -o /dev/null /tmp/z.c 2>&1 |
		    grep redundant |
		    sed $sedcmd |
		    sort`
		if test "$expect" != "$got"
		then
			echo expect \"$expect\", got \"$got\"
			echo $i bad
			mv /tmp/z.c /tmp/`basename $i .gz`.c
		fi
	else
		echo $i bad
		mv /tmp/z.c /tmp/`basename $i .gz`.c
	fi
done