path: root/automake.in

#!@PERL@ -w
# -*- perl -*-
# @configure_input@

eval 'case $# in 0) exec @PERL@ -S "$0";; *) exec @PERL@ -S "$0" "$@";; esac'
    if 0;

# automake - create Makefile.in from Makefile.am
# Copyright (C) 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003
# Free Software Foundation, Inc.

# This program is free software; you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation; either version 2, or (at your option)
# any later version.

# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
# GNU General Public License for more details.

# You should have received a copy of the GNU General Public License
# along with this program; if not, write to the Free Software
# Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA
# 02111-1307, USA.

# Originally written by David Mackenzie <djm@gnu.ai.mit.edu>.
# Perl reimplementation by Tom Tromey <tromey@redhat.com>.

package Language;

BEGIN
{
  my $perllibdir = $ENV{'perllibdir'} || '@datadir@/@PACKAGE@-@APIVERSION@';
  unshift @INC, $perllibdir;

  # Override SHELL.  This is required on DJGPP so that system() uses
  # bash, not COMMAND.COM which doesn't quote arguments properly.
  # Other systems aren't expected to use $SHELL when Automake
  # runs, but it should be safe to drop the `if DJGPP' guard if
  # it turns up other systems need the same thing.  After all,
  # if SHELL is used, ./configure's SHELL is always better than
  # the user's SHELL (which may be something like tcsh).
  $ENV{'SHELL'} = '@SHELL@' if exists $ENV{'DJGPP'};
}

use Automake::Struct;
struct (# Short name of the language (c, f77...).
        'name' => "\$",
        # Nice name of the language (C, Fortran 77...).
        'Name' => "\$",

	# List of configure variables which must be defined.
	'config_vars' => '@',

        'ansi'    => "\$",
	# `pure' is `1' or `'.  A `pure' language is one where, if
	# all the files in a directory are of that language, then we
	# do not require the C compiler or any code to call it.
	'pure'   => "\$",

	'autodep' => "\$",

	# Name of the compiling variable (COMPILE).
        'compiler'  => "\$",
        # Content of the compiling variable.
        'compile'  => "\$",
        # Flag to require compilation without linking (-c).
        'compile_flag' => "\$",
        'extensions' => '@',
	# A subroutine to compute a list of possible extensions of
	# the product given the input extensions.
	# (defaults to a subroutine which returns ('.$(OBJEXT)', '.lo'))
	'output_extensions' => "\$",
	# A list of flag variables used in 'compile'.
	# (defaults to [])
        'flags' => "@",

	# The file to use when generating rules for this language.
	# The default is 'depend2'.
	'rule_file' => "\$",

        # Name of the linking variable (LINK).
        'linker' => "\$",
        # Content of the linking variable.
        'link' => "\$",

        # Name of the linker variable (LD).
        'lder' => "\$",
        # Content of the linker variable ($(CC)).
        'ld' => "\$",

        # Flag to specify the output file (-o).
        'output_flag' => "\$",
        '_finish' => "\$",

	# This is a subroutine which is called whenever we finally
	# determine the context in which a source file will be
	# compiled.
	'_target_hook' => "\$");


sub finish ($)
{
  my ($self) = @_;
  if (defined $self->_finish)
    {
      &{$self->_finish} ();
    }
}

sub target_hook ($$$$)
{
    my ($self) = @_;
    if (defined $self->_target_hook)
    {
	&{$self->_target_hook} (@_);
    }
}

package Automake;

use strict 'vars', 'subs';
use Automake::General;
use Automake::XFile;
use Automake::Channels;
use File::Basename;
use Carp;

## ----------- ##
## Constants.  ##
## ----------- ##

# Parameters set by configure.  Not to be changed.  NOTE: assign
# VERSION as string so that eg version 0.30 will print correctly.
my $VERSION = '@VERSION@';
my $PACKAGE = '@PACKAGE@';
my $libdir = '@datadir@/@PACKAGE@-@APIVERSION@';

# Some regular expressions.  One reason to put them here is that it
# makes indentation work better in Emacs.

# Writting singled-quoted-$-terminated regexes is a pain because
# perl-mode thinks of $' as the ${'} variable (intead of a $ followed
# by a closing quote.  Letting perl-mode think the quote is not closed
# leads to all sort of misindentations.  On the other hand, defining
# regexes as double-quoted strings is far less readable.  So usually
# we will write:
#
#  $REGEX = '^regex_value' . "\$";

my $IGNORE_PATTERN = '^\s*##([^#\n].*)?\n';
my $WHITE_PATTERN = '^\s*' . "\$";
my $COMMENT_PATTERN = '^#';
my $TARGET_PATTERN='[$a-zA-Z_.@%][-.a-zA-Z0-9_(){}/$+@%]*';
# A rule has three parts: a list of targets, a list of dependencies,
# and optionally actions.
my $RULE_PATTERN =
  "^($TARGET_PATTERN(?:(?:\\\\\n|\\s)+$TARGET_PATTERN)*) *:([^=].*|)\$";

my $SUFFIX_RULE_PATTERN =
    '^(\.[a-zA-Z0-9_(){}$+@]+)(\.[a-zA-Z0-9_(){}$+@]+)' . "\$";
# Only recognize leading spaces, not leading tabs.  If we recognize
# leading tabs here then we need to make the reader smarter, because
# otherwise it will think rules like `foo=bar; \' are errors.
my $MACRO_PATTERN = '^[.A-Za-z0-9_@]+' . "\$";
my $ASSIGNMENT_PATTERN = '^ *([^ \t=:+]*)\s*([:+]?)=\s*(.*)' . "\$";
# This pattern recognizes a Gnits version id and sets $1 if the
# release is an alpha release.  We also allow a suffix which can be
# used to extend the version number with a "fork" identifier.
my $GNITS_VERSION_PATTERN = '\d+\.\d+([a-z]|\.\d+)?(-[A-Za-z0-9]+)?';

my $IF_PATTERN = '^if\s+(!?)\s*([A-Za-z][A-Za-z0-9_]*)\s*(?:#.*)?' . "\$";
my $ELSE_PATTERN =
  '^else(?:\s+(!?)\s*([A-Za-z][A-Za-z0-9_]*))?\s*(?:#.*)?' . "\$";
my $ENDIF_PATTERN =
  '^endif(?:\s+(!?)\s*([A-Za-z][A-Za-z0-9_]*))?\s*(?:#.*)?' . "\$";
my $PATH_PATTERN = '(\w|[/.-])+';
# This will pass through anything not of the prescribed form.
my $INCLUDE_PATTERN = ('^include\s+'
		       . '((\$\(top_srcdir\)/' . $PATH_PATTERN . ')'
		       . '|(\$\(srcdir\)/' . $PATH_PATTERN . ')'
		       . '|([^/\$]' . $PATH_PATTERN . '))\s*(#.*)?' . "\$");

# This handles substitution references like ${foo:.a=.b}.
my $SUBST_REF_PATTERN = "^([^:]*):([^=]*)=(.*)\$";

# Match `-d' as a command-line argument in a string.
my $DASH_D_PATTERN = "(^|\\s)-d(\\s|\$)";
# Directories installed during 'install-exec' phase.
my $EXEC_DIR_PATTERN =
  '^(?:bin|sbin|libexec|sysconf|localstate|lib|pkglib|.*exec.*)' . "\$";

# Constants to define the "strictness" level.
use constant FOREIGN => 0;
use constant GNU     => 1;
use constant GNITS   => 2;

# Values for AC_CANONICAL_*
use constant AC_CANONICAL_HOST   => 1;
use constant AC_CANONICAL_SYSTEM => 2;

# Values indicating when something should be cleaned.
use constant MOSTLY_CLEAN     => 0;
use constant CLEAN            => 1;
use constant DIST_CLEAN       => 2;
use constant MAINTAINER_CLEAN => 3;

# Libtool files.
my @libtool_files = qw(ltmain.sh config.guess config.sub);
# ltconfig appears here for compatibility with old versions of libtool.
my @libtool_sometimes = qw(ltconfig ltcf-c.sh ltcf-cxx.sh ltcf-gcj.sh);

# Commonly found files we look for and automatically include in
# DISTFILES.
my @common_files =
    (qw(ABOUT-GNU ABOUT-NLS AUTHORS BACKLOG COPYING COPYING.DOC COPYING.LIB
	COPYING.LESSER ChangeLog INSTALL NEWS README THANKS TODO acinclude.m4
	ansi2knr.1 ansi2knr.c compile config.guess config.rpath config.sub
	configure configure.ac configure.in depcomp elisp-comp
	install-sh libversion.in mdate-sh missing mkinstalldirs
	py-compile texinfo.tex ylwrap),
     @libtool_files, @libtool_sometimes);

# Commonly used files we auto-include, but only sometimes.
my @common_sometimes =
    qw(aclocal.m4 acconfig.h config.h.top config.h.bot stamp-vti);

# Standard directories from the GNU Coding Standards, and additional
# pkg* directories from Automake.  Stored in a hash for fast member check.
my %standard_prefix =
    map { $_ => 1 } (qw(bin data exec include info lib libexec lisp
			localstate man man1 man2 man3 man4 man5 man6
			man7 man8 man9 oldinclude pkgdatadir
			pkgincludedir pkglibdir sbin sharedstate
			sysconf));

# Declare the macros that define known variables, so we can
# hint the user if she try to use one of these variables.

# Macros accessible via aclocal.
my %am_macro_for_var =
  (
   ANSI2KNR => 'AM_C_PROTOTYPES',
   CCAS => 'AM_PROG_AS',
   CCASFLAGS => 'AM_PROG_AS',
   EMACS => 'AM_PATH_LISPDIR',
   GCJ => 'AM_PROG_GCJ',
   LEX => 'AM_PROG_LEX',
   LIBTOOL => 'AC_PROG_LIBTOOL',
   lispdir => 'AM_PATH_LISPDIR',
   pkgpyexecdir => 'AM_PATH_PYTHON',
   pkgpythondir => 'AM_PATH_PYTHON',
   pyexecdir => 'AM_PATH_PYTHON',
   PYTHON => 'AM_PATH_PYTHON',
   pythondir => 'AM_PATH_PYTHON',
   U => 'AM_C_PROTOTYPES',
   );

# Macros shipped with Autoconf.
my %ac_macro_for_var =
  (
   CC => 'AC_PROG_CC',
   CFLAGS => 'AC_PROG_CC',
   CXX => 'AC_PROG_CXX',
   CXXFLAGS => 'AC_PROG_CXX',
   F77 => 'AC_PROG_F77',
   F77FLAGS => 'AC_PROG_F77',
   RANLIB => 'AC_PROG_RANLIB',
   YACC => 'AC_PROG_YACC',
   );

# Copyright on generated Makefile.ins.
my $gen_copyright = "\
# Copyright 1994, 1995, 1996, 1997, 1998, 1999, 2000, 2001, 2002, 2003
# Free Software Foundation, Inc.
# This Makefile.in is free software; the Free Software Foundation
# gives unlimited permission to copy and/or distribute it,
# with or without modifications, as long as this notice is preserved.

# This program is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY, to the extent permitted by law; without
# even the implied warranty of MERCHANTABILITY or FITNESS FOR A
# PARTICULAR PURPOSE.
";

# These constants are returned by lang_*_rewrite functions.
# LANG_SUBDIR means that the resulting object file should be in a
# subdir if the source file is.  In this case the file name cannot
# have `..' components.
use constant LANG_IGNORE  => 0;
use constant LANG_PROCESS => 1;
use constant LANG_SUBDIR  => 2;

# These are used when keeping track of whether an object can be built
# by two different paths.
use constant COMPILE_LIBTOOL  => 1;
use constant COMPILE_ORDINARY => 2;



## ---------------------------------- ##
## Variables related to the options.  ##
## ---------------------------------- ##

# TRUE if we should always generate Makefile.in.
my $force_generation = 1;

# Strictness level as set on command line.
my $default_strictness = GNU;

# Name of strictness level, as set on command line.
my $default_strictness_name = 'gnu';

# This is TRUE if automatic dependency generation code should be
# included in generated Makefile.in.
my $cmdline_use_dependencies = 1;

# From the Perl manual.
my $symlink_exists = (eval 'symlink ("", "");', $@ eq '');

# TRUE if missing standard files should be installed.
my $add_missing = 0;

# TRUE if we should copy missing files; otherwise symlink if possible.
my $copy_missing = 0;

# TRUE if we should always update files that we know about.
my $force_missing = 0;


## ---------------------------------------- ##
## Variables filled during files scanning.  ##
## ---------------------------------------- ##

# Name of the top autoconf input: `configure.ac' or `configure.in'.
my $configure_ac = '';

# Files found by scanning configure.ac for LIBOBJS.
my %libsources = ();

# Names used in AC_CONFIG_HEADER call.
my @config_headers = ();
# Where AC_CONFIG_HEADER appears.
my $config_header_location;

# Directory where output files go.  Actually, output files are
# relative to this directory.
my $output_directory;

# List of Makefile.am's to process, and their corresponding outputs.
my @input_files = ();
my %output_files = ();

# Complete list of Makefile.am's that exist.
my @configure_input_files = ();

# List of files in AC_CONFIG_FILES/AC_OUTPUT without Makefile.am's,
# and their outputs.
my @other_input_files = ();
# Where the last AC_CONFIG_FILES/AC_OUTPUT appears.
my $ac_config_files_location;

# List of directories to search for configure-required files.  This
# can be set by AC_CONFIG_AUX_DIR.
my @config_aux_path = qw(. .. ../..);
my $config_aux_dir = '';
my $config_aux_dir_set_in_configure_in = 0;

# Whether AM_GNU_GETTEXT has been seen in configure.ac.
my $seen_gettext = 0;
# Whether AM_GNU_GETTEXT([external]) is used.
my $seen_gettext_external = 0;
# Where AM_GNU_GETTEXT appears.
my $ac_gettext_location;

# TRUE if we've seen AC_CANONICAL_(HOST|SYSTEM).
my $seen_canonical = 0;
my $canonical_location;

# Where AM_MAINTAINER_MODE appears.
my $seen_maint_mode;

# Actual version we've seen.
my $package_version = '';

# Where version is defined.
my $package_version_location;

# TRUE if we've seen AC_ENABLE_MULTILIB.
my $seen_multilib = 0;

# TRUE if we've seen AM_PROG_CC_C_O
my $seen_cc_c_o = 0;

# Where AM_INIT_AUTOMAKE is called;
my $seen_init_automake = 0;

# TRUE if we've seen AM_AUTOMAKE_VERSION.
my $seen_automake_version = 0;

# Hash table of discovered configure substitutions.  Keys are names,
# values are `FILE:LINE' strings which are used by error message
# generation.
my %configure_vars = ();

# This is used to keep track of which variable definitions we are
# scanning.  It is only used in certain limited ways, but it has to be
# global.  It is declared just for documentation purposes.
my %vars_scanned = ();

# TRUE if --cygnus seen.
my $cygnus_mode = 0;

# Hash table of AM_CONDITIONAL variables seen in configure.
my %configure_cond = ();

# This maps extensions onto language names.
my %extension_map = ();

# List of the DIST_COMMON files we discovered while reading
# configure.in
my $configure_dist_common = '';

# This maps languages names onto objects.
my %languages = ();

# List of targets we must always output.
# FIXME: Complete, and remove falsely required targets.
my %required_targets =
  (
   'all'          => 1,
   'dvi'	  => 1,
   'pdf'	  => 1,
   'ps'		  => 1,
   'info'	  => 1,
   'install-info' => 1,
   'install'      => 1,
   'install-data' => 1,
   'install-exec' => 1,
   'uninstall'    => 1,

   # FIXME: Not required, temporary hacks.
   # Well, actually they are sort of required: the -recursive
   # targets will run them anyway...
   'dvi-am'          => 1,
   'pdf-am'          => 1,
   'ps-am'           => 1,
   'info-am'         => 1,
   'install-data-am' => 1,
   'install-exec-am' => 1,
   'installcheck-am' => 1,
   'uninstall-am' => 1,

   'install-man' => 1,
  );

# This is set to 1 when Automake needs to be run again.
# (For instance, this happens when an auxiliary file such as
# depcomp is added after the toplevel Makefile.in -- which
# should distribute depcomp -- has been generated.)
my $automake_needs_to_reprocess_all_files = 0;

# Options set via AM_INIT_AUTOMAKE.
my $global_options = '';

# Same as $suffix_rules (declared below), but records only the
# default rules supplied by the languages Automake supports.
my $suffix_rules_default;

# If a file name appears as a key in this hash, then it has already
# been checked for.  This variable is local to the "require file"
# functions.
my %require_file_found = ();


################################################################

## ------------------------------------------ ##
## Variables reset by &initialize_per_input.  ##
## ------------------------------------------ ##

# Basename and relative dir of the input file.
my $am_file_name;
my $am_relative_dir;

# Same but wrt Makefile.in.
my $in_file_name;
my $relative_dir;

# These two variables are used when generating each Makefile.in.
# They hold the Makefile.in until it is ready to be printed.
my $output_rules;
my $output_vars;
my $output_trailer;
my $output_all;
my $output_header;

# Suffixes found during a run.
my @suffixes;

# Handling the variables.
#
# For a $VAR:
# - $var_value{$VAR}{$COND} is its value associated to $COND,
# - $var_location{$VAR}{$COND} is where it was defined,
# - $var_comment{$VAR}{$COND} are the comments associated to it.
# - $var_type{$VAR}{$COND} is how it has been defined (`', `+', or `:'),
# - $var_owner{$VAR}{$COND} tells who owns the variable (VAR_AUTOMAKE,
#     VAR_CONFIGURE, or VAR_MAKEFILE).
my %var_value;
my %var_location;
my %var_comment;
my %var_type;
my %var_owner;
# Possible values for var_owner.  Defined so that the owner of
# a variable can only be increased (e.g Automake should not
# override a configure or Makefile variable).
use constant VAR_AUTOMAKE => 0; # Variable defined by Automake.
use constant VAR_CONFIGURE => 1;# Variable defined in configure.ac.
use constant VAR_MAKEFILE => 2; # Variable defined in Makefile.am.

# This holds a 1 if a particular variable was examined.
my %content_seen;

# This holds the names which are targets.  These also appear in
# %contents.  $targets{TARGET}{COND} is the location of the definition
# of TARGET for condition COND.  TARGETs should not include
# a trailing $(EXEEXT), we record this in %target_name.
my %targets;

# $target_source{TARGET}{COND} is the filename where TARGET
# were defined for condition COND.  Note this must be a
# filename, *without* any line number.
my %target_source;

# $target_name{TARGET}{COND} is the real name of TARGET (in condition COND).
# The real name is often TARGET or TARGET$(EXEEXT), and TARGET never
# contain $(EXEEXT)
my %target_name;

# $target_owner{TARGET}{COND} the owner of TARGET in condition COND.
my %target_owner;
use constant TARGET_AUTOMAKE => 0; # Target defined by Automake.
use constant TARGET_USER => 1;	# Target defined in the user's Makefile.am.

# This is the conditional stack.
my @cond_stack;

# This holds the set of included files.
my @include_stack;

# This holds a list of directories which we must create at `dist'
# time.  This is used in some strange scenarios involving weird
# AC_OUTPUT commands.
my %dist_dirs;

# List of dependencies for the obvious targets.
my @all;
my @check;
my @check_tests;

# Holds the dependencies of targets which dependencies are factored.
# Typically, `.PHONY' will appear in plenty of *.am files, but must
# be output once.  Arguably all pure dependencies could be subject
# to this factorization, but it is not unpleasant to have paragraphs
# in Makefile: keeping related stuff altogether.
my %dependencies;

# Holds the factored actions.  Tied to %DEPENDENCIES, i.e., filled
# only when keys exists in %DEPENDENCIES.
my %actions;

# Keys in this hash table are files to delete.  The associated
# value tells when this should happen (MOSTLY_CLEAN, DIST_CLEAN, etc.)
my %clean_files;

# Keys in this hash table are object files or other files in
# subdirectories which need to be removed.  This only holds files
# which are created by compilations.  The value in the hash indicates
# when the file should be removed.
my %compile_clean_files;

# Keys in this hash table are directories where we expect to build a
# libtool object.  We use this information to decide what directories
# to delete.
my %libtool_clean_directories;

# Value of `$(SOURCES)', used by tags.am.
my @sources;
# Sources which go in the distribution.
my @dist_sources;

# This hash maps object file names onto their corresponding source
# file names.  This is used to ensure that each object is created
# by a single source file.
my %object_map;

# This hash maps object file names onto an integer value representing
# whether this object has been built via ordinary compilation or
# libtool compilation (the COMPILE_* constants).
my %object_compilation_map;


# This keeps track of the directories for which we've already
# created dirstamp code.
my %directory_map;

# All .P files.
my %dep_files;

# Strictness levels.
my $strictness;
my $strictness_name;

# Options from AUTOMAKE_OPTIONS.
my %options;

# Whether or not dependencies are handled.  Can be further changed
# in handle_options.
my $use_dependencies;

# This is a list of all targets to run during "make dist".
my @dist_targets;

# Keys in this hash are the basenames of files which must depend on
# ansi2knr.  Values are either the empty string, or the directory in
# which the ANSI source file appears; the directory must have a
# trailing `/'.
my %de_ansi_files;

# This maps the source extension for all suffix rule seen to
# a \hash whose keys are the possible output extensions.
#
# Note that this is transitively closed by construction:
# if we have
#       exists $suffix_rules{$ext1}{$ext2}
#    && exists $suffix_rules{$ext2}{$ext3}
# then we also have
#       exists $suffix_rules{$ext1}{$ext3}
#
# So it's easy to check whether '.foo' can be transformed to '.$(OBJEXT)'
# by checking whether $suffix_rules{'.foo'}{'.$(OBJEXT)'} exist.  This
# will work even if transforming '.foo' to '.$(OBJEXT)' involves a chain
# of several suffix rules.
#
# The value of `$suffix_rules{$ext1}{$ext2}' is the a pair
# `[ $next_sfx, $dist ]' where `$next_sfx' is target suffix
# for the next rule to use to reach '$ext2', and `$dist' the
# distance to `$ext2'.
my $suffix_rules;

# This is the name of the redirect `all' target to use.
my $all_target;

# This keeps track of which extensions we've seen (that we care
# about).
my %extension_seen;

# This is random scratch space for the language finish functions.
# Don't randomly overwrite it; examine other uses of keys first.
my %language_scratch;

# We keep track of which objects need special (per-executable)
# handling on a per-language basis.
my %lang_specific_files;

# This is set when `handle_dist' has finished.  Once this happens,
# we should no longer push on dist_common.
my $handle_dist_run;

# Used to store a set of linkers needed to generate the sources currently
# under consideration.
my %linkers_used;

# True if we need `LINK' defined.  This is a hack.
my $need_link;

# This is the list of such variables to output.
# FIXME: Might be useless actually.
my @var_list;

# Was get_object_extension run?
# FIXME: This is a hack. a better switch should be found.
my $get_object_extension_was_run;

# Contains a stack of `from' parts of variable substitutions currently in
# force.
my @substfroms;

# Contains a stack of `to' parts of variable substitutions currently in
# force.
my @substtos;

# This keeps track of all variables defined by subobjname.
# The value stored is the variable names.
# The key has the form "(COND1)VAL1(COND2)VAL2..." where VAL1 and VAL2
# are the values of the variable for condition COND1 and COND2.
my %subobjvar = ();

# This hash records helper variables used to implement '+=' in conditionals.
# Keys have the form "VAR:CONDITIONS".  The value associated to a key is
# the named of the helper variable used to append to VAR in CONDITIONS.
my %appendvar = ();


## --------------------------------- ##
## Forward subroutine declarations.  ##
## --------------------------------- ##
sub register_language (%);
sub file_contents_internal ($$%);
sub define_objects_from_sources ($$$$$$$);


# &initialize_per_input ()
# ------------------------
# (Re)-Initialize per-Makefile.am variables.
sub initialize_per_input ()
{
    reset_local_duplicates ();

    $am_file_name = '';
    $am_relative_dir = '';

    $in_file_name = '';
    $relative_dir = '';

    $output_rules = '';
    $output_vars = '';
    $output_trailer = '';
    $output_all = '';
    $output_header = '';

    @suffixes = ();

    %var_value = ();
    %var_location = ();
    %var_comment = ();
    %var_type = ();
    %var_owner = ();

    %content_seen = ();

    %targets = ();
    %target_source = ();
    %target_name = ();
    %target_owner = ();

    @cond_stack = ();

    @include_stack = ();

    %dist_dirs = ();

    @all = ();
    @check = ();
    @check_tests = ();

    %dependencies =
      (
       # Texinfoing.
       'dvi'      => [],
       'dvi-am'   => [],
       'pdf'      => [],
       'pdf-am'   => [],
       'ps'       => [],
       'ps-am'    => [],
       'info'     => [],
       'info-am'  => [],

       # Installing/uninstalling.
       'install-data-am'      => [],
       'install-exec-am'      => [],
       'uninstall-am'         => [],

       'install-man'	      => [],
       'uninstall-man'	      => [],

       'install-info'         => [],
       'install-info-am'      => [],
       'uninstall-info'       => [],

       'installcheck-am'      => [],

       # Cleaning.
       'clean-am'             => [],
       'mostlyclean-am'       => [],
       'maintainer-clean-am'  => [],
       'distclean-am'         => [],
       'clean'                => [],
       'mostlyclean'          => [],
       'maintainer-clean'     => [],
       'distclean'            => [],

       # Tarballing.
       'dist-all'             => [],

       # Phoning.
       '.PHONY'               => []
      );
    %actions = ();

    %clean_files = ();

    @sources = ();
    @dist_sources = ();

    %object_map = ();
    %object_compilation_map = ();

    %directory_map = ();

    %dep_files = ();

    $strictness = $default_strictness;
    $strictness_name = $default_strictness_name;

    %options = ();

    $use_dependencies = $cmdline_use_dependencies;

    @dist_targets = ();

    %de_ansi_files = ();


    # The first time we initialize the variables,
    # we save the value of $suffix_rules.
    if (defined $suffix_rules_default)
      {
	$suffix_rules = $suffix_rules_default;
      }
    else
      {
	$suffix_rules_default = $suffix_rules;
      }

    $all_target = '';

    %extension_seen = ();

    %language_scratch = ();

    %lang_specific_files = ();

    $handle_dist_run = 0;

    $need_link = 0;

    @var_list = ();

    $get_object_extension_was_run = 0;

    %compile_clean_files = ();

    # We always include `.'.  This isn't strictly correct.
    %libtool_clean_directories = ('.' => 1);

    %subobjvar = ();

    %appendvar = ();
}


################################################################

# Initialize our list of error/warning channels.
# Do not forget to update &usage and the manual
# if you add or change a warning channel.

# Fatal errors.
register_channel 'fatal', type => 'fatal';
# Common errors.
register_channel 'error', type => 'error';
# Errors related to GNU Standards.
register_channel 'error-gnu', type => 'error';
# Errors related to GNU Standards that should be warnings in `foreign' mode.
register_channel 'error-gnu/warn', type => 'error';
# Errors related to GNITS Standards (silent by default).
register_channel 'error-gnits', type => 'error', silent => 1;
# Internal errors.
register_channel 'automake', type => 'fatal', backtrace => 1,
  header => ("####################\n" .
	     "## Internal Error ##\n" .
	     "####################\n"),
  footer => "\nPlease contact <bug-automake\@gnu.org>.";

# Warnings related to GNU Coding Standards.
register_channel 'gnu', type => 'warning';
# Warnings about obsolete features (silent by default).
register_channel 'obsolete', type => 'warning', silent => 1;
# Warnings about non-portable constructs.
register_channel 'portability', type => 'warning', silent => 1;
# Weird syntax, unused variables, typos...
register_channel 'syntax', type => 'warning';
# Warnings about unsupported (or mis-supported) features.
register_channel 'unsupported', type => 'warning';

# For &verb.
register_channel 'verb', type => 'debug', silent => 1;
# Informative messages.
register_channel 'note', type => 'debug', silent => 0;


# Initialize our list of languages that are internally supported.

# C.
register_language ('name' => 'c',
		   'Name' => 'C',
		   'config_vars' => ['CC'],
		   'ansi' => 1,
		   'autodep' => '',
		   'flags' => ['CFLAGS', 'CPPFLAGS'],
		   'compiler' => 'COMPILE',
		   'compile' => '$(CC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CFLAGS) $(CFLAGS)',
		   'lder' => 'CCLD',
		   'ld' => '$(CC)',
		   'linker' => 'LINK',
		   'link' => '$(CCLD) $(AM_CFLAGS) $(CFLAGS) $(AM_LDFLAGS) $(LDFLAGS) -o $@',
		   'compile_flag' => '-c',
		   'extensions' => ['.c'],
		   '_finish' => \&lang_c_finish);

# C++.
register_language ('name' => 'cxx',
		   'Name' => 'C++',
		   'config_vars' => ['CXX'],
		   'linker' => 'CXXLINK',
		   'link' => '$(CXXLD) $(AM_CXXFLAGS) $(CXXFLAGS) $(AM_LDFLAGS) $(LDFLAGS) -o $@',
		   'autodep' => 'CXX',
		   'flags' => ['CXXFLAGS', 'CPPFLAGS'],
		   'compile' => '$(CXX) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_CXXFLAGS) $(CXXFLAGS)',
		   'compiler' => 'CXXCOMPILE',
		   'compile_flag' => '-c',
		   'output_flag' => '-o',
		   'lder' => 'CXXLD',
		   'ld' => '$(CXX)',
		   'pure' => 1,
		   'extensions' => ['.c++', '.cc', '.cpp', '.cxx', '.C']);

# Objective C.
register_language ('name' => 'objc',
		   'Name' => 'Objective C',
		   'config_vars' => ['OBJC'],
		   'linker' => 'OBJCLINK',,
 		   'link' => '$(OBJCLD) $(AM_OBJCFLAGS) $(OBJCFLAGS) $(AM_LDFLAGS) $(LDFLAGS) -o $@',
		   'autodep' => 'OBJC',
		   'flags' => ['OBJCFLAGS', 'CPPFLAGS'],
		   'compile' => '$(OBJC) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_OBJCFLAGS) $(OBJCFLAGS)',
		   'compiler' => 'OBJCCOMPILE',
		   'compile_flag' => '-c',
		   'output_flag' => '-o',
		   'lder' => 'OBJCLD',
		   'ld' => '$(OBJC)',
		   'pure' => 1,
		   'extensions' => ['.m']);

# Headers.
register_language ('name' => 'header',
		   'Name' => 'Header',
		   'extensions' => ['.h', '.H', '.hxx', '.h++', '.hh',
				    '.hpp', '.inc'],
		   # No output.
		   'output_extensions' => sub { return () },
		   # Nothing to do.
		   '_finish' => sub { });

# Yacc (C & C++).
register_language ('name' => 'yacc',
		   'Name' => 'Yacc',
		   'config_vars' => ['YACC'],
		   'flags' => ['YFLAGS'],
		   'compile' => '$(YACC) $(YFLAGS) $(AM_YFLAGS)',
		   'compiler' => 'YACCCOMPILE',
		   'extensions' => ['.y'],
		   'output_extensions' => sub { (my $ext = $_[0]) =~ tr/y/c/;
						return ($ext,) },
		   'rule_file' => 'yacc',
		   '_finish' => \&lang_yacc_finish,
		   '_target_hook' => \&lang_yacc_target_hook);
register_language ('name' => 'yaccxx',
		   'Name' => 'Yacc (C++)',
		   'config_vars' => ['YACC'],
		   'rule_file' => 'yacc',
		   'flags' => ['YFLAGS'],
		   'compiler' => 'YACCCOMPILE',
		   'compile' => '$(YACC) $(YFLAGS) $(AM_YFLAGS)',
		   'extensions' => ['.y++', '.yy', '.yxx', '.ypp'],
		   'output_extensions' => sub { (my $ext = $_[0]) =~ tr/y/c/;
						return ($ext,) },
		   '_finish' => \&lang_yacc_finish,
		   '_target_hook' => \&lang_yacc_target_hook);

# Lex (C & C++).
register_language ('name' => 'lex',
		   'Name' => 'Lex',
		   'config_vars' => ['LEX'],
		   'rule_file' => 'lex',
		   'flags' => ['LFLAGS'],
		   'compile' => '$(LEX) $(LFLAGS) $(AM_LFLAGS)',
		   'compiler' => 'LEXCOMPILE',
		   'extensions' => ['.l'],
		   'output_extensions' => sub { (my $ext = $_[0]) =~ tr/l/c/;
						return ($ext,) },
		   '_finish' => \&lang_lex_finish,
		   '_target_hook' => \&lang_lex_target_hook);
register_language ('name' => 'lexxx',
		   'Name' => 'Lex (C++)',
		   'config_vars' => ['LEX'],
		   'rule_file' => 'lex',
		   'flags' => ['LFLAGS'],
		   'compile' => '$(LEX) $(LFLAGS) $(AM_LFLAGS)',
		   'compiler' => 'LEXCOMPILE',
		   'extensions' => ['.l++', '.ll', '.lxx', '.lpp'],
		   'output_extensions' => sub { (my $ext = $_[0]) =~ tr/l/c/;
						return ($ext,) },
		   '_finish' => \&lang_lex_finish,
		   '_target_hook' => \&lang_lex_target_hook);

# Assembler.
register_language ('name' => 'asm',
		   'Name' => 'Assembler',
		   'config_vars' => ['CCAS', 'CCASFLAGS'],

		   'flags' => ['CCASFLAGS'],
		   # Users can set AM_ASFLAGS to includes DEFS, INCLUDES,
		   # or anything else required.  They can also set AS.
		   'compile' => '$(CCAS) $(AM_CCASFLAGS) $(CCASFLAGS)',
		   'compiler' => 'CCASCOMPILE',
		   'compile_flag' => '-c',
		   'extensions' => ['.s', '.S'],

		   # With assembly we still use the C linker.
		   '_finish' => \&lang_c_finish);

# Fortran 77
register_language ('name' => 'f77',
		   'Name' => 'Fortran 77',
		   'linker' => 'F77LINK',
		   'link' => '$(F77LD) $(AM_FFLAGS) $(FFLAGS) $(AM_LDFLAGS) $(LDFLAGS) -o $@',
		   'flags' => ['FFLAGS'],
		   'compile' => '$(F77) $(AM_FFLAGS) $(FFLAGS)',
		   'compiler' => 'F77COMPILE',
		   'compile_flag' => '-c',
		   'output_flag' => '-o',
		   'lder' => 'F77LD',
		   'ld' => '$(F77)',
		   'pure' => 1,
		   'extensions' => ['.f', '.for', '.f90']);

# Preprocessed Fortran 77
#
# The current support for preprocessing Fortran 77 just involves
# passing `$(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS)
# $(CPPFLAGS)' as additional flags to the Fortran 77 compiler, since
# this is how GNU Make does it; see the `GNU Make Manual, Edition 0.51
# for `make' Version 3.76 Beta' (specifically, from info file
# `(make)Catalogue of Rules').
#
# A better approach would be to write an Autoconf test
# (i.e. AC_PROG_FPP) for a Fortran 77 preprocessor, because not all
# Fortran 77 compilers know how to do preprocessing.  The Autoconf
# macro AC_PROG_FPP should test the Fortran 77 compiler first for
# preprocessing capabilities, and then fall back on cpp (if cpp were
# available).
register_language ('name' => 'ppf77',
		   'Name' => 'Preprocessed Fortran 77',
		   'config_vars' => ['F77'],
		   'linker' => 'F77LINK',
		   'link' => '$(F77LD) $(AM_FFLAGS) $(FFLAGS) $(AM_LDFLAGS) $(LDFLAGS) -o $@',
		   'lder' => 'F77LD',
		   'ld' => '$(F77)',
		   'flags' => ['FFLAGS', 'CPPFLAGS'],
		   'compiler' => 'PPF77COMPILE',
		   'compile' => '$(F77) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) $(AM_FFLAGS) $(FFLAGS)',
		   'compile_flag' => '-c',
		   'output_flag' => '-o',
		   'pure' => 1,
		   'extensions' => ['.F']);

# Ratfor.
register_language ('name' => 'ratfor',
		   'Name' => 'Ratfor',
		   'config_vars' => ['F77'],
		   'linker' => 'F77LINK',
		   'link' => '$(F77LD) $(AM_FFLAGS) $(FFLAGS) $(AM_LDFLAGS) $(LDFLAGS) -o $@',
		   'lder' => 'F77LD',
		   'ld' => '$(F77)',
		   'flags' => ['RFLAGS', 'FFLAGS'],
		   # FIXME also FFLAGS.
		   'compile' => '$(F77) $(AM_FFLAGS) $(FFLAGS) $(AM_RFLAGS) $(RFLAGS)',
		   'compiler' => 'RCOMPILE',
		   'compile_flag' => '-c',
		   'output_flag' => '-o',
		   'pure' => 1,
		   'extensions' => ['.r']);

# Java via gcj.
register_language ('name' => 'java',
		   'Name' => 'Java',
		   'config_vars' => ['GCJ'],
		   'linker' => 'GCJLINK',
		   'link' => '$(GCJLD) $(AM_GCJFLAGS) $(GCJFLAGS) $(AM_LDFLAGS) $(LDFLAGS) -o $@',
		   'autodep' => 'GCJ',
		   'flags' => ['GCJFLAGS'],
		   'compile' => '$(GCJ) $(AM_GCJFLAGS) $(GCJFLAGS)',
		   'compiler' => 'GCJCOMPILE',
		   'compile_flag' => '-c',
		   'output_flag' => '-o',
		   'lder' => 'GCJLD',
		   'ld' => '$(GCJ)',
		   'pure' => 1,
		   'extensions' => ['.java', '.class', '.zip', '.jar']);

################################################################

# Parse the WARNINGS environnent variable.
&parse_WARNINGS;

# Parse command line.
&parse_arguments;

# Do configure.ac scan only once.
&scan_autoconf_files;

&fatal ("no `Makefile.am' found or specified\n")
  if ! @input_files;

my $automake_has_run = 0;

do
{
  if ($automake_has_run)
    {
      &verb ('processing Makefiles another time to fix them up.');
      &prog_error ('running more than two times should never be needed.')
	if $automake_has_run >= 2;
    }
  $automake_needs_to_reprocess_all_files = 0;

  # Now do all the work on each file.
  # This guy must be local otherwise it's private to the loop.
  use vars '$am_file';
  local $am_file;
  foreach $am_file (@input_files)
    {
      if (! -f ($am_file . '.am'))
	{
	  &error ("`$am_file.am' does not exist");
	}
      else
	{
	  &generate_makefile ($output_files{$am_file}, $am_file);
	}
    }
  ++$automake_has_run;
}
while ($automake_needs_to_reprocess_all_files);

exit $exit_code;

################################################################

# Error reporting functions.

# prog_error ($MESSAGE, [%OPTIONS])
# -------------------------------
# Signal a programming error, display $MESSAGE, and exit 1.
sub prog_error ($;%)
{
  my ($msg, %opts) = @_;
  msg 'automake', '', $msg, %opts;
}

# error ($WHERE, $MESSAGE, [%OPTIONS])
# error ($MESSAGE)
# ------------------------------------
# Uncategorized errors.
sub error ($;$%)
{
  my ($where, $msg, %opts) = @_;
  msg ('error', $where, $msg, %opts);
}

# fatal ($WHERE, $MESSAGE, [%OPTIONS])
# fatal ($MESSAGE)
# ----------------------------------
# Fatal errors.
sub fatal ($;$%)
{
  my ($where, $msg, %opts) = @_;
  msg ('fatal', $where, $msg, %opts);
}

# err_var ($VARNAME, $MESSAGE, [%OPTIONS])
# ----------------------------------------
# Uncategorized errors about variables.
sub err_var ($$;%)
{
  msg_var ('error', @_);
}

# err_target ($TARGETNAME, $MESSAGE, [%OPTIONS])
# ----------------------------------------------
# Uncategorized errors about targets.
sub err_target ($$;%)
{
  msg_target ('error', @_);
}

# err_cond_target ($COND, $TARGETNAME, $MESSAGE, [%OPTIONS])
# ----------------------------------------------------------
# Uncategorized errors about conditional targets.
sub err_cond_target ($$$;%)
{
  msg_cond_target ('error', @_);
}

# err_am ($MESSAGE, [%OPTIONS])
# -----------------------------
# Uncategorized errors about the current Makefile.am.
sub err_am ($;%)
{
  msg_am ('error', @_);
}

# err_ac ($MESSAGE, [%OPTIONS])
# -----------------------------
# Uncategorized errors about configure.ac.
sub err_ac ($;%)
{
  msg_ac ('error', @_);
}

# msg_cond_var ($CHANNEL, $COND, $VARNAME, $MESSAGE, [%OPTIONS])
# --------------------------------------------------------------
# Messages about conditional variable.
sub msg_cond_var ($$$$;%)
{
  my ($channel, $cond, $var, $msg, %opts) = @_;
  msg $channel, $var_location{$var}{$cond}, $msg, %opts;
}

# msg_var ($CHANNEL, $VARNAME, $MESSAGE, [%OPTIONS])
# --------------------------------------------------
# Messages about variables.
sub msg_var ($$$;%)
{
  my ($channel, $var, $msg, %opts) = @_;
  # Don't know which condition is concerned.  Pick any.
  my $cond = (keys %{$var_value{$var}})[0];
  msg_cond_var $channel, $cond, $var, $msg, %opts;
}

# msg_cond_target ($CHANNEL, $COND, $TARGETNAME, $MESSAGE, [%OPTIONS])
# --------------------------------------------------------------------
# Messages about conditional targets.
sub msg_cond_target ($$$$;%)
{
  my ($channel, $cond, $target, $msg, %opts) = @_;
  msg $channel, $targets{$target}{$cond}, $msg, %opts;
}

# msg_target ($CHANNEL, $TARGETNAME, $MESSAGE, [%OPTIONS])
# --------------------------------------------------------
# Messages about targets.
sub msg_target ($$$;%)
{
  my ($channel, $target, $msg, %opts) = @_;
  # Don't know which condition is concerned.  Pick any.
  my $cond = (keys %{$targets{$target}})[0];
  msg_cond_target ($channel, $cond, $target, $msg, %opts);
}

# msg_am ($CHANNEL, $MESSAGE, [%OPTIONS])
# ---------------------------------------
# Messages about about the current Makefile.am.
sub msg_am ($$;%)
{
  my ($channel, $msg, %opts) = @_;
  msg $channel, "${am_file}.am", $msg, %opts;
}

# msg_ac ($CHANNEL, $MESSAGE, [%OPTIONS])
# ---------------------------------------
# Messages about about configure.ac.
sub msg_ac ($$;%)
{
  my ($channel, $msg, %opts) = @_;
  msg $channel, $configure_ac, $msg, %opts;
}

# $BOOL
# reject_var ($VAR, $ERROR_MSG)
# -----------------------------
sub reject_var ($$)
{
  my ($var, $msg) = @_;
  if (variable_defined ($var))
    {
      err_var $var, $msg;
      return 1;
    }
  return 0;
}

# $BOOL
# reject_target ($VAR, $ERROR_MSG)
# --------------------------------
sub reject_target ($$)
{
  my ($target, $msg) = @_;
  if (target_defined ($target))
    {
      err_target $target, $msg;
      return 1;
    }
  return 0;
}

# verb ($MESSAGE, [%OPTIONS])
# ---------------------------
sub verb ($;%)
{
  my ($msg, %opts) = @_;
  msg 'verb', '', $msg, %opts;
}

################################################################

# subst ($TEXT)
# -------------
# Return a configure-style substitution using the indicated text.
# We do this to avoid having the substitutions directly in automake.in;
# when we do that they are sometimes removed and this causes confusion
# and bugs.
sub subst ($)
{
    my ($text) = @_;
    return '@' . $text . '@';
}

################################################################


# $BACKPATH
# &backname ($REL-DIR)
# --------------------
# If I `cd $REL-DIR', then to come back, I should `cd $BACKPATH'.
# For instance `src/foo' => `../..'.
# Works with non strictly increasing paths, i.e., `src/../lib' => `..'.
sub backname ($)
{
    my ($file) = @_;
    my @res;
    foreach (split (/\//, $file))
    {
	next if $_ eq '.' || $_ eq '';
	if ($_ eq '..')
	{
	    pop @res;
	}
	else
	{
	    push (@res, '..');
	}
    }
    return join ('/', @res) || '.';
}

################################################################

# Pattern that matches all know input extensions (i.e. extensions used
# by the languages supported by Automake).  Using this pattern
# (instead of `\..*$') to match extensions allows Automake to support
# dot-less extensions.
my $KNOWN_EXTENSIONS_PATTERN = "";
my @known_extensions_list = ();

# accept_extensions (@EXTS)
# -------------------------
# Update $KNOWN_EXTENSIONS_PATTERN to recognize the extensions
# listed @EXTS.  Extensions should contain a dot if needed.
sub accept_extensions (@)
{
    push @known_extensions_list, @_;
    $KNOWN_EXTENSIONS_PATTERN =
	'(?:' . join ('|', map (quotemeta, @known_extensions_list)) . ')';
}

# var_SUFFIXES_trigger ($TYPE, $VALUE)
# ------------------------------------
# This is called automagically by macro_define() when SUFFIXES
# is defined ($TYPE eq '') or appended ($TYPE eq '+').
# The work here needs to be performed as a side-effect of the
# macro_define() call because SUFFIXES definitions impact
# on $KNOWN_EXTENSIONS_PATTERN, and $KNOWN_EXTENSIONS_PATTERN
# are used when parsing the input am file.
sub var_SUFFIXES_trigger ($$)
{
    my ($type, $value) = @_;
    accept_extensions (split (' ', $value));
}

################################################################


# switch_warning ($CATEGORY)
# --------------------------
# If $CATEGORY is mumble, turn on the mumble channel.
# If it's no-mumble, turn mumble off.
# Alse handle `all' and `none' for completeness.
sub switch_warning ($)
{
  my ($cat) = @_;
  my $has_no = 0;

  if ($cat =~ /^no-(.*)$/)
    {
      $cat = $1;
      $has_no = 1;
    }

  if ($cat eq 'all')
    {
      setup_channel_type 'warning', silent => $has_no;
    }
  elsif ($cat eq 'none')
    {
      setup_channel_type 'warning', silent => ! $has_no;
    }
  elsif ($cat eq 'error')
    {
      $warnings_are_errors = ! $has_no;
      # Set exit code if Perl warns about something
      # (like uninitialized variables).
      $SIG{"__WARN__"} =
	$has_no ? 'DEFAULT' : sub { print STDERR @_; $exit_code = 1; };
    }
  elsif (channel_type ($cat) eq 'warning')
    {
      setup_channel $cat, silent => $has_no;
    }
  else
    {
      return 1;
    }
  return 0;
}

# parse_WARNINGS
# --------------
# Honor the WARNINGS environment variable.
sub parse_WARNINGS ($$)
{
  if (exists $ENV{'WARNINGS'})
    {
      # Ignore unknown categories.  This is required because WARNINGS
      # should be honored by many tools.
      switch_warning $_ foreach (split (',', $ENV{'WARNINGS'}));
    }
}

# parse_warning ($OPTION, $ARGUMENT)
# ----------------------------------
# Parse the argument of --warning=CATEGORY or -WCATEGORY.
sub parse_warnings ($$)
{
  my ($opt, $categories) = @_;

  foreach my $cat (split (',', $categories))
    {
      msg 'unsupported', "unknown warning category `$cat'"
	if switch_warning $cat;
    }
}

# Parse command line.
sub parse_arguments ()
{
  # Start off as gnu.
  &set_strictness ('gnu');

  my %options =
    (
     'libdir:s' 	=> \$libdir,
     'gnu' 		=> sub { &set_strictness ('gnu'); },
     'gnits' 		=> sub { &set_strictness ('gnits'); },
     'cygnus' 	        => \$cygnus_mode,
     'foreign' 	        => sub { &set_strictness ('foreign'); },
     'include-deps' 	=> sub { $cmdline_use_dependencies = 1; },
     'i|ignore-deps' 	=> sub { $cmdline_use_dependencies = 0; },
     'no-force' 	=> sub { $force_generation = 0; },
     'f|force-missing'  => \$force_missing,
     'o|output-dir:s'	=> \$output_directory,
     'a|add-missing' 	=> \$add_missing,
     'c|copy' 	        => \$copy_missing,
     'v|verbose' 	=> sub { setup_channel 'verb', silent => 0; },
     'W|warnings:s'     => \&parse_warnings,
     # These long options (--Werror and --Wno-error) for backward
     # compatibility.  Use -Werror and -Wno-error today.
     'Werror'           => sub { parse_warnings 'W', 'error'; },
     'Wno-error'        => sub { parse_warnings 'W', 'no-error'; },
     );

  use Getopt::Long;
  Getopt::Long::config ("bundling", "pass_through");

  # See if --version or --help is used.  We want to process these before
  # anything else because the GNU Coding Standards require us to
  # `exit 0' after processing these options, and we can't garanty this
  # if we treat other options first.  (Handling other options first
  # could produce error diagnostics, and in this condition it is
  # confusing if Automake `exit 0'.)
  my %options_1st_pass =
    (
     'version' => \&version,
     'help'    => \&usage,
     # Recognize all other options (and their arguments) but do nothing.
     map { $_ => sub {} } (keys %options)
     );
  my @ARGV_backup = @ARGV;
  Getopt::Long::GetOptions %options_1st_pass
    or exit 1;
  @ARGV = @ARGV_backup;

  # Now *really* process the options.  This time we know
  # that --help and --version are not present.
  Getopt::Long::GetOptions %options
    or exit 1;

  if (defined $output_directory)
    {
      msg 'obsolete', "`--output-dir' is deprecated\n";
    }
  else
    {
      # In the next release we'll remove this entirely.
      $output_directory = '.';
    }

  foreach my $arg (@ARGV)
    {
      if ($arg =~ /^-./)
	{
	  fatal ("unrecognized option `$arg'\n"
		 . "Try `$0 --help' for more information.");
	}

      # Handle $local:$input syntax.  Note that we only examine the
      # first ":" file to see if it is automake input; the rest are
      # just taken verbatim.  We still keep all the files around for
      # dependency checking, however.
      my ($local, $input, @rest) = split (/:/, $arg);
      if (! $input)
	{
	  $input = $local;
	}
      else
	{
	  # Strip .in; later on .am is tacked on.  That is how the
	  # automake input file is found.  Maybe not the best way, but
	  # it is easy to explain.
	  $input =~ s/\.in$//
	    or fatal "invalid input file name `$arg'\n.";
	}
      push (@input_files, $input);
      $output_files{$input} = join (':', ($local, @rest));
    }

  # Take global strictness from whatever we currently have set.
  $default_strictness = $strictness;
  $default_strictness_name = $strictness_name;
}

################################################################

# Generate a Makefile.in given the name of the corresponding Makefile and
# the name of the file output by config.status.
sub generate_makefile
{
    my ($output, $makefile) = @_;

    # Reset all the Makefile.am related variables.
    &initialize_per_input;

    # Any warning setting now local to this Makefile.am.
    &dup_channel_setup;
    # AUTOMAKE_OPTIONS can contains -W flags to disable or enable
    # warnings for this file.  So hold any warning issued before
    # we have processed AUTOMAKE_OPTIONS.
    &buffer_messages ('warning');

    # Name of input file ("Makefile.am") and output file
    # ("Makefile.in").  These have no directory components.
    $am_file_name = basename ($makefile) . '.am';
    $in_file_name = basename ($makefile) . '.in';

    # $OUTPUT is encoded.  If it contains a ":" then the first element
    # is the real output file, and all remaining elements are input
    # files.  We don't scan or otherwise deal with these input file,
    # other than to mark them as dependencies.  See
    # &scan_autoconf_files for details.
    my (@secondary_inputs);
    ($output, @secondary_inputs) = split (/:/, $output);

    $relative_dir = dirname ($output);
    $am_relative_dir = dirname ($makefile);

    &read_main_am_file ($makefile . '.am');
    if (&handle_options)
    {
      # Process buffered warnings.
      &flush_messages;
      # Fatal error.  Just return, so we can continue with next file.
      return;
    }
    # Process buffered warnings.
    &flush_messages;

    # There are a few install-related variables that you should not define.
    foreach my $var ('PRE_INSTALL', 'POST_INSTALL', 'NORMAL_INSTALL')
      {
	if (exists $var_owner{$var})
	  {
	    prog_error "\$var_owner{$var}{TRUE} doesn't exist"
	      unless exists $var_owner{$var}{'TRUE'};
	    reject_var $var, "`$var' should not be defined"
	      if $var_owner{$var}{'TRUE'} != VAR_AUTOMAKE;
	  }
      }

    # Catch some obsolete variables.
    msg_var ('obsolete', 'INCLUDES',
	     "`INCLUDES' is the old name for `AM_CPPFLAGS'")
      if variable_defined ('INCLUDES');

    # At the toplevel directory, we might need config.guess, config.sub
    # or libtool scripts (ltconfig and ltmain.sh).
    if ($relative_dir eq '.')
    {
        # AC_CANONICAL_HOST and AC_CANONICAL_SYSTEM need config.guess and
        # config.sub.
        require_conf_file ($canonical_location, FOREIGN,
			   'config.guess', 'config.sub')
	  if $seen_canonical;
    }

    # We still need Makefile.in here, because sometimes the `dist'
    # target doesn't re-run automake.
    if ($am_relative_dir eq $relative_dir)
    {
	# Only distribute the files if they are in the same subdir as
	# the generated makefile.
	&push_dist_common ($in_file_name, $am_file_name);
    }

    push (@sources, '$(SOURCES)')
	if variable_defined ('SOURCES');

    # Must do this after reading .am file.  See read_main_am_file to
    # understand weird tricks we play there with variables.
    &define_variable ('subdir', $relative_dir);

    # Check first, because we might modify some state.
    &check_cygnus;
    &check_gnu_standards;
    &check_gnits_standards;

    &handle_configure ($output, $makefile, @secondary_inputs);
    &handle_gettext;
    &handle_libraries;
    &handle_ltlibraries;
    &handle_programs;
    &handle_scripts;

    # This must run first so that the ANSI2KNR definition is generated
    # before it is used by the _.c rules.  We have to do this because
    # a variable which is used in a dependency must be defined before
    # the target, or else make won't properly see it.
    &handle_compile;
    # This must be run after all the sources are scanned.
    &handle_languages;

    # We have to run this after dealing with all the programs.
    &handle_libtool;

    # Re-init SOURCES.  FIXME: other code shouldn't depend on this
    # (but currently does).
    macro_define ('SOURCES', VAR_AUTOMAKE, '', 'TRUE', "@sources", 'internal');
    define_pretty_variable ('DIST_SOURCES', '', @dist_sources);

    &handle_multilib;
    &handle_texinfo;
    &handle_emacs_lisp;
    &handle_python;
    &handle_java;
    &handle_man_pages;
    &handle_data;
    &handle_headers;
    &handle_subdirs;
    &handle_tags;
    &handle_minor_options;
    &handle_tests;

    # This must come after most other rules.
    &handle_dist ($makefile);

    &handle_footer;
    &do_check_merge_target;
    &handle_all ($output);

    # FIXME: Gross!
    if (variable_defined ('lib_LTLIBRARIES') &&
	variable_defined ('bin_PROGRAMS'))
    {
	$output_rules .= "install-binPROGRAMS: install-libLTLIBRARIES\n\n";
    }

    &handle_install;
    &handle_clean;
    &handle_factored_dependencies;

    check_typos ();

    if (! -d ($output_directory . '/' . $am_relative_dir))
    {
	mkdir ($output_directory . '/' . $am_relative_dir, 0755);
    }

    my ($out_file) = $output_directory . '/' . $makefile . ".in";
    if (! $force_generation && -e $out_file)
    {
	my ($am_time) = (stat ($makefile . '.am'))[9];
	my ($in_time) = (stat ($out_file))[9];
	# FIXME: should cache these times.
	my ($conf_time) = (stat ($configure_ac))[9];
	# FIXME: how to do unsigned comparison?
	if ($am_time < $in_time || $am_time < $conf_time)
	{
	    # No need to update.
	    return;
	}
	if (-f 'aclocal.m4')
	{
	    my ($acl_time) = (stat _)[9];
	    return if ($am_time < $acl_time);
	}
    }

    if (-e "$out_file")
    {
	unlink ($out_file)
	    or fatal "cannot remove $out_file: $!\n";
    }
    my $gm_file = new Automake::XFile "> $out_file";
    verb "creating $makefile.in";

    print $gm_file $output_vars;
    # We make sure that `all:' is the first target.
    print $gm_file $output_all;
    print $gm_file $output_header;
    print $gm_file $output_rules;
    print $gm_file $output_trailer;

    # Back out any warning setting.
    &drop_channel_setup;
}

################################################################

# A version is a string that looks like
#   MAJOR.MINOR[.MICRO][ALPHA][-FORK]
# where
#   MAJOR, MINOR, and MICRO are digits, ALPHA is a character, and
# FORK any alphanumeric word.
# Usually, ALPHA is used to label alpha releases or intermediate snapshots,
# FORK is used for CVS branches or patched releases, and MICRO is used
# for bug fixes releases on the MAJOR.MINOR branch.
#
# For the purpose of ordering, 1.4 is the same as 1.4.0, but 1.4g is
# the same as 1.4.99g.  The FORK identifier is ignored in the
# ordering, except when it looks like -pMINOR[ALPHA]: some versions
# were labelled like 1.4-p3a, this is the same as an alpha release
# labelled 1.4.3a.  Yes it's horrible, but Automake did not support
# two-dot versions in the past.

# version_split (VERSION)
# -----------------------
# Split a version string into the corresponding (MAJOR, MINOR, MICRO,
# ALPHA, FORK) tuple.  For instance "1.4g" would be split into
# (1, 4, 99, 'g', '').
# Return () on error.
sub version_split ($)
{
    my ($ver) = @_;

    # Special case for versions like 1.4-p2a.
    if ($ver =~ /^(\d+)\.(\d+)(?:-p(\d+)([a-z]+)?)$/)
    {
	return ($1, $2, $3, $4 || '', '');
    }
    # Common case.
    elsif ($ver =~ /^(\d+)\.(\d+)(?:\.(\d+))?([a-z])?(?:-([A-Za-z0-9]+))?$/)
    {
	return ($1, $2, $3 || (defined $4 ? 99 : 0), $4 || '', $5 || '');
    }
    return ();
}

# version_compare (\@LVERSION, \@RVERSION)
# ----------------------------------------
# Return 1 if LVERSION > RVERSION,
#       -1 if LVERSION < RVERSION,
#        0 if LVERSION = RVERSION.
sub version_compare (\@\@)
{
    my @l = @{$_[0]};
    my @r = @{$_[1]};

    for my $i (0, 1, 2)
    {
	return 1  if ($l[$i] > $r[$i]);
	return -1 if ($l[$i] < $r[$i]);
    }
    for my $i (3, 4)
    {
	return 1  if ($l[$i] gt $r[$i]);
	return -1 if ($l[$i] lt $r[$i]);
    }
    return 0;
}

# Handles the logic of requiring a version number in AUTOMAKE_OPTIONS.
# Return 0 if the required version is satisfied, 1 otherwise.
sub version_check ($)
{
  my ($required) = @_;
  my @version = version_split $VERSION;
  my @required = version_split $required;

  prog_error "version is incorrect: $VERSION"
    if $#version == -1;

  # This should not happen, because process_option_list and split_version
  # use similar regexes.
  prog_error "required version is incorrect: $required"
    if $#required == -1;

  # If we require 3.4n-foo then we require something
  # >= 3.4n, with the `foo' fork identifier.
  return 1
    if ($required[4] ne '' && $required[4] ne $version[4]);

  return 0 > version_compare @version, @required;
}

# $BOOL
# process_option_list ($CONFIG, @OPTIONS)
# ------------------------------
# Process a list of options.  Return 1 on error, 0 otherwise.
# This is a helper for handle_options.  CONFIG is true if we're
# handling global options.
sub process_option_list
{
  my ($config, @list) = @_;

  # FIXME: We should disallow conditional deffinitions of AUTOMAKE_OPTIONS.
  my $where = ($config ?
	       $seen_init_automake :
	       $var_location{'AUTOMAKE_OPTIONS'}{'TRUE'});

  foreach (@list)
    {
      $options{$_} = $where;
      if ($_ eq 'gnits' || $_ eq 'gnu' || $_ eq 'foreign')
	{
	  &set_strictness ($_);
	}
      elsif ($_ eq 'cygnus')
	{
	  $cygnus_mode = $where;
	}
      elsif (/^(.*\/)?ansi2knr$/)
	{
	  # An option like "../lib/ansi2knr" is allowed.  With no
	  # path prefix, we assume the required programs are in this
	  # directory.  We save the actual option for later.
	  $options{'ansi2knr'} = [$_, $where];
	}
      elsif ($_ eq 'no-installman' || $_ eq 'no-installinfo'
	     || $_ eq 'dist-shar' || $_ eq 'dist-zip'
	     || $_ eq 'dist-tarZ' || $_ eq 'dist-bzip2'
	     || $_ eq 'dejagnu' || $_ eq 'no-texinfo.tex'
	     || $_ eq 'readme-alpha' || $_ eq 'check-news'
	     || $_ eq 'subdir-objects' || $_ eq 'nostdinc'
	     || $_ eq 'no-exeext' || $_ eq 'no-define'
	     || $_ eq 'std-options')
	{
	  # Explicitly recognize these.
	}
      elsif ($_ eq 'no-dependencies')
	{
	  $use_dependencies = 0;
	}
      elsif (/^\d+\.\d+(?:\.\d+)?[a-z]?(?:-[A-Za-z0-9]+)?$/)
	{
	  # Got a version number.
	  if (version_check $&)
	    {
	      error ($where, "require Automake $_, but have $VERSION",
		     uniq_scope => US_GLOBAL);
		return 1;
	    }
	}
      elsif (/^(?:--warnings=|-W)(.*)$/)
	{
	  foreach my $cat (split (',', $1))
	    {
	      msg 'unsupported', $where, "unknown warning category `$cat'"
		if switch_warning $cat;
	    }
	}
      else
	{
	  error ($where, "option `$_' not recognized",
		 uniq_scope => US_GLOBAL);
	  return 1;
	}
    }
}

# Handle AUTOMAKE_OPTIONS variable.  Return 1 on error, 0 otherwise.
sub handle_options
{
    # Process global options first so that more specific options can
    # override.
    if (&process_option_list (1, split (' ', $global_options)))
    {
	return 1;
    }

    if (variable_defined ('AUTOMAKE_OPTIONS'))
    {
	if (&process_option_list (0, &variable_value_as_list_recursive ('AUTOMAKE_OPTIONS', '')))
	{
	    return 1;
	}
    }

    if ($strictness == GNITS)
    {
	$options{'readme-alpha'} = 1;
	$options{'std-options'} = 1;
	$options{'check-news'} = 1;
    }

    return 0;
}


# get_object_extension ($OUT)
# ---------------------------
# Return object extension.  Just once, put some code into the output.
# OUT is the name of the output file
sub get_object_extension
{
    my ($out) = @_;

    # Maybe require libtool library object files.
    my $extension = '.$(OBJEXT)';
    $extension = '.lo' if ($out =~ /\.la$/);

    # Check for automatic de-ANSI-fication.
    $extension = '$U' . $extension
      if defined $options{'ansi2knr'};

    $get_object_extension_was_run = 1;

    return $extension;
}


# Call finish function for each language that was used.
sub handle_languages
{
    if ($use_dependencies)
    {
	# Include auto-dep code.  Don't include it if DEP_FILES would
	# be empty.
	if (&saw_sources_p (0) && keys %dep_files)
	{
	    # Set location of depcomp.
	    &define_variable ('depcomp', "\$(SHELL) $config_aux_dir/depcomp");
	    &define_variable ('am__depfiles_maybe', 'depfiles');

	    require_conf_file ("$am_file.am", FOREIGN, 'depcomp');

	    my @deplist = sort keys %dep_files;

	    # We define this as a conditional variable because BSD
	    # make can't handle backslashes for continuing comments on
	    # the following line.
	    define_pretty_variable ('DEP_FILES', 'AMDEP_TRUE', @deplist);

	    # Generate each `include' individually.  Irix 6 make will
	    # not properly include several files resulting from a
	    # variable expansion; generating many separate includes
	    # seems safest.
	    $output_rules .= "\n";
	    foreach my $iter (@deplist)
 	    {
		$output_rules .= (subst ('AMDEP_TRUE')
				  . subst ('am__include')
				  . ' '
				  . subst ('am__quote')
				  . $iter
				  . subst ('am__quote')
				  . "\n");
	    }

	    # Compute the set of directories to remove in distclean-depend.
	    my @depdirs = uniq (map { dirname ($_) } @deplist);
	    $output_rules .= &file_contents ('depend',
					     DEPDIRS => "@depdirs");
	}
    }
    else
    {
	&define_variable ('depcomp', '');
	&define_variable ('am__depfiles_maybe', '');
    }

    my %done;

    # Is the c linker needed?
    my $needs_c = 0;
    foreach my $ext (sort keys %extension_seen)
    {
	next unless $extension_map{$ext};

	my $lang = $languages{$extension_map{$ext}};

	my $rule_file = $lang->rule_file || 'depend2';

	# Get information on $LANG.
	my $pfx = $lang->autodep;
	my $fpfx = ($pfx eq '') ? 'CC' : $pfx;

	my $AMDEP = (($use_dependencies && $lang->autodep ne 'no')
		     ? 'AMDEP' : 'FALSE');
	my $FASTDEP = (($use_dependencies && $lang->autodep ne 'no')
		       ? ('am__fastdep' . $fpfx) : 'FALSE');

	my %transform = ('EXT'     => $ext,
			 'PFX'     => $pfx,
			 'FPFX'    => $fpfx,
			 'AMDEP'   => $AMDEP,
			 'FASTDEP' => $FASTDEP,
			 '-c'      => $lang->compile_flag || '',
			 'MORE-THAN-ONE'
			           => (count_files_for_language ($lang->name) > 1));

	# Generate the appropriate rules for this extension.
	if (($use_dependencies && $lang->autodep ne 'no')
	    || defined $lang->compile)
	{
	    # Some C compilers don't support -c -o.  Use it only if really
	    # needed.
	    my $output_flag = $lang->output_flag || '';
	    $output_flag = '-o'
	      if (! $output_flag
		  && $lang->name eq 'c'
		  && defined $options{'subdir-objects'});

	    # Compute a possible derived extension.
	    # This is not used by depend2.am.
	    my $der_ext = (&{$lang->output_extensions} ($ext))[0];

	    $output_rules .=
	      file_contents ($rule_file,
			     %transform,
			     'GENERIC'   => 1,

			     'DERIVED-EXT' => $der_ext,

			     # In this situation we know that the
			     # object is in this directory, so
			     # $(DEPDIR) is the correct location for
			     # dependencies.
			     'DEPBASE'   => '$(DEPDIR)/$*',
			     'BASE'      => '$*',
			     'SOURCE'    => '$<',
			     'OBJ'       => '$@',
			     'OBJOBJ'    => '$@',
			     'LTOBJ'     => '$@',

			     'COMPILE'   => '$(' . $lang->compiler . ')',
			     'LTCOMPILE' => '$(LT' . $lang->compiler . ')',
			     '-o'        => $output_flag);
	}

	# Now include code for each specially handled object with this
	# language.
	my %seen_files = ();
	foreach my $file (@{$lang_specific_files{$lang->name}})
	{
	    my ($derived, $source, $obj, $myext) = split (' ', $file);

	    # We might see a given object twice, for instance if it is
	    # used under different conditions.
	    next if defined $seen_files{$obj};
	    $seen_files{$obj} = 1;

	    prog_error ("found " . $lang->name .
			" in handle_languages, but compiler not defined")
	      unless defined $lang->compile;

	    my $obj_compile = $lang->compile;

	    # Rewrite each occurence of `AM_$flag' in the compile
	    # rule into `${derived}_$flag' if it exists.
	    for my $flag (@{$lang->flags})
	      {
		my $val = "${derived}_$flag";
		$obj_compile =~ s/\(AM_$flag\)/\($val\)/
		  if variable_defined ($val);
	      }

	    my $obj_ltcompile = '$(LIBTOOL) --mode=compile ' . $obj_compile;

	    # We _need_ `-o' for per object rules.
	    my $output_flag = $lang->output_flag || '-o';

	    my $depbase = dirname ($obj);
	    $depbase = ''
		if $depbase eq '.';
	    $depbase .= '/'
		unless $depbase eq '';
	    $depbase .= '$(DEPDIR)/' . basename ($obj);

	    # Support for deansified files in subdirectories is ugly
	    # enough to deserve an explanation.
	    #
	    # A Note about normal ansi2knr processing first.  On
	    #
	    #   AUTOMAKE_OPTIONS = ansi2knr
	    #   bin_PROGRAMS = foo
	    #   foo_SOURCES = foo.c
	    #
	    # we generate rules similar to:
	    #
	    #   foo: foo$U.o; link ...
	    #   foo$U.o: foo$U.c; compile ...
	    #   foo_.c: foo.c; ansi2knr ...
	    #
	    # this is fairly compact, and will call ansi2knr depending
	    # on the value of $U (`' or `_').
	    #
	    # It's harder with subdir sources. On
	    #
	    #   AUTOMAKE_OPTIONS = ansi2knr
	    #   bin_PROGRAMS = foo
	    #   foo_SOURCES = sub/foo.c
	    #
	    # we have to create foo_.c in the current directory.
	    # (Unless the user asks 'subdir-objects'.)  This is important
	    # in case the same file (`foo.c') is compiled from other
	    # directories with different cpp options: foo_.c would
	    # be preprocessed for only one set of options if it were
	    # put in the subdirectory.
	    #
	    # Because foo$U.o must be built from either foo_.c or
	    # sub/foo.c we can't be as concise as in the first example.
	    # Instead we output
	    #
	    #   foo: foo$U.o; link ...
	    #   foo_.o: foo_.c; compile ...
	    #   foo.o: sub/foo.c; compile ...
	    #   foo_.c: foo.c; ansi2knr ...
	    #
	    # This is why we'll now transform $rule_file twice
	    # if we detect this case.
	    # A first time we output the compile rule with `$U'
	    # replaced by `_' and the source directory removed,
	    # and another time we simply remove `$U'.
	    #
	    # Note that at this point $source (as computed by
	    # &handle_single_transform_list) is `sub/foo$U.c'.
	    # This can be confusing: it can be used as-is when
	    # subdir-objects is set, otherwise you have to know
	    # it really means `foo_.c' or `sub/foo.c'.
	    my $objdir = dirname ($obj);
	    my $srcdir = dirname ($source);
	    if ($lang->ansi && $obj =~ /\$U/)
	      {
		prog_error "`$obj' contains \$U, but `$source' doesn't."
		  if $source !~ /\$U/;

		(my $source_ = $source) =~ s/\$U/_/g;
		# Explicitely clean the _.c files if they are in
		# a subdirectory. (In the current directory they get
		# erased by a `rm -f *_.c' rule.)
		$clean_files{$source_} = MOSTLY_CLEAN
		  if $objdir ne '.';
		# Output an additional rule if _.c and .c are not in
		# the same directory.  (_.c is always in $objdir.)
		if ($objdir ne $srcdir)
		  {
		    (my $obj_ = $obj) =~ s/\$U/_/g;
		    (my $depbase_ = $depbase) =~ s/\$U/_/g;
		    $source_ = basename ($source_);

		    $output_rules .=
		      file_contents ($rule_file,
				     %transform,
				     GENERIC   => 0,

				     DEPBASE   => $depbase_,
				     BASE      => $obj_,
				     SOURCE    => $source_,
				     OBJ       => "$obj_$myext",
				     OBJOBJ    => "$obj_.obj",
				     LTOBJ     => "$obj_.lo",

				     COMPILE   => $obj_compile,
				     LTCOMPILE => $obj_ltcompile,
				     -o        => $output_flag);
		    $obj =~ s/\$U//g;
		    $depbase =~ s/\$U//g;
		    $source =~ s/\$U//g;
		  }
	      }

	    $output_rules .=
	      file_contents ($rule_file,
			     (%transform,
			      'GENERIC'   => 0,

			      'DEPBASE'   => $depbase,
			      'BASE'      => $obj,
			      'SOURCE'    => $source,
			      # Use $myext and not `.o' here, in case
			      # we are actually building a new source
			      # file -- e.g. via yacc.
			      'OBJ'       => "$obj$myext",
			      'OBJOBJ'    => "$obj.obj",
			      'LTOBJ'     => "$obj.lo",

			      'COMPILE'   => $obj_compile,
			      'LTCOMPILE' => $obj_ltcompile,
			      '-o'        => $output_flag));
	}

	# The rest of the loop is done once per language.
	next if defined $done{$lang};
	$done{$lang} = 1;

	# Load the language dependent Makefile chunks.
	my %lang = map { uc ($_) => 0 } keys %languages;
	$lang{uc ($lang->name)} = 1;
	$output_rules .= file_contents ('lang-compile', %transform, %lang);

	# If the source to a program consists entirely of code from a
	# `pure' language, for instance C++ for Fortran 77, then we
	# don't need the C compiler code.  However if we run into
	# something unusual then we do generate the C code.  There are
	# probably corner cases here that do not work properly.
	# People linking Java code to Fortran code deserve pain.
	$needs_c ||= ! $lang->pure;

	define_compiler_variable ($lang)
	  if ($lang->compile);

	define_linker_variable ($lang)
	  if ($lang->link);

	require_variables ("$am_file.am", $lang->Name . " source seen",
			   'TRUE', @{$lang->config_vars});

	# Call the finisher.
	$lang->finish;

	# Flags listed in `->flags' are user variables (per GNU Standards),
	# they should not be overriden in the Makefile...
	my @dont_override = @{$lang->flags};
	# ... and so is LDFLAGS.
	push @dont_override, 'LDFLAGS' if $lang->link;

	foreach my $flag (@dont_override)
	  {
	    if (exists $var_owner{$flag})
	      {
		for my $cond (keys %{$var_owner{$flag}})
		  {
		    if ($var_owner{$flag}{$cond} == VAR_MAKEFILE)
		      {
			msg_cond_var ('gnu', $cond, $flag,
				      "`$flag' is a user variable, "
				      . "you should not override it;\n"
				      . "use `AM_$flag' instead.");
		      }
		  }
	      }
	  }
    }

    # If the project is entirely C++ or entirely Fortran 77 (i.e., 1
    # suffix rule was learned), don't bother with the C stuff.  But if
    # anything else creeps in, then use it.
    $needs_c = 1
      if $need_link || ((scalar keys %$suffix_rules)
			- (scalar keys %$suffix_rules_default)) > 1;

    if ($needs_c)
      {
	&define_compiler_variable ($languages{'c'})
	  unless defined $done{$languages{'c'}};
	define_linker_variable ($languages{'c'});
      }
}

# Check to make sure a source defined in LIBOBJS is not explicitly
# mentioned.  This is a separate function (as opposed to being inlined
# in handle_source_transform) because it isn't always appropriate to
# do this check.
sub check_libobjs_sources
{
  my ($one_file, $unxformed) = @_;

  foreach my $prefix ('', 'EXTRA_', 'dist_', 'nodist_',
		      'dist_EXTRA_', 'nodist_EXTRA_')
    {
        my @files;
	if (variable_defined ($prefix . $one_file . '_SOURCES'))
	{
	    @files = &variable_value_as_list_recursive (
				($prefix . $one_file . '_SOURCES'),
				'all');
	}
	elsif ($prefix eq '')
	{
	    @files = ($unxformed . '.c');
	}
	else
	{
	    next;
	}

	foreach my $file (@files)
	{
	  err_var ($prefix . $one_file . '_SOURCES',
		   "automatically discovered file `$file' should not" .
		   " be explicitly mentioned")
	    if defined $libsources{$file};
	}
    }
}


# @OBJECTS
# handle_single_transform_list ($VAR, $TOPPARENT, $DERIVED, $OBJ, @FILES)
# -----------------------------------------------------------------------
# Does much of the actual work for handle_source_transform.
# Arguments are:
#   $VAR is the name of the variable that the source filenames come from
#   $TOPPARENT is the name of the _SOURCES variable which is being processed
#   $DERIVED is the name of resulting executable or library
#   $OBJ is the object extension (e.g., `$U.lo')
#   @FILES is the list of source files to transform
# Result is a list of the names of objects
# %linkers_used will be updated with any linkers needed
sub handle_single_transform_list ($$$$@)
{
    my ($var, $topparent, $derived, $obj, @files) = @_;
    my @result = ();
    my $nonansi_obj = $obj;
    $nonansi_obj =~ s/\$U//g;

    # Turn sources into objects.  We use a while loop like this
    # because we might add to @files in the loop.
    while (scalar @files > 0)
    {
	$_ = shift @files;

        # Configure substitutions in _SOURCES variables are errors.
        if (/^\@.*\@$/)
        {
            err_var ($var,
		     "`$var' includes configure substitution `$_', and is " .
		     "referred to\nfrom `$topparent': configure " .
		     "substitutions are not allowed\nin _SOURCES variables");
            next;
        }

        # If the source file is in a subdirectory then the `.o' is put
        # into the current directory, unless the subdir-objects option
        # is in effect.

        # Split file name into base and extension.
        next if ! /^(?:(.*)\/)?([^\/]*)($KNOWN_EXTENSIONS_PATTERN)$/;
        my $full = $_;
        my $directory = $1 || '';
        my $base = $2;
        my $extension = $3;

        # We must generate a rule for the object if it requires its own flags.
        my $renamed = 0;
        my ($linker, $object);

	# This records whether we've seen a derived source file (eg,
	# yacc output).
	my $derived_source = 0;

	# This holds the `aggregate context' of the file we are
	# currently examining.  If the file is compiled with
	# per-object flags, then it will be the name of the object.
	# Otherwise it will be `AM'.  This is used by the target hook
	# language function.
	my $aggregate = 'AM';

        $extension = &derive_suffix ($extension, $nonansi_obj);
        my $lang;
        if ($extension_map{$extension} &&
            ($lang = $languages{$extension_map{$extension}}))
	{
            # Found the language, so see what it says.
            &saw_extension ($extension);

            # Note: computed subr call.  The language rewrite function
            # should return one of the LANG_* constants.  It could
            # also return a list whose first value is such a constant
            # and whose second value is a new source extension which
            # should be applied.  This means this particular language
            # generates another source file which we must then process
            # further.
            my $subr = 'lang_' . $lang->name . '_rewrite';
            my ($r, $source_extension)
		= & $subr ($directory, $base, $extension);
            # Skip this entry if we were asked not to process it.
            next if $r == LANG_IGNORE;

            # Now extract linker and other info.
            $linker = $lang->linker;

            my $this_obj_ext;
	    if (defined $source_extension)
	    {
		$this_obj_ext = $source_extension;
		$derived_source = 1;
	    }
	    elsif ($lang->ansi)
	    {
		$this_obj_ext = $obj;
	    }
	    else
	    {
		$this_obj_ext = $nonansi_obj;
	    }
	    $object = $base . $this_obj_ext;

	    # Do we have per-executable flags for this executable?
	    my $have_per_exec_flags = 0;
	    foreach my $flag (@{$lang->flags})
	      {
		if (variable_defined ("${derived}_$flag"))
		  {
		    $have_per_exec_flags = 1;
		    last;
		  }
	      }

            if ($have_per_exec_flags)
            {
                # We have a per-executable flag in effect for this
                # object.  In this case we rewrite the object's
                # name to ensure it is unique.  We also require
                # the `compile' program to deal with compilers
                # where `-c -o' does not work.

                # We choose the name `DERIVED_OBJECT' to ensure
                # (1) uniqueness, and (2) continuity between
                # invocations.  However, this will result in a
                # name that is too long for losing systems, in
                # some situations.  So we provide _SHORTNAME to
                # override.

                my $dname = $derived;
                if (variable_defined ($derived . '_SHORTNAME'))
                {
                    # FIXME: should use the same conditional as
                    # the _SOURCES variable.  But this is really
                    # silly overkill -- nobody should have
                    # conditional shortnames.
                    $dname = &variable_value ($derived . '_SHORTNAME');
                }
                $object = $dname . '-' . $object;

                require_conf_file ("$am_file.am", FOREIGN, 'compile')
                    if $lang->name eq 'c';

		prog_error ($lang->name . " flags defined without compiler")
		  if ! defined $lang->compile;

                $renamed = 1;
            }

            # If rewrite said it was ok, put the object into a
            # subdir.
            if ($r == LANG_SUBDIR && $directory ne '')
            {
                $object = $directory . '/' . $object;
            }

            # If doing dependency tracking, then we can't print
            # the rule.  If we have a subdir object, we need to
            # generate an explicit rule.  Actually, in any case
            # where the object is not in `.' we need a special
            # rule.  The per-object rules in this case are
            # generated later, by handle_languages.
            if ($renamed || $directory ne '')
            {
                my $obj_sans_ext = substr ($object, 0,
					   - length ($this_obj_ext));
		my $full_ansi = $full;
		if ($lang->ansi && defined $options{'ansi2knr'})
		  {
		    $full_ansi =~ s/$KNOWN_EXTENSIONS_PATTERN$/\$U$&/;
		    $obj_sans_ext .= '$U';
		  }

		my $val = ("$full_ansi $obj_sans_ext "
			   # Only use $this_obj_ext in the derived
			   # source case because in the other case we
			   # *don't* want $(OBJEXT) to appear here.
			   . ($derived_source ? $this_obj_ext : '.o'));

		# If we renamed the object then we want to use the
		# per-executable flag name.  But if this is simply a
		# subdir build then we still want to use the AM_ flag
		# name.
		if ($renamed)
		{
		    $val = "$derived $val";
		    $aggregate = $derived;
		}
		else
		{
		    $val = "AM $val";
		}

		# Each item on this list is a string consisting of
		# four space-separated values: the derived flag prefix
		# (eg, for `foo_CFLAGS', it is `foo'), the name of the
		# source file, the base name of the output file, and
		# the extension for the object file.
                push (@{$lang_specific_files{$lang->name}}, $val);
            }
        }
        elsif ($extension eq $nonansi_obj)
        {
            # This is probably the result of a direct suffix rule.
            # In this case we just accept the rewrite.
            $object = "$base$extension";
            $linker = '';
        }
        else
        {
            # No error message here.  Used to have one, but it was
            # very unpopular.
	    # FIXME: we could potentially do more processing here,
	    # perhaps treating the new extension as though it were a
	    # new source extension (as above).  This would require
	    # more restructuring than is appropriate right now.
            next;
        }

	err_am "object `$object' created by `$full' and `$object_map{$object}'"
	  if (defined $object_map{$object}
	      && $object_map{$object} ne $full);

	my $comp_val = (($object =~ /\.lo$/)
			? COMPILE_LIBTOOL : COMPILE_ORDINARY);
	(my $comp_obj = $object) =~ s/\.lo$/.\$(OBJEXT)/;
	if (defined $object_compilation_map{$comp_obj}
	    && $object_compilation_map{$comp_obj} != 0
	    # Only see the error once.
	    && ($object_compilation_map{$comp_obj}
		!= (COMPILE_LIBTOOL | COMPILE_ORDINARY))
	    && $object_compilation_map{$comp_obj} != $comp_val)
	  {
	    err_am "object `$object' created both with libtool and without";
	  }
	$object_compilation_map{$comp_obj} |= $comp_val;

	if (defined $lang)
	{
	    # Let the language do some special magic if required.
	    $lang->target_hook ($aggregate, $object, $full);
	}

	if ($derived_source)
	  {
	    prog_error ($lang->name . " has automatic dependency tracking")
	      if $lang->autodep ne 'no';
	    # Make sure this new source file is handled next.  That will
	    # make it appear to be at the right place in the list.
	    unshift (@files, $object);
	    # Distribute derived sources unless the source they are
	    # derived from is not.
	    &push_dist_common ($object)
	      unless ($topparent =~ /^(?:nobase_)?nodist_/);
	    next;
	  }

        $linkers_used{$linker} = 1;

        push (@result, $object);

        if (! defined $object_map{$object})
        {
            my @dep_list = ();
            $object_map{$object} = $full;

            # If resulting object is in subdir, we need to make
            # sure the subdir exists at build time.
            if ($object =~ /\//)
            {
                # FIXME: check that $DIRECTORY is somewhere in the
                # project

		# For Java, the way we're handling it right now, a
		# `..' component doesn't make sense.
                if ($lang->name eq 'java' && $object =~ /(\/|^)\.\.\//)
		  {
		    err_am "`$full' should not contain a `..' component";
		  }

		# Make sure object is removed by `make mostlyclean'.
		$compile_clean_files{$object} = MOSTLY_CLEAN;
		# If we have a libtool object then we also must remove
		# the ordinary .o.
		if ($object =~ /\.lo$/)
		{
		    (my $xobj = $object) =~ s,lo$,\$(OBJEXT),;
		    $compile_clean_files{$xobj} = MOSTLY_CLEAN;

		    # Remove any libtool object in this directory.
		    $libtool_clean_directories{$directory} = 1;
		}

                push (@dep_list, require_build_directory ($directory));

                # If we're generating dependencies, we also want
                # to make sure that the appropriate subdir of the
                # .deps directory is created.
		push (@dep_list,
		      require_build_directory ($directory . '/$(DEPDIR)'))
		    if $use_dependencies;
            }

            &pretty_print_rule ($object . ':', "\t", @dep_list)
                if scalar @dep_list > 0;
        }

        # Transform .o or $o file into .P file (for automatic
        # dependency code).
        if ($lang && $lang->autodep ne 'no')
        {
            my $depfile = $object;
            $depfile =~ s/\.([^.]*)$/.P$1/;
            $depfile =~ s/\$\(OBJEXT\)$/o/;
            $dep_files{dirname ($depfile) . '/$(DEPDIR)/'
			   . basename ($depfile)} = 1;
        }
    }

    return @result;
}

# ($LINKER, $OBJVAR)
# define_objects_from_sources ($VAR, $OBJVAR, $NODEFINE, $ONE_FILE,
#                              $OBJ, $PARENT, $TOPPARENT)
# ---------------------------------------------------------------------
# Define an _OBJECTS variable for a _SOURCES variable (or subvariable)
#
# Arguments are:
#   $VAR is the name of the _SOURCES variable
#   $OBJVAR is the name of the _OBJECTS variable if known (otherwise
#     it will be generated and returned).
#   $NODEFINE is a boolean: if true, $OBJVAR will not be defined (but
#     work done to determine the linker will be).
#   $ONE_FILE is the canonical (transformed) name of object to build
#   $OBJ is the object extension (ie either `.o' or `.lo').
#   $PARENT is the variable in which $VAR is used, or $VAR if not applicable.
#   $TOPPARENT is the _SOURCES variable being processed.
#
# Result is a pair ($LINKER, $OBJVAR):
#    $LINKER is a boolean, true if a linker is needed to deal with the objects,
#    $OBJVAR is the name of the variable defined to hold the objects.
#
# %linkers_used, %vars_scanned, @substfroms and @substtos should be cleared
# before use:
#   %linkers_used variable will be set to contain the linkers desired.
#   %vars_scanned will be used to check for recursive definitions.
#   @substfroms and @substtos will be used to keep a stack of variable
#   substitutions to be applied.
#
sub define_objects_from_sources ($$$$$$$)
{
    my ($var, $objvar, $nodefine, $one_file, $obj, $parent, $topparent) = @_;

    if (defined $vars_scanned{$var})
    {
	err_var $var, "variable `$var' recursively defined";
	return "";
    }
    $vars_scanned{$var} = 1;

    my $needlinker = "";
    my @allresults = ();
    foreach my $cond (variable_conditions ($var))
    {
	my @result;
	foreach my $val (&variable_value_as_list ($var, $cond, $parent))
	{
	    # If $val is a variable (i.e. ${foo} or $(bar), not a filename),
	    # handle the sub variable recursively.
	    if ($val =~ /^\$\{([^}]*)\}$/ || $val =~ /^\$\(([^)]*)\)$/)
	    {
		my $subvar = $1;

		# If the user uses a losing variable name, just ignore it.
		# This isn't ideal, but people have requested it.
		next if ($subvar =~ /\@.*\@/);

		# See if the variable is actually a substitution reference
		my ($from, $to);
		my @temp_list;
		if ($subvar =~ /$SUBST_REF_PATTERN/o)
		{
		    $subvar = $1;
		    $to = $3;
		    $from = quotemeta $2;
		}
		push @substfroms, $from;
		push @substtos, $to;

		my ($temp, $varname)
		    = define_objects_from_sources ($subvar, undef,
						   $nodefine, $one_file,
						   $obj, $var, $topparent);

		push (@result, '$('. $varname . ')');
		$needlinker ||= $temp;

		pop @substfroms;
		pop @substtos;
	    }
	    else # $var is a filename
	    {
	        my $substnum=$#substfroms;
	        while ($substnum >= 0)
		{
		    $val =~ s/$substfroms[$substnum]$/$substtos[$substnum]/
			if defined $substfroms[$substnum];
		    $substnum -= 1;
		}

		my (@transformed) =
		      &handle_single_transform_list ($var, $topparent, $one_file, $obj, $val);
		push (@result, @transformed);
		$needlinker = "true" if @transformed;
	    }
	}
	push (@allresults, [$cond, @result]);
    }
    # Find a name for the variable, unless imposed.
    $objvar = subobjname (@allresults) unless defined $objvar;
    # Define _OBJECTS conditionally
    unless ($nodefine)
    {
	foreach my $pair (@allresults)
	{
	    my ($cond, @result) = @$pair;
	    define_pretty_variable ($objvar, $cond, @result);
	}
    }

    delete $vars_scanned{$var};
    return ($needlinker, $objvar);
}


# $VARNAME
# subobjname (@DEFINITIONS)
# -------------------------
# Return a name for an object variable that with definitions @DEFINITIONS.
# @DEFINITIONS is a list of pair [$COND, @OBJECTS].
#
# If we already have an object variable containing @DEFINITIONS, reuse it.
# This way, we avoid combinatorial explosion of the generated
# variables.  Especially, in a Makefile such as:
#
# | if FOO1
# | A1=1
# | endif
# |
# | if FOO2
# | A2=2
# | endif
# |
# | ...
# |
# | if FOON
# | AN=N
# | endif
# |
# | B=$(A1) $(A2) ... $(AN)
# |
# | c_SOURCES=$(B)
# | d_SOURCES=$(B)
#
# The generated c_OBJECTS and d_OBJECTS will share the same variable
# definitions.
#
# This setup can be the case of a testsuite containing lots (>100) of
# small C programs, all testing the same set of source files.
sub subobjname (@)
{
    my $key = '';
    foreach my $pair (@_)
    {
	my ($cond, @values) = @$pair;
	$key .= "($cond)@values";
    }

    return $subobjvar{$key} if exists $subobjvar{$key};

    my $num = 1 + keys (%subobjvar);
    my $name = "am__objects_${num}";
    $subobjvar{$key} = $name;
    return $name;
}


# Handle SOURCE->OBJECT transform for one program or library.
# Arguments are:
#   canonical (transformed) name of object to build
#   actual name of object to build
#   object extension (ie either `.o' or `$o'.
# Return result is name of linker variable that must be used.
# Empty return means just use `LINK'.
sub handle_source_transform
{
    # one_file is canonical name.  unxformed is given name.  obj is
    # object extension.
    my ($one_file, $unxformed, $obj) = @_;

    my ($linker) = '';

    # No point in continuing if _OBJECTS is defined.
    return if reject_var ($one_file . '_OBJECTS',
			  $one_file . '_OBJECTS should not be defined');

    my %used_pfx = ();
    my $needlinker;
    %linkers_used = ();
    foreach my $prefix ('', 'EXTRA_', 'dist_', 'nodist_',
			'dist_EXTRA_', 'nodist_EXTRA_')
    {
	my $var = $prefix . $one_file . "_SOURCES";
	next
	  if !variable_defined ($var);

	# We are going to define _OBJECTS variables using the prefix.
	# Then we glom them all together.  So we can't use the null
	# prefix here as we need it later.
	my $xpfx = ($prefix eq '') ? 'am_' : $prefix;

	# Keep track of which prefixes we saw.
	$used_pfx{$xpfx} = 1
	  unless $prefix =~ /EXTRA_/;

	push @sources, "\$($var)";
	if ($prefix !~ /^nodist_/)
	  {
	    # If the VAR wasn't definined conditionally, we add
	    # it to DIST_SOURCES as is.  Otherwise we create a
	    # am__VAR_DIST variable which contains all possible values,
	    # and add this variable to DIST_SOURCES.
	    my $distvar = "$var";
	    my @conds = variable_conditions_recursive ($var);
	    if (@conds && $conds[0] ne 'TRUE')
	      {
		$distvar = "am__${var}_DIST";
		my @files =
		  uniq (variable_value_as_list_recursive ($var, 'all'));
		define_pretty_variable ($distvar, '', @files);
	      }
	    push @dist_sources, "\$($distvar)"
	  }

	@substfroms = ();
	@substtos = ();
	%vars_scanned = ();
	my ($temp, $objvar) =
	    define_objects_from_sources ($var,
					 $xpfx . $one_file . '_OBJECTS',
					 $prefix =~ /EXTRA_/,
					 $one_file, $obj, $var, $var);
	$needlinker ||= $temp;
    }
    if ($needlinker)
    {
	$linker ||= &resolve_linker (%linkers_used);
    }

    my @keys = sort keys %used_pfx;
    if (scalar @keys == 0)
    {
	&define_variable ($one_file . "_SOURCES", $unxformed . ".c");
	push (@sources, $unxformed . '.c');
	push (@dist_sources, $unxformed . '.c');

	%linkers_used = ();
	my (@result) =
	  &handle_single_transform_list ($one_file . '_SOURCES',
					 $one_file . '_SOURCES',
					 $one_file, $obj,
					 "$unxformed.c");
	$linker ||= &resolve_linker (%linkers_used);
	define_pretty_variable ($one_file . "_OBJECTS", '', @result)
    }
    else
    {
	grep ($_ = '$(' . $_ . $one_file . '_OBJECTS)', @keys);
	define_pretty_variable ($one_file . '_OBJECTS', '', @keys);
    }

    # If we want to use `LINK' we must make sure it is defined.
    if ($linker eq '')
    {
	$need_link = 1;
    }

    return $linker;
}


# handle_lib_objects ($XNAME, $VAR)
# ---------------------------------
# Special-case ALLOCA and LIBOBJS substitutions in _LDADD or _LIBADD variables.
# Also, generate _DEPENDENCIES variable if appropriate.
# Arguments are:
#   transformed name of object being built, or empty string if no object
#   name of _LDADD/_LIBADD-type variable to examine
# Returns 1 if LIBOBJS seen, 0 otherwise.
sub handle_lib_objects
{
    my ($xname, $var) = @_;

    prog_error "handle_lib_objects: $var undefined"
      if ! variable_defined ($var);

    my $ret = 0;
    foreach my $cond (variable_conditions_recursive ($var))
      {
	if (&handle_lib_objects_cond ($xname, $var, $cond))
	  {
	    $ret = 1;
	  }
      }
    return $ret;
}

# Subroutine of handle_lib_objects: handle a particular condition.
sub handle_lib_objects_cond
{
    my ($xname, $var, $cond) = @_;

    # We recognize certain things that are commonly put in LIBADD or
    # LDADD.
    my @dep_list = ();

    my $seen_libobjs = 0;
    my $flagvar = 0;

    foreach my $lsearch (&variable_value_as_list_recursive ($var, $cond))
    {
	# Skip -lfoo and -Ldir; these are explicitly allowed.
	next if $lsearch =~ /^-[lL]/;
	if (! $flagvar && $lsearch =~ /^-/)
	{
	    if ($var =~ /^(.*)LDADD$/)
	    {
		# Skip -dlopen and -dlpreopen; these are explicitly allowed.
		next if $lsearch =~ /^-dl(pre)?open$/;
		my $prefix = $1 || 'AM_';
		err_var ($var, "linker flags such as `$lsearch' belong in "
			 . "`${prefix}LDFLAGS");
	    }
	    else
	    {
		$var =~ /^(.*)LIBADD$/;
		# Only get this error once.
		$flagvar = 1;
		err_var ($var, "linker flags such as `$lsearch' belong in "
			 . "`${1}LDFLAGS");
	    }
	}

	# Assume we have a file of some sort, and push it onto the
	# dependency list.  Autoconf substitutions are not pushed;
	# rarely is a new dependency substituted into (eg) foo_LDADD
	# -- but "bad things (eg -lX11) are routinely substituted.
	# Note that LIBOBJS and ALLOCA are exceptions to this rule,
	# and handled specially below.
	push (@dep_list, $lsearch)
	    unless $lsearch =~ /^\@.*\@$/;

	# Automatically handle LIBOBJS and ALLOCA substitutions.
	# Basically this means adding entries to dep_files.
	if ($lsearch =~ /^\@(LT)?LIBOBJS\@$/)
	{
	    my $lt = $1 ? $1 : '';
	    my $myobjext = ($1 ? 'l' : '') . 'o';

	    push (@dep_list, $lsearch);
	    $seen_libobjs = 1;
	    if (! keys %libsources
		&& ! variable_defined ($lt . 'LIBOBJS'))
	    {
	        err_var ($var, "\@${lt}LIBOBJS\@ seen but never set in "
			 . "`$configure_ac'");
	    }

	    foreach my $iter (keys %libsources)
	    {
		if ($iter =~ /\.[cly]$/)
		{
		    &saw_extension ($&);
		    &saw_extension ('.c');
		}

		if ($iter =~ /\.h$/)
		{
		    require_file_with_macro ($cond, $var, FOREIGN, $iter);
		}
		elsif ($iter ne 'alloca.c')
		{
		    my $rewrite = $iter;
		    $rewrite =~ s/\.c$/.P$myobjext/;
		    $dep_files{'$(DEPDIR)/' . $rewrite} = 1;
		    $rewrite = "^" . quotemeta ($iter) . "\$";
		    # Only require the file if it is not a built source.
		    if (! variable_defined ('BUILT_SOURCES')
			|| ! grep (/$rewrite/,
				   &variable_value_as_list_recursive (
				   	'BUILT_SOURCES', 'all')))
		    {
			require_file_with_macro ($cond, $var, FOREIGN, $iter);
		    }
		}
	    }
	}
	elsif ($lsearch =~ /^\@(LT)?ALLOCA\@$/)
	{
	    my $lt = $1 ? $1 : '';
	    my $myobjext = ($1 ? 'l' : '') . 'o';

	    push (@dep_list, $lsearch);
	    err_var ($var, "\@${lt}ALLOCA\@ seen but `AC_FUNC_ALLOCA' not in "
		     . "`$configure_ac'")
	      if ! defined $libsources{'alloca.c'};
	    $dep_files{'$(DEPDIR)/alloca.P' . $myobjext} = 1;
	    require_file_with_macro ($cond, $var, FOREIGN, 'alloca.c');
	    &saw_extension ('c');
	}
    }

  if ($xname ne '')
    {
      my $depvar = $xname . '_DEPENDENCIES';
      if ((conditional_ambiguous_p ($depvar, $cond,
				    keys %{$var_value{$depvar}}))[0] ne '')
	{
	  # Note that we've examined this.
	  &examine_variable ($depvar);
	}
      else
	{
	  define_pretty_variable ($depvar, $cond, @dep_list);
	}
    }

  return $seen_libobjs;
}

# Canonicalize the input parameter
sub canonicalize
{
    my ($string) = @_;
    $string =~ tr/A-Za-z0-9_\@/_/c;
    return $string;
}

# Canonicalize a name, and check to make sure the non-canonical name
# is never used.  Returns canonical name.  Arguments are name and a
# list of suffixes to check for.
sub check_canonical_spelling
{
  my ($name, @suffixes) = @_;

  my $xname = &canonicalize ($name);
  if ($xname ne $name)
    {
      foreach my $xt (@suffixes)
	{
	  reject_var ("$name$xt", "use `$xname$xt', not `$name$xt'");
	}
    }

  return $xname;
}


# handle_compile ()
# -----------------
# Set up the compile suite.
sub handle_compile ()
{
    return
      unless $get_object_extension_was_run;

    # Boilerplate.
    my $default_includes = '';
    if (! defined $options{'nostdinc'})
      {
	$default_includes = ' -I. -I$(srcdir)';

	if (variable_defined ('CONFIG_HEADER'))
	  {
	    foreach my $hdr (split (' ', &variable_value ('CONFIG_HEADER')))
	      {
		$default_includes .= ' -I' . dirname ($hdr);
	      }
	  }
      }

    my (@mostly_rms, @dist_rms);
    foreach my $item (sort keys %compile_clean_files)
    {
	if ($compile_clean_files{$item} == MOSTLY_CLEAN)
	{
	    push (@mostly_rms, "\t-rm -f $item");
	}
	elsif ($compile_clean_files{$item} == DIST_CLEAN)
	{
	    push (@dist_rms, "\t-rm -f $item");
	}
	else
	{
	  prog_error 'invalid entry in %compile_clean_files';
	}
    }

    my ($coms, $vars, $rules) =
      &file_contents_internal (1, "$libdir/am/compile.am",
			       ('DEFAULT_INCLUDES' => $default_includes,
				'MOSTLYRMS' => join ("\n", @mostly_rms),
				'DISTRMS' => join ("\n", @dist_rms)));
    $output_vars .= $vars;
    $output_rules .= "$coms$rules";

    # Check for automatic de-ANSI-fication.
    if (defined $options{'ansi2knr'})
      {
	my ($ansi2knr_filename, $ansi2knr_where) = @{$options{'ansi2knr'}};
	my $ansi2knr_dir = '';

	require_variables ($ansi2knr_where, "option `ansi2knr' is used",
			   'TRUE', "ANSI2KNR", "U");

	# topdir is where ansi2knr should be.
	if ($ansi2knr_filename eq 'ansi2knr')
	  {
	    # Only require ansi2knr files if they should appear in
	    # this directory.
	    require_file ($ansi2knr_where, FOREIGN,
			  'ansi2knr.c', 'ansi2knr.1');

	    # ansi2knr needs to be built before subdirs, so unshift it.
	    unshift (@all, '$(ANSI2KNR)');
	  }
	else
	  {
	    $ansi2knr_dir = dirname ($ansi2knr_filename);
	  }

	$output_rules .= &file_contents ('ansi2knr',
					 ('ANSI2KNR-DIR' => $ansi2knr_dir));
    }
}

# handle_libtool ()
# -----------------
# Handle libtool rules.
sub handle_libtool
{
  return unless variable_defined ('LIBTOOL');

  # Libtool requires some files, but only at top level.
  require_conf_file_with_macro ('TRUE', 'LIBTOOL', FOREIGN, @libtool_files)
    if $relative_dir eq '.';

  my @libtool_rms;
  foreach my $item (sort keys %libtool_clean_directories)
    {
      my $dir = ($item eq '.') ? '' : "$item/";
      # .libs is for Unix, _libs for DOS.
      push (@libtool_rms, "\t-rm -rf ${dir}.libs ${dir}_libs");
    }

  # Output the libtool compilation rules.
  $output_rules .= &file_contents ('libtool',
				   ('LTRMS' => join ("\n", @libtool_rms)));
}

# handle_programs ()
# ------------------
# Handle C programs.
sub handle_programs
{
  my @proglist = &am_install_var ('progs', 'PROGRAMS',
				  'bin', 'sbin', 'libexec', 'pkglib',
				  'noinst', 'check');
  return if ! @proglist;

  my $seen_global_libobjs =
    variable_defined ('LDADD') && &handle_lib_objects ('', 'LDADD');

  foreach my $one_file (@proglist)
    {
      my $seen_libobjs = 0;
      my $obj = &get_object_extension ($one_file);

      # Canonicalize names and check for misspellings.
      my $xname = &check_canonical_spelling ($one_file, '_LDADD', '_LDFLAGS',
					     '_SOURCES', '_OBJECTS',
					     '_DEPENDENCIES');

      my $linker = &handle_source_transform ($xname, $one_file, $obj);

      my $xt = '';
      if (variable_defined ($xname . "_LDADD"))
	{
	  $seen_libobjs = &handle_lib_objects ($xname, $xname . '_LDADD');
	  $xt = '_LDADD';
	}
      else
	{
	  # User didn't define prog_LDADD override.  So do it.
	  &define_variable ($xname . '_LDADD', '$(LDADD)');

	  # This does a bit too much work.  But we need it to
	  # generate _DEPENDENCIES when appropriate.
	  if (variable_defined ('LDADD'))
	    {
	      $seen_libobjs = &handle_lib_objects ($xname, 'LDADD');
	    }
	  elsif (! variable_defined ($xname . '_DEPENDENCIES'))
	    {
	      &define_variable ($xname . '_DEPENDENCIES', '');
	    }
	  $xt = '_SOURCES';
	}

      reject_var ($xname . '_LIBADD',
		  "use `${xname}_LDADD', not `${xname}_LIBADD'");

      if (! variable_defined ($xname . '_LDFLAGS'))
	{
	  # Define the prog_LDFLAGS variable.
	  &define_variable ($xname . '_LDFLAGS', '');
	}

      # Determine program to use for link.
      my $xlink;
      if (variable_defined ($xname . '_LINK'))
	{
	  $xlink = $xname . '_LINK';
	}
      else
	{
	  $xlink = $linker ? $linker : 'LINK';
	}

      # If the resulting program lies into a subdirectory,
      # make sure this directory will exist.
      my $dirstamp = require_build_directory_maybe ($one_file);

      # Don't add $(EXEEXT) if user already did.
      my $extension = ($one_file !~ /\$\(EXEEXT\)$/
		       ? "\$(EXEEXT)"
		       : '');

      $output_rules .= &file_contents ('program',
				       ('PROGRAM'  => $one_file,
					'XPROGRAM' => $xname,
					'XLINK'    => $xlink,
					'DIRSTAMP' => $dirstamp,
					'EXEEXT'   => $extension));

      if ($seen_libobjs || $seen_global_libobjs)
	{
	  if (variable_defined ($xname . '_LDADD'))
	    {
	      &check_libobjs_sources ($xname, $xname . '_LDADD');
	    }
	  elsif (variable_defined ('LDADD'))
	    {
	      &check_libobjs_sources ($xname, 'LDADD');
	    }
	}
    }
}


# handle_libraries ()
# -------------------
# Handle libraries.
sub handle_libraries
{
  my @liblist = &am_install_var ('libs', 'LIBRARIES',
				 'lib', 'pkglib', 'noinst', 'check');
  return if ! @liblist;

  my @prefix = am_primary_prefixes ('LIBRARIES', 0, 'lib', 'pkglib',
				    'noinst', 'check');

  require_variables_for_macro ($prefix[0] . '_LIBRARIES',
			       'library used', 'RANLIB')
    if (@prefix);

  foreach my $onelib (@liblist)
    {
      my $seen_libobjs = 0;
      # Check that the library fits the standard naming convention.
      if (basename ($onelib) !~ /^lib.*\.a/)
	{
	  # FIXME should put line number here.  That means mapping
	  # from library name back to variable name.
	  err_am "`$onelib' is not a standard library name";
	}

      my $obj = &get_object_extension ($onelib);

      # Canonicalize names and check for misspellings.
      my $xlib = &check_canonical_spelling ($onelib, '_LIBADD', '_SOURCES',
					    '_OBJECTS', '_DEPENDENCIES',
					    '_AR');

      if (! variable_defined ($xlib . '_AR'))
	{
	  &define_variable ($xlib . '_AR', '$(AR) cru');
	}

      if (variable_defined ($xlib . '_LIBADD'))
	{
	  if (&handle_lib_objects ($xlib, $xlib . '_LIBADD'))
	    {
	      $seen_libobjs = 1;
	    }
	}
      else
	{
	  # Generate support for conditional object inclusion in
	  # libraries.
	  &define_variable ($xlib . "_LIBADD", '');
	}

      reject_var ($xlib . '_LDADD',
		  "use `${xlib}_LIBADD', not `${xlib}_LDADD'");

      # Make sure we at look at this.
      &examine_variable ($xlib . '_DEPENDENCIES');

      &handle_source_transform ($xlib, $onelib, $obj);

      # If the resulting library lies into a subdirectory,
      # make sure this directory will exist.
      my $dirstamp = require_build_directory_maybe ($onelib);

      $output_rules .= &file_contents ('library',
				       ('LIBRARY'  => $onelib,
					'XLIBRARY' => $xlib,
					'DIRSTAMP' => $dirstamp));

      if ($seen_libobjs)
	{
	  if (variable_defined ($xlib . '_LIBADD'))
	    {
	      &check_libobjs_sources ($xlib, $xlib . '_LIBADD');
	    }
	}
    }
}


# handle_ltlibraries ()
# ---------------------
# Handle shared libraries.
sub handle_ltlibraries
{
  my @liblist = &am_install_var ('ltlib', 'LTLIBRARIES',
				 'noinst', 'lib', 'pkglib', 'check');
  return if ! @liblist;

  my %instdirs;
  my @prefix = am_primary_prefixes ('LTLIBRARIES', 0, 'lib', 'pkglib',
				    'noinst', 'check');

  require_variables_for_macro ($prefix[0] . '_LTLIBRARIES',
			       'Libtool library used', 'LIBTOOL')
    if (@prefix);

  foreach my $key (@prefix)
    {
      # Get the installation directory of each library.
      (my $dir = $key) =~ s/^nobase_//;
      for (variable_value_as_list_recursive ($key . '_LTLIBRARIES', 'all'))
	{
	  # We reject libraries which are installed in several places,
	  # because we don't handle this in the rules (think `-rpath').
	  #
	  # However, we allow the same library to be listed many times
	  # for the same directory.  This is for users who need setups
	  # like
	  #   if COND1
	  #     lib_LTLIBRARIES = libfoo.la
	  #   endif
	  #   if COND2
	  #     lib_LTLIBRARIES = libfoo.la
	  #   endif
	  #
	  # Actually this will also allow
	  #   lib_LTLIBRARIES = libfoo.la libfoo.la
	  # Diagnosing this case doesn't seem worth the plain (we'd
	  # have to fill $instdirs on a per-condition basis, check
	  # implied conditions, etc.)
	  if (defined $instdirs{$_} && $instdirs{$_} ne $dir)
	    {
	      err_am ("`$_' is already going to be installed in "
		      . "`$instdirs{$_}'");
	    }
	  else
	    {
	      $instdirs{$_} = $dir;
	    }
	}
    }

  foreach my $onelib (@liblist)
    {
      my $seen_libobjs = 0;
      my $obj = &get_object_extension ($onelib);

      # Canonicalize names and check for misspellings.
      my $xlib = &check_canonical_spelling ($onelib, '_LIBADD', '_LDFLAGS',
					    '_SOURCES', '_OBJECTS',
					    '_DEPENDENCIES');

      if (! variable_defined ($xlib . '_LDFLAGS'))
	{
	  # Define the lib_LDFLAGS variable.
	  &define_variable ($xlib . '_LDFLAGS', '');
	}

      # Check that the library fits the standard naming convention.
      my $libname_rx = "^lib.*\.la";
      if ((variable_defined ($xlib . '_LDFLAGS')
	   && grep (/-module/,
		    &variable_value_as_list_recursive ($xlib . '_LDFLAGS',
						       'all')))
	  || (variable_defined ('LDFLAGS')
	      && grep (/-module/,
		       &variable_value_as_list_recursive ('LDFLAGS', 'all'))))
	{
	  # Relax name checking for libtool modules.
	  $libname_rx = "\.la";
	}
      if (basename ($onelib) !~ /$libname_rx$/)
	{
	  # FIXME should put line number here.  That means mapping
	  # from library name back to variable name.
	  msg_am ('error-gnu/warn',
		  "`$onelib' is not a standard libtool library name");
	}

      if (variable_defined ($xlib . '_LIBADD'))
	{
	  if (&handle_lib_objects ($xlib, $xlib . '_LIBADD'))
	    {
	      $seen_libobjs = 1;
	    }
	}
      else
	{
	  # Generate support for conditional object inclusion in
	  # libraries.
	  &define_variable ($xlib . "_LIBADD", '');
	}

      reject_var ("${xlib}_LDADD",
		  "use `${xlib}_LIBADD', not `${xlib}_LDADD'");

      # Make sure we at look at this.
      &examine_variable ($xlib . '_DEPENDENCIES');

      my $linker = &handle_source_transform ($xlib, $onelib, $obj);

      # Determine program to use for link.
      my $xlink;
      if (variable_defined ($xlib . '_LINK'))
	{
	  $xlink = $xlib . '_LINK';
	}
      else
	{
	  $xlink = $linker ? $linker : 'LINK';
	}

      my $rpath;
      if ($instdirs{$onelib} eq 'EXTRA'
	  || $instdirs{$onelib} eq 'noinst'
	  || $instdirs{$onelib} eq 'check')
	{
	  # It's an EXTRA_ library, so we can't specify -rpath,
	  # because we don't know where the library will end up.
	  # The user probably knows, but generally speaking automake
	  # doesn't -- and in fact configure could decide
	  # dynamically between two different locations.
	  $rpath = '';
	}
      else
	{
	  $rpath = ('-rpath $(' . $instdirs{$onelib} . 'dir)');
	}

      # If the resulting library lies into a subdirectory,
      # make sure this directory will exist.
      my $dirstamp = require_build_directory_maybe ($onelib);

      # Remember to cleanup .libs/ in this directory.
      my $dirname = dirname $onelib;
      $libtool_clean_directories{$dirname} = 1;

      $output_rules .= &file_contents ('ltlibrary',
				       ('LTLIBRARY'  => $onelib,
					'XLTLIBRARY' => $xlib,
					'RPATH'      => $rpath,
					'XLINK'      => $xlink,
					'DIRSTAMP'   => $dirstamp));
      if ($seen_libobjs)
	{
	  if (variable_defined ($xlib . '_LIBADD'))
	    {
	      &check_libobjs_sources ($xlib, $xlib . '_LIBADD');
	    }
	}
    }
}

# See if any _SOURCES variable were misspelled.
sub check_typos ()
{
  # It is ok if the user sets this particular variable.
  &examine_variable ('AM_LDFLAGS');

  foreach my $varname (keys %var_value)
    {
      foreach my $primary ('_SOURCES', '_LIBADD', '_LDADD', '_LDFLAGS',
			   '_DEPENDENCIES')
	{
	  msg_var 'syntax', $varname, "unused variable: `$varname'"
	    # Note that a configure variable is always legitimate.
	    if ($varname =~ /$primary$/ && ! $content_seen{$varname}
		&& ! exists $configure_vars{$varname});
	}
    }
}


# Handle scripts.
sub handle_scripts
{
    # NOTE we no longer automatically clean SCRIPTS, because it is
    # useful to sometimes distribute scripts verbatim.  This happens
    # eg in Automake itself.
    &am_install_var ('-candist', 'scripts', 'SCRIPTS',
		     'bin', 'sbin', 'libexec', 'pkgdata',
		     'noinst', 'check');
}


# ($OUTFILE, $VFILE, @CLEAN_FILES)
# &scan_texinfo_file ($FILENAME)
# ------------------------------
# $OUTFILE is the name of the info file produced by $FILENAME.
# $VFILE is the name of the version.texi file used (empty if none).
# @CLEAN_FILES is the list of by products (indexes etc.)
sub scan_texinfo_file
{
    my ($filename) = @_;

    # Some of the following extensions are always created, no matter
    # whether indexes are used or not.  Other (like cps, fns, ... pgs)
    # are only created when they are used.  We used to scan $FILENAME
    # for their use, but that is not enough: they could be used in
    # included files.  We can't scan included files because we don't
    # know the include path.  Therefore we always erase these files,
    # no matter whether they are used or not.
    #
    # (tmp is only created if an @macro is used and a certain e-TeX
    # feature is not available.)
    my %clean_suffixes =
      map { $_ => 1 } (qw(aux log toc tmp
			  cp cps
			  fn fns
			  ky kys
			  vr vrs
			  tp tps
			  pg pgs)); # grep 'new.*index' texinfo.tex

    my $texi = new Automake::XFile "< $filename";
    verb "reading $filename";

    my ($outfile, $vfile);
    while ($_ = $texi->getline)
    {
      if (/^\@setfilename +(\S+)/)
      {
	# Honor only the first @setfilename.  (It's possible to have
	# more occurences later if the manual shows examples of how
	# to use @setfilename...)
	next if $outfile;

        $outfile = $1;
        if ($outfile =~ /\.(.+)$/ && $1 ne 'info')
          {
	    error ("$filename:$.",
		   "output `$outfile' has unrecognized extension");
            return;
          }
      }
      # A "version.texi" file is actually any file whose name
      # matches "vers*.texi".
      elsif (/^\@include\s+(vers[^.]*\.texi)\s*$/)
      {
        $vfile = $1;
      }

      # Try to find new or unused indexes.

      # Creating a new category of index.
      elsif (/^\@def(code)?index (\w+)/)
      {
	$clean_suffixes{$2} = 1;
	$clean_suffixes{"$2s"} = 1;
      }

      # Merging an index into an another.
      elsif (/^\@syn(code)?index (\w+) (\w+)/)
      {
	delete $clean_suffixes{"$2s"};
	$clean_suffixes{"$3s"} = 1;
      }

    }

    if ($outfile eq '')
      {
	err_am "`$filename' missing \@setfilename";
	return;
      }

    my $infobase = basename ($filename);
    $infobase =~ s/\.te?xi(nfo)?$//;
    return ($outfile, $vfile,
	    map { "$infobase.$_" } (sort keys %clean_suffixes));
}

# ($DIRSTAMP, @CLEAN_FILES)
# output_texinfo_build_rules ($SOURCE, $DEST, @DEPENDENCIES)
# ----------------------------------------------------------
# SOURCE - the source Texinfo file
# DEST - the destination Info file
# DEPENDENCIES - known dependencies
sub output_texinfo_build_rules ($$@)
{
  my ($source, $dest, @deps) = @_;

  # Split `a.texi' into `a' and `.texi'.
  my ($spfx, $ssfx) = ($source =~ /^(.*?)(\.[^.]*)?$/);
  my ($dpfx, $dsfx) = ($dest =~ /^(.*?)(\.[^.]*)?$/);

  $ssfx ||= "";
  $dsfx ||= "";

  # We can output two kinds of rules: the "generic" rules
  # use Make suffix rules and are appropritate when
  # $source and $dest lie in the current directory; the "specifix"
  # rules is needed in the other case.
  #
  # The former are output only once (this is not really apparent
  # here, but just remember that some logic deeper in Automake will
  # not output the same rule twice); while the later need to be output
  # for each Texinfo source.
  my $generic;
  my $makeinfoflags;
  my $sdir = dirname $source;
  if ($sdir eq '.' && dirname ($dest) eq '.')
    {
      $generic = 1;
      $makeinfoflags = '-I $(srcdir)';
    }
  else
    {
      $generic = 0;
      $makeinfoflags = "-I $sdir -I \$(srcdir)/$sdir";
    }

  # We cannot use a suffix rule to build info files with
  # an empty extension.  Otherwise we would output a single suffix
  # inference rule, with separate dependencies, as in
  #    .texi:
  #	       $(MAKEINFO) ...
  #    foo.info: foo.texi
  # which confuse Solaris make.  (See the Autoconf manual for details.)
  # Therefore we use a specific rule in this case.  This applies
  # to info files only (dvi and pdf files always have an extension).
  my $generic_info = ($generic && $dsfx) ? 1 : 0;

  # If the resulting file lie into a subdirectory,
  # make sure this directory will exist.
  my $dirstamp = require_build_directory_maybe ($dest);

  $output_rules .= &file_contents ('texibuild',
                                   GENERIC       => $generic,
                                   GENERIC_INFO  => $generic_info,
                                   SOURCE_SUFFIX => $ssfx,
                                   SOURCE => ($generic ? '$<' : $source),
                                   SOURCE_INFO   => ($generic_info ?
						     '$<' : $source),
                                   SOURCE_REAL   => $source,
                                   DEST_PREFIX   => $dpfx,
                                   DEST_SUFFIX   => $dsfx,
                                   MAKEINFOFLAGS => $makeinfoflags,
                                   DEPS          => "@deps",
				   DIRSTAMP      => $dirstamp);
  return ($dirstamp, "$dpfx.dvi", "$dpfx.pdf", "$dpfx.ps");
}


# ($DO-SOMETHING, $TEXICLEANS)
# handle_texinfo_helper ()
# ------------------------
# Handle all Texinfo source; helper for handle_texinfo
sub handle_texinfo_helper
{
    reject_var 'TEXINFOS', "`TEXINFOS' is an anachronism; use `info_TEXINFOS'";
    reject_var 'html_TEXINFOS', "HTML generation not yet supported";

    return (0, '') if ! variable_defined ('info_TEXINFOS');

    my @texis = &variable_value_as_list_recursive ('info_TEXINFOS', 'all');

    my (@info_deps_list, @dvis_list, @pdfs_list, @pss_list, @texi_deps);
    my %versions;
    my $done = 0;
    my @texi_cleans;
    my $canonical;

    foreach my $info_cursor (@texis)
    {
        my $infobase = $info_cursor;
        $infobase =~ s/\.(txi|texinfo|texi)$//;

	if ($infobase eq $info_cursor)
	  {
	    # FIXME: report line number.
	    err_am "texinfo file `$info_cursor' has unrecognized extension";
	    next;
	  }

	# If 'version.texi' is referenced by input file, then include
	# automatic versioning capability.
	my ($out_file, $vtexi, @clean_files) =
	  &scan_texinfo_file ("$relative_dir/$info_cursor")
	    or next;
	push (@texi_cleans, @clean_files);

	# If the Texinfo source is in a subdirectory, create the
	# resulting info in this subdirectory.  If it is in the
	# current directory, try hard to not prefix "./" because
	# it breaks the generic rules.
	my $outdir = dirname ($info_cursor) . '/';
	$outdir = "" if $outdir eq './';
	$out_file =  $outdir . $out_file;

	# If user specified file_TEXINFOS, then use that as explicit
	# dependency list.
	@texi_deps = ();
	push (@texi_deps, "$outdir$vtexi") if $vtexi;

	my $canonical = &canonicalize ($infobase);
	if (variable_defined ($canonical . "_TEXINFOS"))
	{
	    push (@texi_deps, '$(' . $canonical . '_TEXINFOS)');
	    &push_dist_common ('$(' . $canonical . '_TEXINFOS)');
	}

	my ($dirstamp, @cfiles) =
	  output_texinfo_build_rules ($info_cursor, $out_file, @texi_deps);
	push (@texi_cleans, @cfiles);

	push (@info_deps_list, $out_file);
	push (@dvis_list, $infobase . '.dvi');
	push (@pdfs_list, $infobase . '.pdf');
	push (@pss_list, $infobase . '.ps');

	# If a vers*.texi file is needed, emit the rule.
	if ($vtexi)
	{
	    err_am ("`$vtexi', included in `$info_cursor', "
		    . "also included in `$versions{$vtexi}'")
	      if defined $versions{$vtexi};
	    $versions{$vtexi} = $info_cursor;

	    # We number the stamp-vti files.  This is doable since the
	    # actual names don't matter much.  We only number starting
	    # with the second one, so that the common case looks nice.
	    my $vti = ($done ? $done : 'vti');
	    ++$done;

	    # This is ugly, but it is our historical practice.
	    if ($config_aux_dir_set_in_configure_in)
	    {
		require_conf_file_with_macro ('TRUE', 'info_TEXINFOS', FOREIGN,
					      'mdate-sh');
	    }
	    else
	    {
		require_file_with_macro ('TRUE', 'info_TEXINFOS',
					 FOREIGN, 'mdate-sh');
	    }

	    my $conf_dir;
	    if ($config_aux_dir_set_in_configure_in)
	    {
		$conf_dir = $config_aux_dir;
		$conf_dir .= '/' unless $conf_dir =~ /\/$/;
	    }
	    else
	    {
		$conf_dir = '$(srcdir)/';
	    }
	    $output_rules .= &file_contents ('texi-vers',
					     TEXI     => $info_cursor,
					     VTI      => $vti,
					     STAMPVTI => "${outdir}stamp-$vti",
					     VTEXI    => "$outdir$vtexi",
					     MDDIR    => $conf_dir,
					     DIRSTAMP => $dirstamp);
	}
    }

    # Handle location of texinfo.tex.
    my $need_texi_file = 0;
    my $texinfodir;
    if ($cygnus_mode)
    {
        $texinfodir = '$(top_srcdir)/../texinfo';
	&define_variable ('TEXINFO_TEX', "$texinfodir/texinfo.tex");
    }
    elsif ($config_aux_dir_set_in_configure_in)
    {
        $texinfodir = $config_aux_dir;
	&define_variable ('TEXINFO_TEX', "$texinfodir/texinfo.tex");
	$need_texi_file = 2; # so that we require_conf_file later
    }
    elsif (variable_defined ('TEXINFO_TEX'))
    {
	# The user defined TEXINFO_TEX so assume he knows what he is
	# doing.
        $texinfodir = ('$(srcdir)/'
		       . dirname (&variable_value ('TEXINFO_TEX')));
    }
    else
    {
        $texinfodir = '$(srcdir)';
	$need_texi_file = 1;
    }
    &define_variable ('am__TEXINFO_TEX_DIR', $texinfodir);

    # The return value.
    my $texiclean = &pretty_print_internal ("", "\t  ", @texi_cleans);

    push (@dist_targets, 'dist-info');

    if (! defined $options{'no-installinfo'})
    {
	# Make sure documentation is made and installed first.  Use
	# $(INFO_DEPS), not 'info', because otherwise recursive makes
	# get run twice during "make all".
	unshift (@all, '$(INFO_DEPS)');
    }

    &define_variable ("INFO_DEPS", "@info_deps_list");
    &define_variable ("DVIS", "@dvis_list");
    &define_variable ("PDFS", "@pdfs_list");
    &define_variable ("PSS", "@pss_list");
    # This next isn't strictly needed now -- the places that look here
    # could easily be changed to look in info_TEXINFOS.  But this is
    # probably better, in case noinst_TEXINFOS is ever supported.
    &define_variable ("TEXINFOS", &variable_value ('info_TEXINFOS'));

    # Do some error checking.  Note that this file is not required
    # when in Cygnus mode; instead we defined TEXINFO_TEX explicitly
    # up above.
    if ($need_texi_file && ! defined $options{'no-texinfo.tex'})
    {
	if ($need_texi_file > 1)
	{
	    require_conf_file_with_macro ('TRUE', 'info_TEXINFOS', FOREIGN,
					  'texinfo.tex');
	}
	else
	{
	    require_file_with_macro ('TRUE', 'info_TEXINFOS', FOREIGN,
				     'texinfo.tex');
	}
    }

    return (1, $texiclean);
}

# handle_texinfo ()
# -----------------
# Handle all Texinfo source.
sub handle_texinfo
{
    my ($do_something, $texiclean) = handle_texinfo_helper ();
    $output_rules .=  &file_contents ('texinfos',
				      ('TEXICLEAN' => $texiclean,
				       'LOCAL-TEXIS' => $do_something));
}

# Handle any man pages.
sub handle_man_pages
{
    reject_var 'MANS', "`MANS' is an anachronism; use `man_MANS'";

    # Find all the sections in use.  We do this by first looking for
    # "standard" sections, and then looking for any additional
    # sections used in man_MANS.
    my (%sections, %vlist);
    # We handle nodist_ for uniformity.  man pages aren't distributed
    # by default so it isn't actually very important.
    foreach my $pfx ('', 'dist_', 'nodist_')
    {
	# Add more sections as needed.
	foreach my $section ('0'..'9', 'n', 'l')
	{
	    if (variable_defined ($pfx . 'man' . $section . '_MANS'))
	    {
		$sections{$section} = 1;
		$vlist{'$(' . $pfx . 'man' . $section . '_MANS)'} = 1;

		&push_dist_common ('$(' . $pfx . 'man' . $section . '_MANS)')
		    if $pfx eq 'dist_';
	    }
	}

	if (variable_defined ($pfx . 'man_MANS'))
	{
	    $vlist{'$(' . $pfx . 'man_MANS)'} = 1;
	    foreach (&variable_value_as_list_recursive ($pfx . 'man_MANS', 'all'))
	    {
		# A page like `foo.1c' goes into man1dir.
		if (/\.([0-9a-z])([a-z]*)$/)
		{
		    $sections{$1} = 1;
		}
	    }

	    &push_dist_common ('$(' . $pfx . 'man_MANS)')
		if $pfx eq 'dist_';
	}
    }

    return unless %sections;

    # Now for each section, generate an install and unintall rule.
    # Sort sections so output is deterministic.
    foreach my $section (sort keys %sections)
    {
	$output_rules .= &file_contents ('mans', ('SECTION' => $section));
    }

    my @mans = sort keys %vlist;
    $output_vars .= file_contents ('mans-vars',
				   ('MANS' => "@mans"));

    if (! defined $options{'no-installman'})
    {
	push (@all, '$(MANS)');
    }
}

# Handle DATA variables.
sub handle_data
{
    &am_install_var ('-noextra', '-candist', 'data', 'DATA',
		     'data', 'sysconf', 'sharedstate', 'localstate',
		     'pkgdata', 'noinst', 'check');
}

# Handle TAGS.
sub handle_tags
{
    my @tag_deps = ();
    my @ctag_deps = ();
    if (variable_defined ('SUBDIRS'))
    {
	$output_rules .= ("tags-recursive:\n"
			  . "\tlist=\'\$(SUBDIRS)\'; for subdir in \$\$list; do \\\n"
			  # Never fail here if a subdir fails; it
			  # isn't important.
			  . "\t  test \"\$\$subdir\" = . || (cd \$\$subdir"
			  . " && \$(MAKE) \$(AM_MAKEFLAGS) tags); \\\n"
			  . "\tdone\n");
	push (@tag_deps, 'tags-recursive');
	&depend ('.PHONY', 'tags-recursive');

	$output_rules .= ("ctags-recursive:\n"
			  . "\tlist=\'\$(SUBDIRS)\'; for subdir in \$\$list; do \\\n"
			  # Never fail here if a subdir fails; it
			  # isn't important.
			  . "\t  test \"\$\$subdir\" = . || (cd \$\$subdir"
			  . " && \$(MAKE) \$(AM_MAKEFLAGS) ctags); \\\n"
			  . "\tdone\n");
	push (@ctag_deps, 'ctags-recursive');
	&depend ('.PHONY', 'ctags-recursive');
    }

    if (&saw_sources_p (1)
	|| variable_defined ('ETAGS_ARGS')
	|| @tag_deps)
    {
	my @config;
	foreach my $spec (@config_headers)
	{
	    my ($out, @ins) = split_config_file_spec ($spec);
	    foreach my $in (@ins)
	      {
		# If the config header source is in this directory,
		# require it.
		push @config, basename ($in)
		  if $relative_dir eq dirname ($in);
	      }
	}
	$output_rules .= &file_contents ('tags',
					 ('CONFIG' => "@config",
					  'TAGSDIRS'   => "@tag_deps",
					  'CTAGSDIRS'  => "@ctag_deps"));
	&examine_variable ('TAGS_DEPENDENCIES');
    }
    elsif (reject_var ('TAGS_DEPENDENCIES',
		       "doesn't make sense to define `TAGS_DEPENDENCIES'"
		       . "without\nsources or `ETAGS_ARGS'"))
    {
    }
    else
    {
	# Every Makefile must define some sort of TAGS rule.
	# Otherwise, it would be possible for a top-level "make TAGS"
	# to fail because some subdirectory failed.
	$output_rules .= "tags: TAGS\nTAGS:\n\n";
	# Ditto ctags.
	$output_rules .= "ctags: CTAGS\nCTAGS:\n\n";
    }
}

# Handle multilib support.
sub handle_multilib
{
    if ($seen_multilib && $relative_dir eq '.')
    {
	$output_rules .= &file_contents ('multilib');
    }
}


# $BOOLEAN
# &for_dist_common ($A, $B)
# -------------------------
# Subroutine for &handle_dist: sort files to dist.
#
# We put README first because it then becomes easier to make a
# Usenet-compliant shar file (in these, README must be first).
#
# FIXME: do more ordering of files here.
sub for_dist_common
{
    return 0
        if $a eq $b;
    return -1
        if $a eq 'README';
    return 1
        if $b eq 'README';
    return $a cmp $b;
}


# handle_dist ($MAKEFILE)
# -----------------------
# Handle 'dist' target.
sub handle_dist
{
    my ($makefile) = @_;

    # `make dist' isn't used in a Cygnus-style tree.
    # Omit the rules so that people don't try to use them.
    return if $cygnus_mode;

    # Look for common files that should be included in distribution.
    # If the aux dir is set, and it does not have a Makefile.am, then
    # we check for these files there as well.
    my $check_aux = 0;
    my $auxdir = '';
    if ($relative_dir eq '.'
	&& $config_aux_dir_set_in_configure_in)
    {
	($auxdir = $config_aux_dir) =~ s,^\$\(top_srcdir\)/,,;
	if (! &is_make_dir ($auxdir))
	{
	    $check_aux = 1;
	}
    }
    foreach my $cfile (@common_files)
    {
	if (-f ($relative_dir . "/" . $cfile)
	    # The file might be absent, but if it can be built it's ok.
	    || exists $targets{$cfile})
	{
	    &push_dist_common ($cfile);
	}

	# Don't use `elsif' here because a file might meaningfully
	# appear in both directories.
	if ($check_aux && -f ($auxdir . '/' . $cfile))
	{
	    &push_dist_common ($auxdir . '/' . $cfile);
	}
    }

    # We might copy elements from $configure_dist_common to
    # %dist_common if we think we need to.  If the file appears in our
    # directory, we would have discovered it already, so we don't
    # check that.  But if the file is in a subdir without a Makefile,
    # we want to distribute it here if we are doing `.'.  Ugly!
    if ($relative_dir eq '.')
    {
       foreach my $file (split (' ' , $configure_dist_common))
       {
	   push_dist_common ($file)
	     unless is_make_dir (dirname ($file));
       }
    }



    # Files to distributed.  Don't use &variable_value_as_list_recursive
    # as it recursively expands `$(dist_pkgdata_DATA)' etc.
    check_variable_defined_unconditionally ('DIST_COMMON');
    my @dist_common = split (' ', variable_value ('DIST_COMMON', 'TRUE'));
    @dist_common = uniq (sort for_dist_common (@dist_common));
    pretty_print ('DIST_COMMON = ', "\t", @dist_common);

    # Now that we've processed DIST_COMMON, disallow further attempts
    # to set it.
    $handle_dist_run = 1;

    # Scan EXTRA_DIST to see if we need to distribute anything from a
    # subdir.  If so, add it to the list.  I didn't want to do this
    # originally, but there were so many requests that I finally
    # relented.
    if (variable_defined ('EXTRA_DIST'))
    {
	# FIXME: This should be fixed to work with conditionals.  That
	# will require only making the entries in %dist_dirs under the
	# appropriate condition.  This is meaningful if the nature of
	# the distribution should depend upon the configure options
	# used.
	foreach (&variable_value_as_list_recursive ('EXTRA_DIST', ''))
	{
	    next if /^\@.*\@$/;
	    next unless s,/+[^/]+$,,;
	    $dist_dirs{$_} = 1
		unless $_ eq '.';
	}
    }

    # We have to check DIST_COMMON for extra directories in case the
    # user put a source used in AC_OUTPUT into a subdir.
    my $topsrcdir = backname ($relative_dir);
    foreach (&variable_value_as_list_recursive ('DIST_COMMON', 'all'))
    {
	next if /^\@.*\@$/;
	s/\$\(top_srcdir\)/$topsrcdir/;
	s/\$\(srcdir\)/./;
	next unless s,/+[^/]+$,,;
	$dist_dirs{$_} = 1
	    unless $_ eq '.';
    }

    # Rule to check whether a distribution is viable.
    my %transform = ('DISTCHECK-HOOK' => &target_defined ('distcheck-hook'),
		     'GETTEXT' => $seen_gettext && !$seen_gettext_external);

    # Prepend $(distdir) to each directory given.
    my %rewritten = map { '$(distdir)/' . "$_" => 1 } keys %dist_dirs;
    $transform{'DISTDIRS'} = join (' ', sort keys %rewritten);

    # If we have SUBDIRS, create all dist subdirectories and do
    # recursive build.
    if (variable_defined ('SUBDIRS'))
    {
	# If SUBDIRS is conditionally defined, then set DIST_SUBDIRS
	# to all possible directories, and use it.  If DIST_SUBDIRS is
	# defined, just use it.
	my $dist_subdir_name;
	# Note that we check DIST_SUBDIRS first on purpose.  At least
	# one project uses so many conditional subdirectories that
	# calling variable_conditionally_defined on SUBDIRS will cause
	# automake to grow to 150Mb.  Sigh.
	if (variable_defined ('DIST_SUBDIRS')
	    || variable_conditionally_defined ('SUBDIRS'))
	{
	    $dist_subdir_name = 'DIST_SUBDIRS';
	    if (! variable_defined ('DIST_SUBDIRS'))
	    {
		define_pretty_variable
		  ('DIST_SUBDIRS', '',
		   uniq (&variable_value_as_list_recursive ('SUBDIRS', 'all')));
	    }
	}
	else
	{
	    $dist_subdir_name = 'SUBDIRS';
	    # We always define this because that is what `distclean'
	    # wants.
	    define_pretty_variable ('DIST_SUBDIRS', '', '$(SUBDIRS)');
	}

	$transform{'DIST_SUBDIR_NAME'} = $dist_subdir_name;
    }

    # If the target `dist-hook' exists, make sure it is run.  This
    # allows users to do random weird things to the distribution
    # before it is packaged up.
    push (@dist_targets, 'dist-hook')
      if &target_defined ('dist-hook');
    $transform{'DIST-TARGETS'} = join(' ', @dist_targets);

    # Defining $(DISTDIR).
    $transform{'DISTDIR'} = !variable_defined('distdir');
    $transform{'TOP_DISTDIR'} = backname ($relative_dir);

    $output_rules .= &file_contents ('distdir', %transform);
}


# Handle subdirectories.
sub handle_subdirs
{
    return
      unless variable_defined ('SUBDIRS');

    my @subdirs = &variable_value_as_list_recursive ('SUBDIRS', 'all');
    my @dsubdirs = ();
    @dsubdirs = &variable_value_as_list_recursive ('DIST_SUBDIRS', 'all')
      if variable_defined ('DIST_SUBDIRS');

    # If an `obj/' directory exists, BSD make will enter it before
    # reading `Makefile'.  Hence the `Makefile' in the current directory
    # will not be read.
    #
    #  % cat Makefile
    #  all:
    #          echo Hello
    #  % cat obj/Makefile
    #  all:
    #          echo World
    #  % make      # GNU make
    #  echo Hello
    #  Hello
    #  % pmake     # BSD make
    #  echo World
    #  World
    msg_var ('portability', 'SUBDIRS',
	     "naming a subdirectory `obj' causes troubles with BSD make")
      if grep ($_ eq 'obj', @subdirs);
    msg_var ('portability', 'DIST_SUBDIRS',
	     "naming a subdirectory `obj' causes troubles with BSD make")
      if grep ($_ eq 'obj', @dsubdirs);

    # Make sure each directory mentioned in SUBDIRS actually exists.
    foreach my $dir (@subdirs)
    {
	# Skip directories substituted by configure.
	next if $dir =~ /^\@.*\@$/;

	if (! -d $am_relative_dir . '/' . $dir)
	{
	    err_var ('SUBDIRS', "required directory $am_relative_dir/$dir "
		     . "does not exist");
	    next;
	}

	err_var 'SUBDIRS', "directory should not contain `/'"
	  if $dir =~ /\//;
    }

    $output_rules .= &file_contents ('subdirs');
    variable_pretty_output ('RECURSIVE_TARGETS', 'TRUE');
}


# ($REGEN, @DEPENDENCIES)
# &scan_aclocal_m4
# ----------------
# If aclocal.m4 creation is automated, return the list of its dependencies.
sub scan_aclocal_m4
{
    my $regen_aclocal = 0;

    if (-f 'aclocal.m4')
    {
       # When using aclocal.m4, define this variable even in subdirectories,
       # because every Makefile.in depends on $(ACLOCAL_M4).
       &define_variable ("ACLOCAL_M4", '$(top_srcdir)/aclocal.m4');
    }

    return (0, ())
      unless $relative_dir eq '.';

    &examine_variable ('CONFIG_STATUS_DEPENDENCIES');
    &examine_variable ('CONFIGURE_DEPENDENCIES');

    if (-f 'aclocal.m4')
    {
	&push_dist_common ('aclocal.m4');

	my $aclocal = new Automake::XFile "< aclocal.m4";
	my $line = $aclocal->getline;
	$regen_aclocal = $line =~ 'generated automatically by aclocal';
    }

    my @ac_deps = ();

    if (-f 'acinclude.m4')
    {
	$regen_aclocal = 1;
	push @ac_deps, 'acinclude.m4';
    }

    if (variable_defined ('ACLOCAL_M4_SOURCES'))
    {
	push (@ac_deps, '$(ACLOCAL_M4_SOURCES)');
    }
    elsif (variable_defined ('ACLOCAL_AMFLAGS'))
    {
	# Scan all -I directories for m4 files.  These are our
	# dependencies.
	my $examine_next = 0;
	foreach my $amdir (&variable_value_as_list_recursive ('ACLOCAL_AMFLAGS', ''))
	{
	    if ($examine_next)
	    {
		$examine_next = 0;
		if ($amdir !~ /^\// && -d $amdir)
		{
		    foreach my $ac_dep (&my_glob ($amdir . '/*.m4'))
		    {
			$ac_dep =~ s/^\.\/+//;
			push (@ac_deps, $ac_dep)
			  unless $ac_dep eq "aclocal.m4"
			    || $ac_dep eq "acinclude.m4";
		    }
		}
	    }
	    elsif ($amdir eq '-I')
	    {
		$examine_next = 1;
	    }
	}
    }

    # Note that it might be possible that aclocal.m4 doesn't exist but
    # should be auto-generated.  This case probably isn't very
    # important.

    return ($regen_aclocal, @ac_deps);
}


# @DEPENDENCY
# &rewrite_inputs_into_dependencies ($ADD_SRCDIR, @INPUTS)
# --------------------------------------------------------
# Rewrite a list of input files into a form suitable to put on a
# dependency list.  The idea is that if an input file has a directory
# part the same as the current directory, then the directory part is
# simply removed.  But if the directory part is different, then
# $(top_srcdir) is prepended.  Among other things, this is used to
# generate the dependency list for the output files generated by
# AC_OUTPUT.  Consider what the dependencies should look like in this
# case:
#   AC_OUTPUT(src/out:src/in1:lib/in2)
# The first argument, ADD_SRCDIR, is 1 if $(top_srcdir) should be added.
# If 0 then files that require this addition will simply be ignored.
sub rewrite_inputs_into_dependencies ($@)
{
    my ($add_srcdir, @inputs) = @_;
    my @newinputs;

    foreach my $single (@inputs)
    {
	if (dirname ($single) eq $relative_dir)
	{
	    push (@newinputs, basename ($single));
	}
	elsif ($add_srcdir)
	{
	    push (@newinputs, '$(top_srcdir)/' . $single);
	}
    }

    return @newinputs;
}

# Handle remaking and configure stuff.
# We need the name of the input file, to do proper remaking rules.
sub handle_configure
{
    my ($local, $input, @secondary_inputs) = @_;

    my $input_base = basename ($input);
    my $local_base = basename ($local);

    my $amfile = $input_base . '.am';
    # We know we can always add '.in' because it really should be an
    # error if the .in was missing originally.
    my $infile = '$(srcdir)/' . $input_base . '.in';
    my $colon_infile = '';
    if ($local ne $input || @secondary_inputs)
    {
	$colon_infile = ':' . $input . '.in';
    }
    $colon_infile .= ':' . join (':', @secondary_inputs)
	if @secondary_inputs;

    my @rewritten = rewrite_inputs_into_dependencies (1, @secondary_inputs);

    my ($regen_aclocal_m4, @aclocal_m4_deps) = scan_aclocal_m4 ();

    $output_rules .=
      &file_contents ('configure',
		      ('MAKEFILE'
		       => $local_base,
		       'MAKEFILE-DEPS'
		       => "@rewritten",
		       'CONFIG-MAKEFILE'
		       => ($relative_dir eq '.') ? '$@' : '$(subdir)/$@',
		       'MAKEFILE-IN'
		       => $infile,
		       'MAKEFILE-IN-DEPS'
		       => "@include_stack",
		       'MAKEFILE-AM'
		       => $amfile,
		       'STRICTNESS'
		       => $cygnus_mode ? 'cygnus' : $strictness_name,
		       'USE-DEPS'
		       => $cmdline_use_dependencies ? '' : ' --ignore-deps',
		       'MAKEFILE-AM-SOURCES'
		       =>  "$input$colon_infile",
		       'REGEN-ACLOCAL-M4'
		       => $regen_aclocal_m4,
		       'ACLOCAL_M4_DEPS'
		       => "@aclocal_m4_deps"));

    if ($relative_dir eq '.')
    {
	&push_dist_common ('acconfig.h')
	    if -f 'acconfig.h';
    }

    # If we have a configure header, require it.
    my $hdr_index = 0;
    my @distclean_config;
    foreach my $spec (@config_headers)
      {
	$hdr_index += 1;
	# $CONFIG_H_PATH: config.h from top level.
	my ($config_h_path, @ins) = split_config_file_spec ($spec);
	my $config_h_dir = dirname ($config_h_path);

	# If the header is in the current directory we want to build
	# the header here.  Otherwise, if we're at the topmost
	# directory and the header's directory doesn't have a
	# Makefile, then we also want to build the header.
	if ($relative_dir eq $config_h_dir
	    || ($relative_dir eq '.' && ! &is_make_dir ($config_h_dir)))
	{
	    my ($cn_sans_dir, $stamp_dir);
	    if ($relative_dir eq $config_h_dir)
	    {
		$cn_sans_dir = basename ($config_h_path);
		$stamp_dir = '';
	    }
	    else
	    {
		$cn_sans_dir = $config_h_path;
		if ($config_h_dir eq '.')
		{
		    $stamp_dir = '';
		}
		else
		{
		    $stamp_dir = $config_h_dir . '/';
		}
	    }

	    # Compute relative path from directory holding output
	    # header to directory holding input header.  FIXME:
	    # doesn't handle case where we have multiple inputs.
	    my $in0_sans_dir;
	    if (dirname ($ins[0]) eq $relative_dir)
	    {
		$in0_sans_dir = basename ($ins[0]);
	    }
	    else
	    {
	        $in0_sans_dir = backname ($relative_dir) . '/' . $ins[0];
	    }

	    require_file ($config_header_location, FOREIGN, $in0_sans_dir);

	    # Header defined and in this directory.
	    my @files;
	    if (-f $config_h_path . '.top')
	    {
		push (@files, "$cn_sans_dir.top");
	    }
	    if (-f $config_h_path . '.bot')
	    {
		push (@files, "$cn_sans_dir.bot");
	    }

	    push_dist_common (@files);

	    # For now, acconfig.h can only appear in the top srcdir.
	    if (-f 'acconfig.h')
	    {
	        push (@files, '$(top_srcdir)/acconfig.h');
	    }

	    my $stamp = "${stamp_dir}stamp-h${hdr_index}";
            $output_rules .=
	      file_contents ('remake-hdr',
			     ('FILES'         => "@files",
			      'CONFIG_H'      => $cn_sans_dir,
			      'CONFIG_HIN'    => $in0_sans_dir,
			      'CONFIG_H_PATH' => $config_h_path,
			      'STAMP'         => "$stamp"));

	    push @distclean_config, $cn_sans_dir, $stamp;
	}
    }

    $output_rules .= file_contents ('clean-hdr',
				    ('FILES' => "@distclean_config"))
      if @distclean_config;

    # Set location of mkinstalldirs.
    define_variable ('mkinstalldirs',
		     ('$(SHELL) ' . $config_aux_dir . '/mkinstalldirs'));

    reject_var ('CONFIG_HEADER',
		"`CONFIG_HEADER' is an anachronism; now determined "
		. "automatically\nfrom `$configure_ac'");

    my @config_h;
    foreach my $spec (@config_headers)
      {
	my ($out, @ins) = split_config_file_spec ($spec);
	# Generate CONFIG_HEADER define.
	if ($relative_dir eq dirname ($out))
	{
	    push @config_h, basename ($out);
	}
	else
	{
	    push @config_h, "\$(top_builddir)/$out";
	}
    }
    define_variable ("CONFIG_HEADER", "@config_h")
      if @config_h;

    # Now look for other files in this directory which must be remade
    # by config.status, and generate rules for them.
    my @actual_other_files = ();
    foreach my $lfile (@other_input_files)
    {
        my $file;
	my @inputs;
	if ($lfile =~ /^([^:]*):(.*)$/)
	{
	    # This is the ":" syntax of AC_OUTPUT.
	    $file = $1;
	    @inputs = split (':', $2);
	}
	else
	{
	    # Normal usage.
	    $file = $lfile;
	    @inputs = $file . '.in';
	}

	# Automake files should not be stored in here, but in %MAKE_LIST.
        prog_error "$lfile in \@other_input_files"
	  if -f $file . '.am';

	my $local = basename ($file);

	# Make sure the dist directory for each input file is created.
	# We only have to do this at the topmost level though.  This
	# is a bit ugly but it easier than spreading out the logic,
	# especially in cases like AC_OUTPUT(foo/out:bar/in), where
	# there is no Makefile in bar/.
	if ($relative_dir eq '.')
	{
	    foreach (@inputs)
	    {
		$dist_dirs{dirname ($_)} = 1;
	    }
	}

	# We skip files that aren't in this directory.  However, if
	# the file's directory does not have a Makefile, and we are
	# currently doing `.', then we create a rule to rebuild the
	# file in the subdir.
	my $fd = dirname ($file);
	if ($fd ne $relative_dir)
	{
	    if ($relative_dir eq '.' && ! &is_make_dir ($fd))
	    {
		$local = $file;
	    }
	    else
	    {
		next;
	    }
	}

	my @rewritten_inputs = rewrite_inputs_into_dependencies (1, @inputs);
	$output_rules .= ($local . ': '
			  . '$(top_builddir)/config.status '
			  . "@rewritten_inputs\n"
			  . "\t"
			  . 'cd $(top_builddir) && '
			  . '$(SHELL) ./config.status '
			  . ($relative_dir eq '.' ? '' : '$(subdir)/')
   			  . '$@'
			  . "\n");
	push (@actual_other_files, $local);

	# Require all input files.
	require_file ($ac_config_files_location, FOREIGN,
		      rewrite_inputs_into_dependencies (0, @inputs));
    }

    # These files get removed by "make clean".
    define_pretty_variable ('CONFIG_CLEAN_FILES', '', @actual_other_files);
}

# Handle C headers.
sub handle_headers
{
    my @r = &am_install_var ('-defaultdist', 'header', 'HEADERS', 'include',
			     'oldinclude', 'pkginclude',
			     'noinst', 'check');
    foreach (@r)
    {
	next unless /\..*$/;
	&saw_extension ($&);
    }
}

sub handle_gettext
{
  return if ! $seen_gettext || $relative_dir ne '.';

  if (! variable_defined ('SUBDIRS'))
    {
      err_ac "AM_GNU_GETTEXT used but SUBDIRS not defined";
      return;
    }

  # Perform some sanity checks to help users get the right setup.
  # We disable these tests when po/ doesn't exist in order not to disallow
  # unusual gettext setups.
  #
  # Bruno Haible:
  # | The idea is:
  # |
  # |  1) If a package doesn't have a directory po/ at top level, it
  # |     will likely have multiple po/ directories in subpackages.
  # |
  # |  2) It is useful to warn for the absence of intl/ if AM_GNU_GETTEXT
  # |     is used without 'external'. It is also useful to warn for the
  # |     presence of intl/ if AM_GNU_GETTEXT([external]) is used. Both
  # |     warnings apply only to the usual layout of packages, therefore
  # |     they should both be disabled if no po/ directory is found at
  # |     top level.

  if (-d 'po')
    {
      my @subdirs = &variable_value_as_list_recursive ('SUBDIRS', 'all');

      msg_var ('syntax', 'SUBDIRS',
	       "AM_GNU_GETTEXT used but `po' not in SUBDIRS")
	if ! grep ($_ eq 'po', @subdirs);

      # intl/ is not required when AM_GNU_GETTEXT is called with
      # the `external' option.
      msg_var ('syntax', 'SUBDIRS',
	       "AM_GNU_GETTEXT used but `intl' not in SUBDIRS")
	if (! $seen_gettext_external
	    && ! grep ($_ eq 'intl', @subdirs));

      # intl/ should not be used with AM_GNU_GETTEXT([external])
      msg_var ('syntax', 'SUBDIRS',
	       "`intl' should not be in SUBDIRS when "
	       . "AM_GNU_GETTEXT([external]) is used")
	if ($seen_gettext_external && grep ($_ eq 'intl', @subdirs));
    }

  require_file ($ac_gettext_location, GNU, 'ABOUT-NLS');
}

# Handle footer elements.
sub handle_footer
{
    # NOTE don't use define_pretty_variable here, because
    # $contents{...} is already defined.
    $output_vars .= 'SOURCES = ' . variable_value ('SOURCES') . "\n\n"
      if variable_value ('SOURCES');

    reject_target ('.SUFFIXES',
		   "use variable `SUFFIXES', not target `.SUFFIXES'");

    # Note: AIX 4.1 /bin/make will fail if any suffix rule appears
    # before .SUFFIXES.  So we make sure that .SUFFIXES appears before
    # anything else, by sticking it right after the default: target.
    $output_header .= ".SUFFIXES:\n";
    if (@suffixes || variable_defined ('SUFFIXES'))
    {
	# Make sure suffixes has unique elements.  Sort them to ensure
	# the output remains consistent.  However, $(SUFFIXES) is
	# always at the start of the list, unsorted.  This is done
	# because make will choose rules depending on the ordering of
	# suffixes, and this lets the user have some control.  Push
	# actual suffixes, and not $(SUFFIXES).  Some versions of make
	# do not like variable substitutions on the .SUFFIXES line.
	my @user_suffixes = (variable_defined ('SUFFIXES')
			     ? &variable_value_as_list_recursive ('SUFFIXES', '')
			     : ());

	my %suffixes = map { $_ => 1 } @suffixes;
	delete @suffixes{@user_suffixes};

	$output_header .= (".SUFFIXES: "
			   . join (' ', @user_suffixes, sort keys %suffixes)
			   . "\n");
    }

    $output_trailer .= file_contents ('footer');
}


# Generate `make install' rules.
sub handle_install ()
{
  $output_rules .= &file_contents
    ('install',
     maybe_BUILT_SOURCES => (variable_defined ('BUILT_SOURCES')
			     ? (" \$(BUILT_SOURCES)\n"
				. "\t\$(MAKE) \$(AM_MAKEFLAGS)")
			     : ''),
     'installdirs-local' => (target_defined ('installdirs-local')
			     ? ' installdirs-local' : ''),
     am__installdirs => variable_value ('am__installdirs') || '');
}


# Deal with all and all-am.
sub handle_all ($)
{
    my ($makefile) = @_;

    # Output `all-am'.

    # Put this at the beginning for the sake of non-GNU makes.  This
    # is still wrong if these makes can run parallel jobs.  But it is
    # right enough.
    unshift (@all, basename ($makefile));

    foreach my $spec (@config_headers)
      {
        my ($out, @ins) = split_config_file_spec ($spec);
	push (@all, basename ($out))
	  if dirname ($out) eq $relative_dir;
      }

    # Install `all' hooks.
    if (&target_defined ("all-local"))
    {
      push (@all, "all-local");
      &depend ('.PHONY', "all-local");
    }

    &pretty_print_rule ("all-am:", "\t\t", @all);
    &depend ('.PHONY', 'all-am', 'all');


    # Output `all'.

    my @local_headers = ();
    push @local_headers, '$(BUILT_SOURCES)'
      if variable_defined ('BUILT_SOURCES');
    foreach my $spec (@config_headers)
      {
        my ($out, @ins) = split_config_file_spec ($spec);
	push @local_headers, basename ($out)
	  if dirname ($out) eq $relative_dir;
      }

    if (@local_headers)
      {
	# We need to make sure config.h is built before we recurse.
	# We also want to make sure that built sources are built
	# before any ordinary `all' targets are run.  We can't do this
	# by changing the order of dependencies to the "all" because
	# that breaks when using parallel makes.  Instead we handle
	# things explicitly.
	$output_all .= ("all: @local_headers"
			. "\n\t"
			. '$(MAKE) $(AM_MAKEFLAGS) '
			. (variable_defined ('SUBDIRS')
			   ? 'all-recursive' : 'all-am')
			. "\n\n");
      }
    else
      {
	$output_all .= "all: " . (variable_defined ('SUBDIRS')
				  ? 'all-recursive' : 'all-am') . "\n\n";
      }
}


# Handle check merge target specially.
sub do_check_merge_target
{
    if (&target_defined ('check-local'))
    {
	# User defined local form of target.  So include it.
	push (@check_tests, 'check-local');
	&depend ('.PHONY', 'check-local');
    }

    # In --cygnus mode, check doesn't depend on all.
    if ($cygnus_mode)
    {
	# Just run the local check rules.
	&pretty_print_rule ('check-am:', "\t\t", @check);
    }
    else
    {
	# The check target must depend on the local equivalent of
	# `all', to ensure all the primary targets are built.  Then it
	# must build the local check rules.
	$output_rules .= "check-am: all-am\n";
	&pretty_print_rule ("\t\$(MAKE) \$(AM_MAKEFLAGS)", "\t  ",
			    @check)
	    if @check;
    }
    &pretty_print_rule ("\t\$(MAKE) \$(AM_MAKEFLAGS)", "\t  ",
			@check_tests)
	if @check_tests;

    &depend ('.PHONY', 'check', 'check-am');
    # Handle recursion.  We have to honor BUILT_SOURCES like for `all:'.
    $output_rules .= ("check: "
		      . (variable_defined ('BUILT_SOURCES')
			 ? "\$(BUILT_SOURCES)\n\t\$(MAKE) \$(AM_MAKEFLAGS) "
			 : '')
		      . (variable_defined ('SUBDIRS')
			 ? 'check-recursive' : 'check-am')
		      . "\n");
}

# Handle all 'clean' targets.
sub handle_clean
{
  # Clean the files listed in user variables if they exist.
  $clean_files{'$(MOSTLYCLEANFILES)'} = MOSTLY_CLEAN
    if variable_defined ('MOSTLYCLEANFILES');
  $clean_files{'$(CLEANFILES)'} = CLEAN
    if variable_defined ('CLEANFILES');
  $clean_files{'$(DISTCLEANFILES)'} = DIST_CLEAN
    if variable_defined ('DISTCLEANFILES');
  $clean_files{'$(MAINTAINERCLEANFILES)'} = MAINTAINER_CLEAN
    if variable_defined ('MAINTAINERCLEANFILES');

  # Built sources are automatically removed by maintainer-clean.
  $clean_files{'$(BUILT_SOURCES)'} = MAINTAINER_CLEAN
    if variable_defined ('BUILT_SOURCES');

  # Compute a list of "rm"s to run for each target.
  my %rms = (MOSTLY_CLEAN, [],
	     CLEAN, [],
	     DIST_CLEAN, [],
	     MAINTAINER_CLEAN, []);

  foreach my $file (keys %clean_files)
    {
      my $when = $clean_files{$file};
      prog_error 'invalid entry in %clean_files'
	unless exists $rms{$when};

      my $rm = "rm -f $file";
      # If file is a variable, make sure when don't call `rm -f' without args.
      $rm ="test -z \"$file\" || $rm"
	if ($file =~ /^\s*\$(\(.*\)|\{.*\})\s*$/);

      push @{$rms{$when}}, "\t-$rm\n";
    }

  $output_rules .= &file_contents
    ('clean',
     MOSTLYCLEAN_RMS      => join ('', @{$rms{&MOSTLY_CLEAN}}),
     CLEAN_RMS            => join ('', @{$rms{&CLEAN}}),
     DISTCLEAN_RMS        => join ('', @{$rms{&DIST_CLEAN}}),
     MAINTAINER_CLEAN_RMS => join ('', @{$rms{&MAINTAINER_CLEAN}}));
}


# &depend ($CATEGORY, @DEPENDENDEES)
# ----------------------------------
# The target $CATEGORY depends on @DEPENDENDEES.
sub depend
{
    my ($category, @dependendees) = @_;
    {
      push (@{$dependencies{$category}}, @dependendees);
    }
}


# &target_cmp ($A, $B)
# --------------------
# Subroutine for &handle_factored_dependencies to let `.PHONY' be last.
sub target_cmp
{
    return 0
        if $a eq $b;
    return -1
        if $b eq '.PHONY';
    return 1
        if $a eq '.PHONY';
    return $a cmp $b;
}


# &handle_factored_dependencies ()
# --------------------------------
# Handle everything related to gathered targets.
sub handle_factored_dependencies
{
  # Reject bad hooks.
  foreach my $utarg ('uninstall-data-local', 'uninstall-data-hook',
		     'uninstall-exec-local', 'uninstall-exec-hook')
    {
      my $x = $utarg;
      $x =~ s/(data|exec)-//;
      reject_target ($utarg, "use `$x', not `$utarg'");
    }

  reject_target ('install-local',
		 "use `install-data-local' or `install-exec-local', "
		 . "not `install-local'");

  reject_target ('install-info-local',
		 "`install-info-local' target defined but "
		 . "`no-installinfo' option not in use")
    unless defined $options{'no-installinfo'};

  # Install the -local hooks.
  foreach (keys %dependencies)
    {
      # Hooks are installed on the -am targets.
      s/-am$// or next;
      if (&target_defined ("$_-local"))
	{
	  depend ("$_-am", "$_-local");
	  &depend ('.PHONY', "$_-local");
	}
    }

  # Install the -hook hooks.
  # FIXME: Why not be as liberal as we are with -local hooks?
  foreach ('install-exec', 'install-data', 'uninstall')
    {
      if (&target_defined ("$_-hook"))
	{
	  $actions{"$_-am"} .=
	    ("\t\@\$(NORMAL_INSTALL)\n"
	     . "\t" . '$(MAKE) $(AM_MAKEFLAGS) ' . "$_-hook\n");
	}
    }

  # All the required targets are phony.
  depend ('.PHONY', keys %required_targets);

  # Actually output gathered targets.
  foreach (sort target_cmp keys %dependencies)
    {
      # If there is nothing about this guy, skip it.
      next
	unless (@{$dependencies{$_}}
		|| $actions{$_}
		|| $required_targets{$_});
      &pretty_print_rule ("$_:", "\t",
			  uniq (sort @{$dependencies{$_}}));
      $output_rules .= $actions{$_}
      if defined $actions{$_};
      $output_rules .= "\n";
    }
}


# &handle_tests_dejagnu ()
# ------------------------
sub handle_tests_dejagnu
{
    push (@check_tests, 'check-DEJAGNU');
    $output_rules .= file_contents ('dejagnu');
}


# Handle TESTS variable and other checks.
sub handle_tests
{
  if (defined $options{'dejagnu'})
    {
      &handle_tests_dejagnu;
    }
  else
    {
      foreach my $c ('DEJATOOL', 'RUNTEST', 'RUNTESTFLAGS')
	{
	  reject_var ($c, "`$c' defined but `dejagnu' not in "
		      . "`AUTOMAKE_OPTIONS'");
	}
    }

  if (variable_defined ('TESTS'))
    {
      push (@check_tests, 'check-TESTS');
      $output_rules .= &file_contents ('check');
    }
}

# Handle Emacs Lisp.
sub handle_emacs_lisp
{
  my @elfiles = &am_install_var ('-candist', 'lisp', 'LISP',
				 'lisp', 'noinst');

  return if ! @elfiles;

  # Generate .elc files.
  my @elcfiles = map { $_ . 'c' } @elfiles;
  define_pretty_variable ('ELCFILES', '', @elcfiles);

  define_pretty_variable ('am__ELFILES', '', @elfiles);

  # Do not depend on the build rules if ELCFILES is empty.
  # This is necessary because overriding ELCFILES= is a documented
  # idiom to disable byte-compilation.
  if (variable_value ('ELCFILES'))
    {
      # It's important that all depends on elc-stamp so that
      # all .elc files get recompiled whenever a .el changes.
      # It's important that all depends on $(ELCFILES) so that
      # we can recover if any of them is deleted.
      push (@all, 'elc-stamp', '$(ELCFILES)');
    }

  require_variables ("$am_file.am", "Emacs Lisp sources seen", 'TRUE',
		     'EMACS', 'lispdir');
  require_conf_file ("$am_file.am", FOREIGN, 'elisp-comp');
  &define_variable ('elisp_comp', $config_aux_dir . '/elisp-comp');
}

# Handle Python
sub handle_python
{
  my @pyfiles = &am_install_var ('-defaultdist', 'python', 'PYTHON',
				 'noinst');
  return if ! @pyfiles;

  require_variables ("$am_file.am", "Python sources seen", 'TRUE',
		     'PYTHON');
  require_conf_file ("$am_file.am", FOREIGN, 'py-compile');
  &define_variable ('py_compile', $config_aux_dir . '/py-compile');
}

# Handle Java.
sub handle_java
{
    my @sourcelist = &am_install_var ('-candist',
				      'java', 'JAVA',
				      'java', 'noinst', 'check');
    return if ! @sourcelist;

    my @prefix = am_primary_prefixes ('JAVA', 1,
				      'java', 'noinst', 'check');

    my $dir;
    foreach my $curs (@prefix)
      {
	next
	  if $curs eq 'EXTRA';

	err_var "${curs}_JAVA", "multiple _JAVA primaries in use"
	  if defined $dir;
	$dir = $curs;
      }


    push (@all, 'class' . $dir . '.stamp');
}


# Handle some of the minor options.
sub handle_minor_options
{
  if (defined $options{'readme-alpha'})
    {
      if ($relative_dir eq '.')
	{
	  if ($package_version !~ /^$GNITS_VERSION_PATTERN$/)
	    {
	      msg ('error-gnits', $package_version_location,
		   "version `$package_version' doesn't follow " .
		   "Gnits standards");
	    }
	  if (defined $1 && -f 'README-alpha')
	    {
	      # This means we have an alpha release.  See
	      # GNITS_VERSION_PATTERN for details.
	      push_dist_common ('README-alpha');
	    }
	}
    }
}

################################################################

# ($OUTPUT, @INPUTS)
# &split_config_file_spec ($SPEC)
# -------------------------------
# Decode the Autoconf syntax for config files (files, headers, links
# etc.).
sub split_config_file_spec ($)
{
  my ($spec) = @_;
  my ($output, @inputs) = split (/:/, $spec);

  push @inputs, "$output.in"
    unless @inputs;

  return ($output, @inputs);
}


my %make_list;

# &scan_autoconf_config_files ($CONFIG-FILES)
# -------------------------------------------
# Study $CONFIG-FILES which is the first argument to AC_CONFIG_FILES
# (or AC_OUTPUT).
sub scan_autoconf_config_files
{
    my ($config_files) = @_;
    # Look at potential Makefile.am's.
    foreach (split ' ', $config_files)
    {
        # Must skip empty string for Perl 4.
        next if $_ eq "\\" || $_ eq '';

        # Handle $local:$input syntax.  Note that we ignore
        # every input file past the first, though we keep
        # those around for later.
        my ($local, $input, @rest) = split (/:/);
        if (! $input)
        {
            $input = $local;
        }
        else
        {
            # FIXME: should be error if .in is missing.
            $input =~ s/\.in$//;
        }

        if (-f $input . '.am')
        {
            # We have a file that automake should generate.
            $make_list{$input} = join (':', ($local, @rest));
        }
        else
        {
            # We have a file that automake should cause to be
            # rebuilt, but shouldn't generate itself.
            push (@other_input_files, $_);
        }
    }
}


# &scan_autoconf_traces ($FILENAME)
# ---------------------------------
sub scan_autoconf_traces ($)
{
  my ($filename) = @_;

  my %traced = (
		AC_CANONICAL_HOST => 0,
		AC_CANONICAL_SYSTEM => 0,
		AC_CONFIG_AUX_DIR => 1,
		AC_CONFIG_FILES => 1,
		AC_CONFIG_HEADERS => 1,
		AC_INIT => 0,
		AC_LIBSOURCE => 1,
		AC_SUBST => 1,
		AM_AUTOMAKE_VERSION => 1,
		AM_CONDITIONAL => 2,
		AM_GNU_GETTEXT => 0,
		AM_INIT_AUTOMAKE => 0,
		AM_MAINTAINER_MODE => 0,
		AM_PROG_CC_C_O => 0,
		);

  my $traces = ($ENV{AUTOCONF} || 'autoconf') . " ";

  # Use a separator unlikely to be used, not `:', the default, which
  # has a precise meaning for AC_CONFIG_FILES and so on.
  $traces .= join (' ',
		   map { "--trace=$_" . ':\$f:\$l::\$n::\${::}%' }
		   (keys %traced));

  my $tracefh = new Automake::XFile ("$traces $filename |");
  verb "reading $traces";

  while ($_ = $tracefh->getline)
    {
      chomp;
      my ($here, @args) = split /::/;
      my $macro = $args[0];

      prog_error ("unrequested trace `$macro'")
	unless exists $traced{$macro};

      # Skip and diagnose malformed calls.
      if ($#args < $traced{$macro})
	{
	  msg ('syntax', $here, "not enough arguments for $macro");
	  next;
	}

      # Alphabetical ordering please.
      if ($macro eq 'AC_CANONICAL_HOST')
	{
	  if (! $seen_canonical)
	    {
	      $seen_canonical = AC_CANONICAL_HOST;
	      $canonical_location = $here;
	    };
	}
      elsif ($macro eq 'AC_CANONICAL_SYSTEM')
	{
	  $seen_canonical = AC_CANONICAL_SYSTEM;
	  $canonical_location = $here;
	}
      elsif ($macro eq 'AC_CONFIG_AUX_DIR')
	{
	  @config_aux_path = $args[1];
	  $config_aux_dir_set_in_configure_in = 1;
	}
      elsif ($macro eq 'AC_CONFIG_FILES')
	{
	  # Look at potential Makefile.am's.
	  $ac_config_files_location = $here;
	  &scan_autoconf_config_files ($args[1]);
	}
      elsif ($macro eq 'AC_CONFIG_HEADERS')
	{
	  $config_header_location = $here;
	  push @config_headers, split (' ', $args[1]);
	}
      elsif ($macro eq 'AC_INIT')
        {
	  if (defined $args[2])
	    {
	      $package_version = $args[2];
	      $package_version_location = $here;
	    }
	}
      elsif ($macro eq 'AC_LIBSOURCE')
	{
	  $libsources{$args[1]} = $here;
	}
      elsif ($macro eq 'AC_SUBST')
	{
	  # Just check for alphanumeric in AC_SUBST.  If you do
	  # AC_SUBST(5), then too bad.
	  $configure_vars{$args[1]} = $here
	    if $args[1] =~ /^\w+$/;
	}
      elsif ($macro eq 'AM_AUTOMAKE_VERSION')
        {
	  error ($here,
		 "version mismatch.  This is Automake $VERSION,\n" .
		 "but the definition used by this AM_INIT_AUTOMAKE\n" .
		 "comes from Automake $args[1].  You should recreate\n" .
		 "aclocal.m4 with aclocal and run automake again.\n")
	    if $VERSION ne $args[1];

	  $seen_automake_version = 1;
        }
      elsif ($macro eq 'AM_CONDITIONAL')
	{
	  $configure_cond{$args[1]} = $here;
	}
      elsif ($macro eq 'AM_GNU_GETTEXT')
	{
	  $seen_gettext = $here;
	  $ac_gettext_location = $here;
	  $seen_gettext_external = grep ($_ eq 'external', @args);
	}
      elsif ($macro eq 'AM_INIT_AUTOMAKE')
	{
	  $seen_init_automake = $here;
	  if (defined $args[2])
	    {
	      $package_version = $args[2];
	      $package_version_location = $here;
	    }
	  elsif (defined $args[1])
	    {
	      $global_options = $args[1];
	    }
	}
      elsif ($macro eq 'AM_MAINTAINER_MODE')
	{
	  $seen_maint_mode = $here;
	}
      elsif ($macro eq 'AM_PROG_CC_C_O')
	{
	  $seen_cc_c_o = $here;
	}
   }
}


# &scan_autoconf_files ()
# -----------------------
# Check whether we use `configure.ac' or `configure.in'.
# Scan it (and possibly `aclocal.m4') for interesting things.
# We must scan aclocal.m4 because there might be AC_SUBSTs and such there.
sub scan_autoconf_files
{
    # Reinitialize libsources here.  This isn't really necessary,
    # since we currently assume there is only one configure.ac.  But
    # that won't always be the case.
    %libsources = ();

    $configure_ac = find_configure_ac;
    fatal "`configure.ac' or `configure.in' is required\n"
        if !$configure_ac;

    scan_autoconf_traces ($configure_ac);

    # Set input and output files if not specified by user.
    if (! @input_files)
    {
	@input_files = sort keys %make_list;
	%output_files = %make_list;
    }

    @configure_input_files = sort keys %make_list;

    if (! $seen_init_automake)
      {
	err_ac ("no proper invocation of AM_INIT_AUTOMAKE was found.\nYou "
		. "should verify that $configure_ac invokes AM_INIT_AUTOMAKE,"
		. "\nthat aclocal.m4 is present in the top-level directory,\n"
		. "and that aclocal.m4 was recently regenerated "
		. "(using aclocal).");
      }
    else
      {
	if (! $seen_automake_version)
	  {
	    if (-f 'aclocal.m4')
	      {
		error ($seen_init_automake,
		       "your implementation of AM_INIT_AUTOMAKE comes from " .
		       "an\nold Automake version.  You should recreate " .
		       "aclocal.m4\nwith aclocal and run automake again.\n");
	      }
	    else
	      {
		error ($seen_init_automake,
		       "no proper implementation of AM_INIT_AUTOMAKE was " .
		       "found,\nprobably because aclocal.m4 is missing...\n" .
		       "You should run aclocal to create this file, then\n" .
		       "run automake again.\n");
	      }
	  }
      }

    # Look for some files we need.  Always check for these.  This
    # check must be done for every run, even those where we are only
    # looking at a subdir Makefile.  We must set relative_dir so that
    # the file-finding machinery works.
    # FIXME: Is this broken because it needs dynamic scopes.
    # My tests seems to show it's not the case.
    $relative_dir = '.';
    require_conf_file ($configure_ac, FOREIGN,
		       'install-sh', 'mkinstalldirs', 'missing');
    err_am "`install.sh' is an anachronism; use `install-sh' instead"
      if -f $config_aux_path[0] . '/install.sh';

    # Preserve dist_common for later.
    $configure_dist_common = variable_value ('DIST_COMMON', 'TRUE') || '';
}

################################################################

# Set up for Cygnus mode.
sub check_cygnus
{
  return unless $cygnus_mode;

  &set_strictness ('foreign');
  $options{'no-installinfo'} = $cygnus_mode;
  $options{'no-dependencies'} = $cygnus_mode;
  $use_dependencies = 0;

  err_ac "`AM_MAINTAINER_MODE' required when --cygnus specified"
    if !$seen_maint_mode;
}

# Do any extra checking for GNU standards.
sub check_gnu_standards
{
  if ($relative_dir eq '.')
    {
      # In top level (or only) directory.

      # Accept one of these three licenses; default to COPYING.
      my $license = 'COPYING';
      foreach (qw /COPYING.LIB COPYING.LESSER/)
	{
	  $license = $_ if -f $_;
	}
      require_file ("$am_file.am", GNU, $license,
		    qw/INSTALL NEWS README AUTHORS ChangeLog/);
    }

  for my $opt ('no-installman', 'no-installinfo')
    {
      msg ('error-gnu', $options{$opt},
	   "option `$opt' disallowed by GNU standards")
	if (defined $options{$opt});
    }
}

# Do any extra checking for GNITS standards.
sub check_gnits_standards
{
  if ($relative_dir eq '.')
    {
      # In top level (or only) directory.
      require_file ("$am_file.am", GNITS, 'THANKS');
    }
}

################################################################
#
# Functions to handle files of each language.

# Each `lang_X_rewrite($DIRECTORY, $BASE, $EXT)' function follows a
# simple formula: Return value is LANG_SUBDIR if the resulting object
# file should be in a subdir if the source file is, LANG_PROCESS if
# file is to be dealt with, LANG_IGNORE otherwise.

# Much of the actual processing is handled in
# handle_single_transform_list.  These functions exist so that
# auxiliary information can be recorded for a later cleanup pass.
# Note that the calls to these functions are computed, so don't bother
# searching for their precise names in the source.

# This is just a convenience function that can be used to determine
# when a subdir object should be used.
sub lang_sub_obj
{
    return defined $options{'subdir-objects'} ? LANG_SUBDIR : LANG_PROCESS;
}

# Rewrite a single C source file.
sub lang_c_rewrite
{
  my ($directory, $base, $ext) = @_;

  if (defined $options{'ansi2knr'} && $base =~ /_$/)
    {
      # FIXME: include line number in error.
      err_am "C source file `$base.c' would be deleted by ansi2knr rules";
    }

  my $r = LANG_PROCESS;
  if (defined $options{'subdir-objects'})
    {
      $r = LANG_SUBDIR;
      $base = $directory . '/' . $base
	unless $directory eq '.' || $directory eq '';

      err_am ("C objects in subdir but `AM_PROG_CC_C_O' "
	      . "not in `$configure_ac'",
	      uniq_scope => US_GLOBAL)
	unless $seen_cc_c_o;

      require_conf_file ("$am_file.am", FOREIGN, 'compile');

      # In this case we already have the directory information, so
      # don't add it again.
      $de_ansi_files{$base} = '';
    }
  else
    {
      $de_ansi_files{$base} = (($directory eq '.' || $directory eq '')
			       ? ''
			       : "$directory/");
    }

    return $r;
}

# Rewrite a single C++ source file.
sub lang_cxx_rewrite
{
    return &lang_sub_obj;
}

# Rewrite a single header file.
sub lang_header_rewrite
{
    # Header files are simply ignored.
    return LANG_IGNORE;
}

# Rewrite a single yacc file.
sub lang_yacc_rewrite
{
    my ($directory, $base, $ext) = @_;

    my $r = &lang_sub_obj;
    (my $newext = $ext) =~ tr/y/c/;
    return ($r, $newext);
}

# Rewrite a single yacc++ file.
sub lang_yaccxx_rewrite
{
    my ($directory, $base, $ext) = @_;

    my $r = &lang_sub_obj;
    (my $newext = $ext) =~ tr/y/c/;
    return ($r, $newext);
}

# Rewrite a single lex file.
sub lang_lex_rewrite
{
    my ($directory, $base, $ext) = @_;

    my $r = &lang_sub_obj;
    (my $newext = $ext) =~ tr/l/c/;
    return ($r, $newext);
}

# Rewrite a single lex++ file.
sub lang_lexxx_rewrite
{
    my ($directory, $base, $ext) = @_;

    my $r = &lang_sub_obj;
    (my $newext = $ext) =~ tr/l/c/;
    return ($r, $newext);
}

# Rewrite a single assembly file.
sub lang_asm_rewrite
{
    return &lang_sub_obj;
}

# Rewrite a single Fortran 77 file.
sub lang_f77_rewrite
{
    return LANG_PROCESS;
}

# Rewrite a single preprocessed Fortran 77 file.
sub lang_ppf77_rewrite
{
    return LANG_PROCESS;
}

# Rewrite a single ratfor file.
sub lang_ratfor_rewrite
{
    return LANG_PROCESS;
}

# Rewrite a single Objective C file.
sub lang_objc_rewrite
{
    return &lang_sub_obj;
}

# Rewrite a single Java file.
sub lang_java_rewrite
{
    return LANG_SUBDIR;
}

# The lang_X_finish functions are called after all source file
# processing is done.  Each should handle defining rules for the
# language, etc.  A finish function is only called if a source file of
# the appropriate type has been seen.

sub lang_c_finish
{
    # Push all libobjs files onto de_ansi_files.  We actually only
    # push files which exist in the current directory, and which are
    # genuine source files.
    foreach my $file (keys %libsources)
    {
	if ($file =~ /^(.*)\.[cly]$/ && -f "$relative_dir/$file")
	{
	    $de_ansi_files{$1} = ''
	}
    }

    if (defined $options{'ansi2knr'} && keys %de_ansi_files)
    {
 	# Make all _.c files depend on their corresponding .c files.
 	my @objects;
 	foreach my $base (sort keys %de_ansi_files)
 	{
	    # Each _.c file must depend on ansi2knr; otherwise it
	    # might be used in a parallel build before it is built.
	    # We need to support files in the srcdir and in the build
	    # dir (because these files might be auto-generated.  But
	    # we can't use $< -- some makes only define $< during a
	    # suffix rule.
	    my $ansfile = $de_ansi_files{$base} . $base . '.c';
 	    $output_rules .= ($base . "_.c: $ansfile \$(ANSI2KNR)\n\t"
			      . '$(CPP) $(DEFS) $(DEFAULT_INCLUDES) $(INCLUDES) $(AM_CPPFLAGS) $(CPPFLAGS) '
			      . '`if test -f $(srcdir)/' . $ansfile
			      . '; then echo $(srcdir)/' . $ansfile
			      . '; else echo ' . $ansfile . '; fi` '
			      . "| sed 's/^# \\([0-9]\\)/#line \\1/' "
			      . '| $(ANSI2KNR) > $@'
			      # If ansi2knr fails then we shouldn't
			      # create the _.c file
			      . " || rm -f \$\@\n");
 	    push (@objects, $base . '_.$(OBJEXT)');
 	    push (@objects, $base . '_.lo')
	      if variable_defined ('LIBTOOL');
 	}

 	# Make all _.o (and _.lo) files depend on ansi2knr.
 	# Use a sneaky little hack to make it print nicely.
 	&pretty_print_rule ('', '', @objects, ':', '$(ANSI2KNR)');
    }
}

# This is a yacc helper which is called whenever we have decided to
# compile a yacc file.
sub lang_yacc_target_hook
{
    my ($self, $aggregate, $output, $input) = @_;

    my $flag = $aggregate . "_YFLAGS";
    if ((variable_defined ($flag)
	 && &variable_value ($flag) =~ /$DASH_D_PATTERN/o)
	|| (variable_defined ('YFLAGS')
	    && &variable_value ('YFLAGS') =~ /$DASH_D_PATTERN/o))
    {
	(my $output_base = $output) =~ s/$KNOWN_EXTENSIONS_PATTERN$//;
	my $header = $output_base . '.h';

	# Found a `-d' that applies to the compilation of this file.
	# Add a dependency for the generated header file, and arrange
	# for that file to be included in the distribution.
	# FIXME: this fails for `nodist_*_SOURCES'.
	$output_rules .= ("${header}: $output\n"
			  # Recover from removal of $header
			  . "\t\@if test ! -f \$@; then \\\n"
			  . "\t  rm -f $output; \\\n"
			  . "\t  \$(MAKE) $output; \\\n"
			  . "\telse :; fi\n");
	&push_dist_common ($header);
	# If the files are built in the build directory, then we want
	# to remove them with `make clean'.  If they are in srcdir
	# they shouldn't be touched.  However, we can't determine this
	# statically, and the GNU rules say that yacc/lex output files
	# should be removed by maintainer-clean.  So that's what we
	# do.
	$clean_files{$header} = MAINTAINER_CLEAN;
    }
    # Erase $OUTPUT on `make maintainer-clean' (by GNU standards).
    # See the comment above for $HEADER.
    $clean_files{$output} = MAINTAINER_CLEAN;
}

# This is a lex helper which is called whenever we have decided to
# compile a lex file.
sub lang_lex_target_hook
{
    my ($self, $aggregate, $output, $input) = @_;
    # If the files are built in the build directory, then we want to
    # remove them with `make clean'.  If they are in srcdir they
    # shouldn't be touched.  However, we can't determine this
    # statically, and the GNU rules say that yacc/lex output files
    # should be removed by maintainer-clean.  So that's what we do.
    $clean_files{$output} = MAINTAINER_CLEAN;
}

# This is a helper for both lex and yacc.
sub yacc_lex_finish_helper
{
    return if defined $language_scratch{'lex-yacc-done'};
    $language_scratch{'lex-yacc-done'} = 1;

    # If there is more than one distinct yacc (resp lex) source file
    # in a given directory, then the `ylwrap' program is required to
    # allow parallel builds to work correctly.  FIXME: for now, no
    # line number.
    require_conf_file ($configure_ac, FOREIGN, 'ylwrap');
    if ($config_aux_dir_set_in_configure_in)
    {
	&define_variable ('YLWRAP', $config_aux_dir . "/ylwrap");
    }
    else
    {
	&define_variable ('YLWRAP', '$(top_srcdir)/ylwrap');
    }
}

sub lang_yacc_finish
{
  return if defined $language_scratch{'yacc-done'};
  $language_scratch{'yacc-done'} = 1;

  reject_var 'YACCFLAGS', "`YACCFLAGS' obsolete; use `YFLAGS' instead";

  &yacc_lex_finish_helper
    if count_files_for_language ('yacc') > 1;
}


sub lang_lex_finish
{
  return if defined $language_scratch{'lex-done'};
  $language_scratch{'lex-done'} = 1;

  &yacc_lex_finish_helper
    if count_files_for_language ('lex') > 1;
}


# Given a hash table of linker names, pick the name that has the most
# precedence.  This is lame, but something has to have global
# knowledge in order to eliminate the conflict.  Add more linkers as
# required.
sub resolve_linker
{
    my (%linkers) = @_;

    foreach my $l (qw(GCJLINK CXXLINK F77LINK OBJCLINK))
    {
	return $l if defined $linkers{$l};
    }
    return 'LINK';
}

# Called to indicate that an extension was used.
sub saw_extension
{
    my ($ext) = @_;
    if (! defined $extension_seen{$ext})
    {
	$extension_seen{$ext} = 1;
    }
    else
    {
	++$extension_seen{$ext};
    }
}

# Return the number of files seen for a given language.  Knows about
# special cases we care about.  FIXME: this is hideous.  We need
# something that involves real language objects.  For instance yacc
# and yaccxx could both derive from a common yacc class which would
# know about the strange ylwrap requirement.  (Or better yet we could
# just not support legacy yacc!)
sub count_files_for_language
{
    my ($name) = @_;

    my @names;
    if ($name eq 'yacc' || $name eq 'yaccxx')
    {
	@names = ('yacc', 'yaccxx');
    }
    elsif ($name eq 'lex' || $name eq 'lexxx')
    {
	@names = ('lex', 'lexxx');
    }
    else
    {
	@names = ($name);
    }

    my $r = 0;
    foreach $name (@names)
    {
	my $lang = $languages{$name};
	foreach my $ext (@{$lang->extensions})
	{
	    $r += $extension_seen{$ext}
	        if defined $extension_seen{$ext};
	}
    }

    return $r
}

# Called to ask whether source files have been seen . If HEADERS is 1,
# headers can be included.
sub saw_sources_p
{
    my ($headers) = @_;

    # count all the sources
    my $count = 0;
    foreach my $val (values %extension_seen)
    {
	$count += $val;
    }

    if (!$headers)
    {
	$count -= count_files_for_language ('header');
    }

    return $count > 0;
}


# register_language (%ATTRIBUTE)
# ------------------------------
# Register a single language.
# Each %ATTRIBUTE is of the form ATTRIBUTE => VALUE.
sub register_language (%)
{
  my (%option) = @_;

  # Set the defaults.
  $option{'ansi'} = 0
    unless defined $option{'ansi'};
  $option{'autodep'} = 'no'
    unless defined $option{'autodep'};
  $option{'linker'} = ''
    unless defined $option{'linker'};
  $option{'flags'} = []
    unless defined $option{'flags'};
  $option{'output_extensions'} = sub { return ( '.$(OBJEXT)', '.lo' ) }
    unless defined $option{'output_extensions'};

  my $lang = new Language (%option);

  # Fill indexes.
  grep ($extension_map{$_} = $lang->name, @{$lang->extensions});
  $languages{$lang->name} = $lang;

  # Update the pattern of known extensions.
  accept_extensions (@{$lang->extensions});

  # Upate the $suffix_rule map.
  foreach my $suffix (@{$lang->extensions})
    {
      foreach my $dest (&{$lang->output_extensions} ($suffix))
	{
	  &register_suffix_rule ('internal', $suffix, $dest);
	}
    }
}

# derive_suffix ($EXT, $OBJ)
# --------------------------
# This function is used to find a path from a user-specified suffix $EXT
# to $OBJ or to some other suffix we recognize internally, eg `cc'.
sub derive_suffix ($$)
{
  my ($source_ext, $obj) = @_;

  while (! $extension_map{$source_ext}
	 && $source_ext ne $obj
	 && exists $suffix_rules->{$source_ext}
	 && exists $suffix_rules->{$source_ext}{$obj})
    {
      $source_ext = $suffix_rules->{$source_ext}{$obj}[0];
    }

  return $source_ext;
}


################################################################

# Pretty-print something.  HEAD is what should be printed at the
# beginning of the first line, FILL is what should be printed at the
# beginning of every subsequent line.
sub pretty_print_internal
{
    my ($head, $fill, @values) = @_;

    my $column = length ($head);
    my $result = $head;

    # Fill length is number of characters.  However, each Tab
    # character counts for eight.  So we count the number of Tabs and
    # multiply by 7.
    my $fill_length = length ($fill);
    $fill_length += 7 * ($fill =~ tr/\t/\t/d);

    foreach (@values)
    {
	# "71" because we also print a space.
	if ($column + length ($_) > 71)
	{
	    $result .= " \\\n" . $fill;
	    $column = $fill_length;
	}
	$result .= ' ' if $result =~ /\S\z/;
	$result .= $_;
	$column += length ($_) + 1;
    }

    $result .= "\n";
    return $result;
}

# Pretty-print something and append to output_vars.
sub pretty_print
{
    $output_vars .= &pretty_print_internal (@_);
}

# Pretty-print something and append to output_rules.
sub pretty_print_rule
{
    $output_rules .= &pretty_print_internal (@_);
}


################################################################


# $STRING
# &conditional_string(@COND-STACK)
# --------------------------------
# Build a string which denotes the conditional in @COND-STACK.  Some
# simplifications are done: `TRUE' entries are elided, and any `FALSE'
# entry results in a return of `FALSE'.
sub conditional_string
{
  my (@stack) = @_;

  if (grep (/^FALSE$/, @stack))
    {
      return 'FALSE';
    }
  else
    {
      return join (' ', uniq sort grep (!/^TRUE$/, @stack));
    }
}


# $BOOLEAN
# &conditional_true_when ($COND, $WHEN)
# -------------------------------------
# See if a conditional is true.  Both arguments are conditional
# strings.  This returns true if the first conditional is true when
# the second conditional is true.
# For instance with $COND = `BAR FOO', and $WHEN = `BAR BAZ FOO',
# obviously return 1, and 0 when, for instance, $WHEN = `FOO'.
sub conditional_true_when ($$)
{
    my ($cond, $when) = @_;

    # Make a hash holding all the values from $WHEN.
    my %cond_vals = map { $_ => 1 } split (' ', $when);

    # Nothing is true when FALSE (not even FALSE itself, but it
    # shouldn't hurt if you decide to change that).
    return 0 if exists $cond_vals{'FALSE'};

    # Check each component of $cond, which looks `COND1 COND2'.
    foreach my $comp (split (' ', $cond))
    {
	# TRUE is always true.
	next if $comp eq 'TRUE';
	return 0 if ! defined $cond_vals{$comp};
    }

    return 1;
}


# $BOOLEAN
# &conditional_is_redundant ($COND, @WHENS)
# ----------------------------------------
# Determine whether $COND is redundant with respect to @WHENS.
#
# Returns true if $COND is true for any of the conditions in @WHENS.
#
# If there are no @WHENS, then behave as if @WHENS contained a single empty
# condition.
sub conditional_is_redundant ($@)
{
    my ($cond, @whens) = @_;

    @whens = ("") if @whens == 0;

    foreach my $when (@whens)
    {
	return 1 if conditional_true_when ($cond, $when);
    }
    return 0;
}


# $BOOLEAN
# &conditional_implies_any ($COND, @CONDS)
# ----------------------------------------
# Returns true iff $COND implies any of the conditions in @CONDS.
sub conditional_implies_any ($@)
{
    my ($cond, @conds) = @_;

    @conds = ("") if @conds == 0;

    foreach my $c (@conds)
    {
	return 1 if conditional_true_when ($c, $cond);
    }
    return 0;
}


# $NEGATION
# condition_negate ($COND)
# ------------------------
sub condition_negate ($)
{
    my ($cond) = @_;

    $cond =~ s/TRUE$/TRUEO/;
    $cond =~ s/FALSE$/TRUE/;
    $cond =~ s/TRUEO$/FALSE/;

    return $cond;
}


# Compare condition names.
# Issue them in alphabetical order, foo_TRUE before foo_FALSE.
sub by_condition
{
    # Be careful we might be comparing `' or `#'.
    $a =~ /^(.*)_(TRUE|FALSE)$/;
    my ($aname, $abool) = ($1 || '', $2 || '');
    $b =~ /^(.*)_(TRUE|FALSE)$/;
    my ($bname, $bbool) = ($1 || '', $2 || '');
    return ($aname cmp $bname
	    # Don't bother with IFs, given that TRUE is after FALSE
	    # just cmp in the reverse order.
	    || $bbool cmp $abool
	    # Just in case...
	    || $a cmp $b);
}


# &make_condition (@CONDITIONS)
# -----------------------------
# Transform a list of conditions (themselves can be an internal list
# of conditions, e.g., @CONDITIONS = ('cond1 cond2', 'cond3')) into a
# Make conditional (a pattern for AC_SUBST).
# Correctly returns the empty string when there are no conditions.
sub make_condition
{
    my $res = conditional_string (@_);

    # There are no conditions.
    if ($res eq '')
      {
	# Nothing to do.
      }
    # It's impossible.
    elsif ($res eq 'FALSE')
      {
	$res = '#';
      }
    # Build it.
    else
      {
	$res = '@' . $res . '@';
	$res =~ s/ /@@/g;
      }

    return $res;
}



## ------------------------------ ##
## Handling the condition stack.  ##
## ------------------------------ ##


# $COND_STRING
# cond_stack_if ($NEGATE, $COND, $WHERE)
# --------------------------------------
sub cond_stack_if ($$$)
{
  my ($negate, $cond, $where) = @_;

  error $where, "$cond does not appear in AM_CONDITIONAL"
    if ! $configure_cond{$cond} && $cond !~ /^TRUE|FALSE$/;

  $cond = "${cond}_TRUE"
    unless $cond =~ /^TRUE|FALSE$/;
  $cond = condition_negate ($cond)
    if $negate;

  push (@cond_stack, $cond);

  return conditional_string (@cond_stack);
}


# $COND_STRING
# cond_stack_else ($NEGATE, $COND, $WHERE)
# ----------------------------------------
sub cond_stack_else ($$$)
{
  my ($negate, $cond, $where) = @_;

  if (! @cond_stack)
    {
      error $where, "else without if";
      return;
    }

  $cond_stack[$#cond_stack] = condition_negate ($cond_stack[$#cond_stack]);

  # If $COND is given, check against it.
  if (defined $cond)
    {
      $cond = "${cond}_TRUE"
	unless $cond =~ /^TRUE|FALSE$/;
      $cond = condition_negate ($cond)
	if $negate;

      error ($where, "else reminder ($negate$cond) incompatible with "
	     . "current conditional: $cond_stack[$#cond_stack]")
	if $cond_stack[$#cond_stack] ne $cond;
    }

  return conditional_string (@cond_stack);
}


# $COND_STRING
# cond_stack_endif ($NEGATE, $COND, $WHERE)
# -----------------------------------------
sub cond_stack_endif ($$$)
{
  my ($negate, $cond, $where) = @_;
  my $old_cond;

  if (! @cond_stack)
    {
      error $where, "endif without if: $negate$cond";
      return;
    }


  # If $COND is given, check against it.
  if (defined $cond)
    {
      $cond = "${cond}_TRUE"
	unless $cond =~ /^TRUE|FALSE$/;
      $cond = condition_negate ($cond)
	if $negate;

      error ($where, "endif reminder ($negate$cond) incompatible with "
	     . "current conditional: $cond_stack[$#cond_stack]")
	if $cond_stack[$#cond_stack] ne $cond;
    }

  pop @cond_stack;

  return conditional_string (@cond_stack);
}





## ------------------------ ##
## Handling the variables.  ##
## ------------------------ ##


# check_ambiguous_conditional ($VAR, $COND, $WHERE)
# -------------------------------------------------
# Check for an ambiguous conditional.  This is called when a variable
# is being defined conditionally.  If we already know about a
# definition that is true under the same conditions, then we have an
# ambiguity.
sub check_ambiguous_conditional ($$$)
{
  my ($var, $cond, $where) = @_;
  my ($message, $ambig_cond) =
    conditional_ambiguous_p ($var, $cond, keys %{$var_value{$var}});
  if ($message)
    {
      msg 'syntax', $where, "$message ...";
      msg_var ('syntax', $var, "... `$var' previously defined here.");
      verb (macro_dump ($var));
    }
}

# $STRING, $AMBIG_COND
# conditional_ambiguous_p ($WHAT, $COND, @CONDS)
# ----------------------------------------------
# Check for an ambiguous conditional.  Return an error message and
# the other condition involved if we have one, two empty strings otherwise.
#   WHAT:  the thing being defined
#   COND:  the condition under which is is being defined
#   CONDS: the conditons under which is had already been defined
sub conditional_ambiguous_p ($$@)
{
  my ($var, $cond, @conds) = @_;
  foreach my $vcond (@conds)
    {
      # Note that these rules doesn't consider the following
      # example as ambiguous.
      #
      #   if COND1
      #     FOO = foo
      #   endif
      #   if COND2
      #     FOO = bar
      #   endif
      #
      # It's up to the user to not define COND1 and COND2
      # simultaneously.
      my $message;
      if ($vcond eq $cond)
	{
	  return ("$var multiply defined in condition $cond", $vcond);
	}
      elsif (&conditional_true_when ($vcond, $cond))
	{
	  return ("$var was already defined in condition $vcond, "
		  . "which implies condition $cond", $vcond);
	}
      elsif (&conditional_true_when ($cond, $vcond))
	{
	  return ("$var was already defined in condition $vcond, "
		   . "which is implied by condition $cond", $vcond);
	}
    }
  return ('', '');
}

# @MISSING_CONDS
# variable_not_always_defined_in_cond ($VAR, $COND)
# ---------------------------------------------
# Check whether $VAR is always defined for condition $COND.
# Return a list of conditions where the definition is missing.
#
# For instance, given
#
#   if COND1
#     if COND2
#       A = foo
#       D = d1
#     else
#       A = bar
#       D = d2
#     endif
#   else
#     D = d3
#   endif
#   if COND3
#     A = baz
#     B = mumble
#   endif
#   C = mumble
#
# we should have:
#   variable_not_always_defined_in_cond ('A', 'COND1_TRUE COND2_TRUE')
#     => ()
#   variable_not_always_defined_in_cond ('A', 'COND1_TRUE')
#     => ()
#   variable_not_always_defined_in_cond ('A', 'TRUE')
#     => ("COND1_FALSE COND2_FALSE COND3_FALSE",
#         "COND1_FALSE COND2_TRUE COND3_FALSE",
#         "COND1_TRUE COND2_FALSE COND3_FALSE",
#         "COND1_TRUE COND2_TRUE COND3_FALSE")
#   variable_not_always_defined_in_cond ('B', 'COND1_TRUE')
#     => ("COND3_FALSE")
#   variable_not_always_defined_in_cond ('C', 'COND1_TRUE')
#     => ()
#   variable_not_always_defined_in_cond ('D', 'TRUE')
#     => ()
#   variable_not_always_defined_in_cond ('Z', 'TRUE')
#     => ("TRUE")
#
sub variable_not_always_defined_in_cond ($$)
{
  my ($var, $cond) = @_;

  # It's easy to answer if the variable is not defined.
  return ("TRUE",) unless exists $var_value{$var};

  # How does it work?  Let's take the second example:
  #
  #   variable_not_always_defined_in_cond ('A', 'COND1_TRUE')
  #
  # (1) First, we get the list of conditions where A is defined:
  #
  #   ("COND1_TRUE COND2_TRUE", "COND1_TRUE COND2_FALSE", "COND3_TRUE")
  #
  # (2) Then we generate the set of inverted conditions:
  #
  #   ("COND1_FALSE COND2_TRUE COND3_FALSE",
  #    "COND1_FALSE COND2_FALSE COND3_FALSE")
  #
  # (3) Finally we remove these conditions which are not implied by
  #     COND1_TRUE.  This yields an empty list and we are done.

  my @res = ();
  my @cond_defs = keys %{$var_value{$var}}; # (1)
  foreach my $icond (invert_conditions (@cond_defs)) # (2)
    {
      prog_error "invert_conditions returned an input condition"
	if exists $var_value{$var}{$icond};

      push @res, $icond
	if (conditional_true_when ($cond, $icond)); # (3)
    }
  return @res;
}

# &macro_define($VAR, $OWNER, $TYPE, $COND, $VALUE, $WHERE)
# -------------------------------------------------------------
# The $VAR can go from Automake to user, but not the converse.
sub macro_define ($$$$$$)
{
  my ($var, $owner, $type, $cond, $value, $where) = @_;

  # We will adjust the owener of this variable unless told otherwise.
  my $adjust_owner = 1;

  error $where, "bad characters in variable name `$var'"
    if $var !~ /$MACRO_PATTERN/o;

  # NEWS-OS 4.2R complains if a Makefile variable begins with `_'.
  msg ('portability', $where,
       "$var: variable names starting with `_' are not portable")
    if $var =~ /^_/;

  # `:='-style assignments are not acknowledged by POSIX.  Moreover it
  # has multiple meanings.  In GNU make or BSD make it means "assign
  # with immediate expansion", while in OSF make it is used for
  # conditional assignments.
  msg ('portability', $where, "`:='-style assignments are not portable")
    if $type eq ':';

  check_variable_expansions ($value, $where);

  $cond ||= 'TRUE';

  # An Automake variable must be consistently defined with the same
  # sign by Automake.  A user variable must be set by either `=' or
  # `:=', and later promoted to `+='.
  if ($owner == VAR_AUTOMAKE)
    {
      if (exists $var_type{$var}
	  && exists $var_type{$var}{$cond}
	  && $var_type{$var}{$cond} ne $type)
	{
	  error ($where, "$var was set with `$var_type{$var}=' "
		 . "and is now set with `$type='");
	}
    }
  else
    {
      if (!exists $var_type{$var} && $type eq '+')
	{
	  error $where, "$var must be set with `=' before using `+='";
	}
    }
  $var_type{$var}{$cond} = $type;

  # Differentiate assignment types.

  # 1. append (+=) to a variable defined for current condition
  if ($type eq '+' && exists $var_value{$var}{$cond})
    {
      if (chomp $var_value{$var}{$cond})
	{
	  # Insert a backslash before a trailing newline.
	  $var_value{$var}{$cond} .= "\\\n";
	}
      elsif ($var_value{$var}{$cond})
	{
	  # Insert a separator.
	  $var_value{$var}{$cond} .= ' ';
	}
       $var_value{$var}{$cond} .= $value;
    }
  # 2. append (+=) to a variable defined for *another* condition
  elsif ($type eq '+' && keys %{$var_value{$var}})
    {
      # * Generally, $cond is not TRUE.  For instance:
      #     FOO = foo
      #     if COND
      #       FOO += bar
      #     endif
      #   In this case, we declare an helper variable conditionally,
      #   and append it to FOO:
      #     FOO = foo $(am__append_1)
      #     @COND_TRUE@am__append_1 = bar
      #   Of course if FOO is defined under several conditions, we add
      #   $(am__append_1) to each definitions.
      #
      # * If $cond is TRUE, we don't need the helper variable.  E.g., in
      #     if COND1
      #       FOO = foo1
      #     else
      #       FOO = foo2
      #     endif
      #     FOO += bar
      #   we can add bar directly to all definition of FOO, and output
      #     @COND_TRUE@FOO = foo1 bar
      #     @COND_FALSE@FOO = foo2 bar

      # Do we need an helper variable?
      if ($cond ne 'TRUE')
        {
	    # Does the helper variable already exists?
	    my $key = "$var:$cond";
	    if (exists $appendvar{$key})
	      {
		# Yes, let's simply append to it.
		$var = $appendvar{$key};
		$owner = VAR_AUTOMAKE;
	      }
	    else
	      {
		# No, create it.
		my $num = 1 + keys (%appendvar);
		my $hvar = "am__append_$num";
		$appendvar{$key} = $hvar;
		&macro_define ($hvar, VAR_AUTOMAKE, '+',
			       $cond, $value, $where);
		push @var_list, $hvar;
		# Now HVAR is to be added to VAR.
		$value = "\$($hvar)";
	      }
	}

      # Add VALUE to all definitions of VAR.
      foreach my $vcond (keys %{$var_value{$var}})
        {
	  # We have a bit of error detection to do here.
	  # This:
	  #   if COND1
	  #     X = Y
	  #   endif
	  #   X += Z
	  # should be rejected because X is not defined for all conditions
	  # where `+=' applies.
	  my @undef_cond = variable_not_always_defined_in_cond $var, $cond;
	  if (@undef_cond != 0)
	    {
	      error ($where,
		     "Cannot apply `+=' because `$var' is not defined "
		     . "in\nthe following conditions:\n  "
		     . join ("\n  ", @undef_cond)
		     . "\nEither define `$var' in these conditions,"
		     . " or use\n`+=' in the same conditions as"
		     . " the definitions.");
	    }
	  else
	    {
	      &macro_define ($var, $owner, '+', $vcond, $value, $where);
	    }
	}
      # Don't adjust the owner.  The above &macro_define did it in the
      # right conditions.
      $adjust_owner = 0;
    }
  # 3. first assignment (=, :=, or +=)
  else
    {
      # If Automake tries to override a value specified by the user,
      # just don't let it do.
      if (exists $var_value{$var}{$cond}
	  && $var_owner{$var}{$cond} != VAR_AUTOMAKE
	  && $owner == VAR_AUTOMAKE)
	{
	  verb ("refusing to override the user definition of:\n"
		. macro_dump ($var)
		."with `$cond' => `$value'");
	}
      else
	{
	  # There must be no previous value unless the user is redefining
	  # an Automake variable or an AC_SUBST variable for an existing
	  # condition.
	  check_ambiguous_conditional ($var, $cond, $where)
	    unless (exists $var_owner{$var}{$cond}
		    && (($var_owner{$var}{$cond} == VAR_AUTOMAKE
			 && $owner != VAR_AUTOMAKE)
			|| $var_owner{$var}{$cond} == VAR_CONFIGURE));

	  $var_value{$var}{$cond} = $value;
	  # Assignments to a macro set its location.  We don't adjust
	  # locations for `+='.  Ideally I suppose we would associate
	  # line numbers with random bits of text.
	  $var_location{$var}{$cond} = $where;
	}
    }

  # The owner of a variable can only increase, because an Automake
  # variable can be given to the user, but not the converse.
  if ($adjust_owner &&
      (! exists $var_owner{$var}{$cond}
       || $owner > $var_owner{$var}{$cond}))
    {
      $var_owner{$var}{$cond} = $owner;
      # Always adjust the location when the owner changes (even for
      # `+=' statements).  The risk otherwise is to warn about
      # a VAR_MAKEFILE variable and locate it in configure.ac...
      $var_location{$var}{$cond} = $where;
    }

  # Call var_VAR_trigger if it's defined.
  # This hook helps to update some internal state *while*
  # parsing the file.  For instance the handling of SUFFIXES
  # requires this (see var_SUFFIXES_trigger).
  my $var_trigger = "var_${var}_trigger";
  &$var_trigger($type, $value) if defined &$var_trigger;
}


# &macro_delete ($VAR, [@CONDS])
# ------------------------------
# Forget about $VAR under the conditions @CONDS, or completely if
# @CONDS is empty.
sub macro_delete ($@)
{
  my ($var, @conds) = @_;

  if (!@conds)
    {
      delete $var_value{$var};
      delete $var_location{$var};
      delete $var_owner{$var};
      delete $var_comment{$var};
      delete $var_type{$var};
    }
  else
    {
      foreach my $cond (@conds)
	{
	  delete $var_value{$var}{$cond};
	  delete $var_location{$var}{$cond};
	  delete $var_owner{$var}{$cond};
	  delete $var_comment{$var}{$cond};
	  delete $var_type{$var}{$cond};
	}
    }
}


# &macro_dump ($VAR)
# ------------------
sub macro_dump ($)
{
  my ($var) = @_;
  my $text = '';

  if (!exists $var_value{$var})
    {
      $text = "  $var does not exist\n";
    }
  else
    {
      $text .= "  $var $var_type{$var}=\n  {\n";
      foreach my $vcond (sort by_condition keys %{$var_value{$var}})
	{
	  prog_error ("`$var' is a key in \$var_value, "
		      . "but not in \$var_owner\n")
	    unless exists $var_owner{$var}{$vcond};

	  my $var_owner;
	  if ($var_owner{$var}{$vcond} == VAR_AUTOMAKE)
	    {
	      $var_owner = 'Automake';
	    }
	  elsif ($var_owner{$var}{$vcond} == VAR_CONFIGURE)
	    {
	      $var_owner = 'Configure';
	    }
	  elsif ($var_owner{$var}{$vcond} == VAR_MAKEFILE)
	    {
	      $var_owner = 'Makefile';
	    }
	  else
	    {
	      prog_error ("unexpected value for `\$var_owner{$var}{$vcond}': "
			  . $var_owner{$var}{$vcond})
		unless defined $var_owner;
	    }

	  my $where = (defined $var_location{$var}{$vcond}
		       ? $var_location{$var}{$vcond} : "undefined");
	  $text .= "$var_comment{$var}{$vcond}"
	    if exists $var_comment{$var}{$vcond};
	  $text .= "    $vcond => $var_value{$var}{$vcond}\n";
	}
      $text .= "  }\n";
    }
  return $text;
}


# &macros_dump ()
# ---------------
sub macros_dump ()
{
  my ($var) = @_;

  my $text = "%var_value =\n{\n";
  foreach my $var (sort (keys %var_value))
    {
      $text .= macro_dump ($var);
    }
  $text .= "}\n";
  return $text;
}


# $BOOLEAN
# variable_defined ($VAR, [$COND])
# ---------------------------------
# See if a variable exists.  $VAR is the variable name, and $COND is
# the condition which we should check.  If no condition is given, we
# currently return true if the variable is defined under any
# condition.
sub variable_defined ($;$)
{
    my ($var, $cond) = @_;

    if (! exists $var_value{$var}
	|| (defined $cond && ! exists $var_value{$var}{$cond}))
      {
	# VAR is not defined.

	# Check there is no target defined with the name of the
	# variable we check.

	# adl> I'm wondering if this error still makes any sense today. I
	# adl> guess it was because targets and variables used to share
	# adl> the same namespace in older versions of Automake?
	# tom> While what you say is definitely part of it, I think it
	# tom> might also have been due to someone making a "spelling error"
	# tom> -- writing "foo:..." instead of "foo = ...".
	# tom> I'm not sure whether it is really worth diagnosing
	# tom> this sort of problem.  In the old days I used to add warnings
	# tom> and errors like this pretty randomly, based on bug reports I
	# tom> got.  But there's a plausible argument that I was trying
	# tom> too hard to prevent people from making mistakes.
	if (exists $targets{$var}
	    && (! defined $cond || exists $targets{$var}{$cond}))
	  {
	    for my $tcond ($cond || keys %{$targets{$var}})
	      {
		prog_error ("\$targets{$var}{$tcond} exists but "
			    . "\$target_owner doesn't")
		  unless exists $target_owner{$var}{$tcond};
		# Diagnose the first user target encountered, if any.
		# Restricting this test to user targets allows Automake
		# to create rules for things like `bin_PROGRAMS = LDADD'.
		if ($target_owner{$var}{$tcond} == TARGET_USER)
		  {
		    msg_cond_target ('syntax', $tcond, $var,
				     "`$var' is a target; "
				     . "expected a variable");
		    return 0;
		  }
	      }
	  }
	return 0;
      }

    # Even a var_value examination is good enough for us.  FIXME:
    # really should maintain examined status on a per-condition basis.
    $content_seen{$var} = 1;
    return 1;
}


# $BOOLEAN
# variable_assert ($VAR, $WHERE)
# ------------------------------
# Make sure a variable exists.  $VAR is the variable name, and $WHERE
# is the name of a macro which refers to $VAR.
sub variable_assert ($$)
{
  my ($var, $where) = @_;

  return 1
    if variable_defined $var;

  require_variables ($where, "variable `$var' is used", 'TRUE', $var);

  return 0;
}

# Mark a variable as examined.
sub examine_variable
{
  my ($var) = @_;
  variable_defined ($var);
}


# &variable_conditions_recursive ($VAR)
# -------------------------------------
# Return the set of conditions for which a variable is defined.

# If the variable is not defined conditionally, and is not defined in
# terms of any variables which are defined conditionally, then this
# returns the empty list.

# If the variable is defined conditionally, but is not defined in
# terms of any variables which are defined conditionally, then this
# returns the list of conditions for which the variable is defined.

# If the variable is defined in terms of any variables which are
# defined conditionally, then this returns a full set of permutations
# of the subvariable conditions.  For example, if the variable is
# defined in terms of a variable which is defined for COND_TRUE,
# then this returns both COND_TRUE and COND_FALSE.  This is
# because we will need to define the variable under both conditions.
sub variable_conditions_recursive ($)
{
    my ($var) = @_;

    %vars_scanned = ();

    my @new_conds = variable_conditions_recursive_sub ($var, '');

    # Now we want to return all permutations of the subvariable
    # conditions.
    my %allconds = ();
    foreach my $item (@new_conds)
    {
	foreach (split (' ', $item))
	{
	    s/^(.*)_(TRUE|FALSE)$/$1_TRUE/;
	    $allconds{$_} = 1;
	}
    }
    @new_conds = variable_conditions_permutations (sort keys %allconds);

    my %uniqify;
    foreach my $cond (@new_conds)
    {
	my $reduce = variable_conditions_reduce (split (' ', $cond));
        next
	    if $reduce eq 'FALSE';
	$uniqify{$cond} = 1;
    }

    # Note we cannot just do `return sort keys %uniqify', because this
    # function is sometimes used in a scalar context.
    my @uniq_list = sort by_condition keys %uniqify;
    return @uniq_list;
}


# @CONDS
# variable_conditions ($VAR)
# --------------------------
# Get the list of conditions that a variable is defined with, without
# recursing through the conditions of any subvariables.
# Argument is $VAR: the variable to get the conditions of.
# Returns the list of conditions.
sub variable_conditions ($)
{
    my ($var) = @_;
    my @conds = keys %{$var_value{$var}};
    return sort by_condition @conds;
}


# $BOOLEAN
# &variable_conditionally_defined ($VAR)
# --------------------------------------
sub variable_conditionally_defined ($)
{
    my ($var) = @_;
    foreach my $cond (variable_conditions_recursive ($var))
      {
	return 1
	  unless $cond =~ /^TRUE|FALSE$/;
      }
    return 0;
}

# @LIST
# &scan_variable_expansions ($TEXT)
# ---------------------------------
# Return the list of variable names expanded in $TEXT.
# Note that unlike some other functions, $TEXT is not split
# on spaces before we check for subvariables.
sub scan_variable_expansions ($)
{
  my ($text) = @_;
  my @result = ();

  # Strip comments.
  $text =~ s/#.*$//;

  # Record each use of ${stuff} or $(stuff) that do not follow a $.
  while ($text =~ /(?<!\$)\$(?:\{([^\}]*)\}|\(([^\)]*)\))/g)
    {
      my $var = $1 || $2;
      # The occurent may look like $(string1[:subst1=[subst2]]) but
      # we want only `string1'.
      $var =~ s/:[^:=]*=[^=]*$//;
      push @result, $var;
    }

  return @result;
}

# &check_variable_expansions ($TEXT, $WHERE)
# ------------------------------------------
# Check variable expansions in $TEXT and warn about any name that
# does not conform to POSIX.  $WHERE is the location of $TEXT for
# the error message.
sub check_variable_expansions ($$)
{
  my ($text, $where) = @_;
  # Catch expansion of variables whose name does not conform to POSIX.
  foreach my $var (scan_variable_expansions ($text))
    {
      if ($var !~ /$MACRO_PATTERN/)
	{
	  # If the variable name contains a space, it's likely
	  # to be a GNU make extension (such as $(addsuffix ...)).
	  # Mention this in the diagnostic.
	  my $gnuext = "";
	  $gnuext = "\n(probably a GNU make extension)" if $var =~ / /;
	  msg ('portability', $where,
	       "$var: non-POSIX variable name$gnuext");
	}
    }
}

# &variable_conditions_recursive_sub ($VAR, $PARENT)
# -------------------------------------------------------
# A subroutine of variable_conditions_recursive.  This returns all the
# conditions of $VAR, including those of any sub-variables.
sub variable_conditions_recursive_sub
{
    my ($var, $parent) = @_;
    my @new_conds = ();

    if (defined $vars_scanned{$var})
    {
        err_var $parent, "variable `$var' recursively defined";
	return ();
    }
    $vars_scanned{$var} = 1;

    my @this_conds = ();
    # Examine every condition under which $VAR is defined.
    foreach my $vcond (keys %{$var_value{$var}})
    {
      push (@this_conds, $vcond);

      # If $VAR references some other variable, then compute the
      # conditions for that subvariable.
      my @subvar_conds = ();
      foreach my $varname (scan_variable_expansions $var_value{$var}{$vcond})
	{
	  if ($varname =~ /$SUBST_REF_PATTERN/o)
	    {
	      $varname = $1;
	    }

	  # Here we compute all the conditions under which the
	  # subvariable is defined.  Then we go through and add
	  # $VCOND to each.
	  my @svc = variable_conditions_recursive_sub ($varname, $var);
	  foreach my $item (@svc)
	    {
	      my $val = conditional_string ($vcond, split (' ', $item));
	      $val ||= 'TRUE';
	      push (@subvar_conds, $val);
	    }
	}

      # If there are no conditional subvariables, then we want to
      # return this condition.  Otherwise, we want to return the
      # permutations of the subvariables, taking into account the
      # conditions of $VAR.
      if (! @subvar_conds)
	{
	  push (@new_conds, $vcond);
	}
      else
	{
	  push (@new_conds, variable_conditions_reduce (@subvar_conds));
	}
    }

    # Unset our entry in vars_scanned.  We only care about recursive
    # definitions.
    delete $vars_scanned{$var};

    # If we are being called on behalf of another variable, we need to
    # return all possible permutations of the conditions.  We have
    # already handled everything in @this_conds along with their
    # subvariables.  We now need to add any permutations that are not
    # in @this_conds.
    foreach my $this_cond (@this_conds)
    {
	my @perms =
	    variable_conditions_permutations (split (' ', $this_cond));
	foreach my $perm (@perms)
	{
	    my $ok = 1;
	    foreach my $scan (@this_conds)
	    {
		if (&conditional_true_when ($perm, $scan)
		    || &conditional_true_when ($scan, $perm))
		{
		    $ok = 0;
		    last;
		}
	    }
	    next if ! $ok;

	    # This permutation was not already handled, and is valid
	    # for the parents.
	    push (@new_conds, $perm);
	}
    }

    return @new_conds;
}


# Filter a list of conditionals so that only the exclusive ones are
# retained.  For example, if both `COND1_TRUE COND2_TRUE' and
# `COND1_TRUE' are in the list, discard the latter.
# If the list is empty, return TRUE
sub variable_conditions_reduce
{
    my (@conds) = @_;
    my @ret = ();
    my $cond;
    while(@conds > 0)
    {
	$cond = shift(@conds);

        # FALSE is absorbent.
	return 'FALSE'
	  if $cond eq 'FALSE';

	if (!conditional_is_redundant ($cond, @ret, @conds))
	  {
	    push (@ret, $cond);
	  }
    }

    return "TRUE" if @ret == 0;
    return @ret;
}

# @CONDS
# invert_conditions (@CONDS)
# --------------------------
# Invert a list of conditionals.  Returns a set of conditionals which
# are never true for any of the input conditionals, and when taken
# together with the input conditionals cover all possible cases.
#
# For example:
#   invert_conditions("A_TRUE B_TRUE", "A_FALSE B_FALSE")
#     => ("A_FALSE B_TRUE", "A_TRUE B_FALSE")
#
#   invert_conditions("A_TRUE B_TRUE", "A_TRUE B_FALSE", "A_FALSE")
#     => ()
sub invert_conditions
{
    my (@conds) = @_;

    my @notconds = ();

    # Generate all permutation for all inputs.
    my @perm =
	map { variable_conditions_permutations (split(' ', $_)); } @conds;
    # Remove redundant conditions.
    @perm = variable_conditions_reduce @perm;

    # Now remove all conditions which imply one of the input conditions.
    foreach my $perm (@perm)
    {
	push @notconds, $perm
	    if ! conditional_implies_any ($perm, @conds);
    }
    return @notconds;
}

# Return a list of permutations of a conditional string.
# (But never output FALSE conditions, they are useless.)
#
# Examples:
#   variable_conditions_permutations ("FOO_FALSE", "BAR_TRUE")
#     => ("FOO_FALSE BAR_FALSE",
#         "FOO_FALSE BAR_TRUE",
#         "FOO_TRUE BAR_FALSE",
#         "FOO_TRUE BAR_TRUE")
#   variable_conditions_permutations ("FOO_FALSE", "TRUE")
#     => ("FOO_FALSE TRUE",
#         "FOO_TRUE TRUE")
#   variable_conditions_permutations ("TRUE")
#     => ("TRUE")
#   variable_conditions_permutations ("FALSE")
#     => ("TRUE")
sub variable_conditions_permutations
{
    my (@comps) = @_;
    return ()
	if ! @comps;
    my $comp = shift (@comps);
    return variable_conditions_permutations (@comps)
	if $comp eq '';
    my $neg = condition_negate ($comp);

    my @ret;
    foreach my $sub (variable_conditions_permutations (@comps))
    {
	push (@ret, "$comp $sub") if $comp ne 'FALSE';
	push (@ret, "$neg $sub") if $neg ne 'FALSE';
    }
    if (! @ret)
    {
	push (@ret, $comp) if $comp ne 'FALSE';
	push (@ret, $neg) if $neg ne 'FALSE';
    }
    return @ret;
}


# $BOOL
# &check_variable_defined_unconditionally($VAR, $PARENT)
# ------------------------------------------------------
# Warn if a variable is conditionally defined.  This is called if we
# are using the value of a variable.
sub check_variable_defined_unconditionally ($$)
{
  my ($var, $parent) = @_;
  foreach my $cond (keys %{$var_value{$var}})
    {
      next
	if $cond =~ /^TRUE|FALSE$/;

      if ($parent)
	{
	  msg_var ('unsupported', $parent,
		   "automake does not support conditional definition of "
		   . "$var in $parent");
	}
      else
	{
	  msg_var ('unsupported', $var,
		   "automake does not support $var being defined "
		   . "conditionally");
	}
    }
}


# Get the TRUE value of a variable, warn if the variable is
# conditionally defined.
sub variable_value
{
    my ($var) = @_;
    &check_variable_defined_unconditionally ($var);
    return $var_value{$var}{'TRUE'};
}


# @VALUES
# &value_to_list ($VAR, $VAL, $COND)
# ----------------------------------
# Convert a variable value to a list, split as whitespace.  This will
# recursively follow $(...) and ${...} inclusions.  It preserves @...@
# substitutions.
#
# If COND is 'all', then all values under all conditions should be
# returned; if COND is a particular condition (all conditions are
# surrounded by @...@) then only the value for that condition should
# be returned; otherwise, warn if VAR is conditionally defined.
# SCANNED is a global hash listing whose keys are all the variables
# already scanned; it is an error to rescan a variable.
sub value_to_list ($$$)
{
    my ($var, $val, $cond) = @_;
    my @result;

    # Strip backslashes
    $val =~ s/\\(\n|$)/ /g;

    foreach (split (' ', $val))
    {
	# If a comment seen, just leave.
	last if /^#/;

	# Handle variable substitutions.
	if (/^\$\{([^}]*)\}$/ || /^\$\(([^)]*)\)$/)
	{
	    my $varname = $1;

	    # If the user uses a losing variable name, just ignore it.
	    # This isn't ideal, but people have requested it.
	    next if ($varname =~ /\@.*\@/);

	    my ($from, $to);
	    my @temp_list;
	    if ($varname =~ /$SUBST_REF_PATTERN/o)
	    {
		$varname = $1;
		$to = $3;
		$from = quotemeta $2;
	    }

	    # Find the value.
	    @temp_list =
	      variable_value_as_list_recursive_worker ($1, $cond, $var);

	    # Now rewrite the value if appropriate.
	    if (defined $from)
	    {
		grep (s/$from$/$to/, @temp_list);
	    }

	    push (@result, @temp_list);
	}
	else
	{
	    push (@result, $_);
	}
    }

    return @result;
}


# @VALUES
# variable_value_as_list ($VAR, $COND, $PARENT)
# ---------------------------------------------
# Get the value of a variable given a specified condition. without
# recursing through any subvariables.
# Arguments are:
#   $VAR    is the variable
#   $COND   is the condition.  If this is not given, the value for the
#           "TRUE" condition will be returned.
#   $PARENT is the variable in which the variable is used: this is used
#           only for error messages.
# Returns the list of conditions.
# For example, if A is defined as "foo $(B) bar", and B is defined as
# "baz", this will return ("foo", "$(B)", "bar")
sub variable_value_as_list
{
    my ($var, $cond, $parent) = @_;
    my @result;

    # Check defined
    return
      unless variable_assert $var, $parent;

    # Get value for given condition
    $cond ||= 'TRUE';
    my $onceflag;
    foreach my $vcond (keys %{$var_value{$var}})
    {
	my $val = $var_value{$var}{$vcond};

	if (&conditional_true_when ($vcond, $cond))
	{
	    # Unless variable is not defined conditionally, there should only
	    # be one value of $vcond true when $cond.
	    &check_variable_defined_unconditionally ($var, $parent)
		    if $onceflag;
	    $onceflag = 1;

	    # Strip backslashes
	    $val =~ s/\\(\n|$)/ /g;

	    foreach (split (' ', $val))
	    {
		# If a comment seen, just leave.
		last if /^#/;

		push (@result, $_);
	    }
	}
    }

    return @result;
}


# @VALUE
# &variable_value_as_list_recursive_worker ($VAR, $COND, $PARENT)
# ---------------------------------------------------------------
# Return contents of VAR as a list, split on whitespace.  This will
# recursively follow $(...) and ${...} inclusions.  It preserves @...@
# substitutions.  If COND is 'all', then all values under all
# conditions should be returned; if COND is a particular condition
# (all conditions are surrounded by @...@) then only the value for
# that condition should be returned; otherwise, warn if VAR is
# conditionally defined.  If PARENT is specified, it is the name of
# the including variable; this is only used for error reports.
sub variable_value_as_list_recursive_worker ($$$)
{
    my ($var, $cond, $parent) = @_;
    my @result = ();

    return
      unless variable_assert $var, $parent;

    if (defined $vars_scanned{$var})
    {
	# `vars_scanned' is a global we use to keep track of which
	# variables we've already examined.
	err_var $parent, "variable `$var' recursively defined";
    }
    elsif ($cond eq 'all')
    {
	$vars_scanned{$var} = 1;
	foreach my $vcond (keys %{$var_value{$var}})
	{
	    my $val = $var_value{$var}{$vcond};
	    push (@result, &value_to_list ($var, $val, $cond));
	}
    }
    else
    {
        $cond ||= 'TRUE';
	$vars_scanned{$var} = 1;
	my $onceflag;
	foreach my $vcond (keys %{$var_value{$var}})
	{
	    my $val = $var_value{$var}{$vcond};
	    if (&conditional_true_when ($vcond, $cond))
	    {
		# Warn if we have an ambiguity.  It's hard to know how
		# to handle this case correctly.
		&check_variable_defined_unconditionally ($var, $parent)
		    if $onceflag;
		$onceflag = 1;
		push (@result, &value_to_list ($var, $val, $cond));
	    }
	}
    }

    # Unset our entry in vars_scanned.  We only care about recursive
    # definitions.
    delete $vars_scanned{$var};

    return @result;
}


# &variable_output ($VAR, [@CONDS])
# ---------------------------------
# Output all the values of $VAR if @COND is not specified, else only
# that corresponding to @COND.
sub variable_output ($@)
{
  my ($var, @conds) = @_;

  @conds = keys %{$var_value{$var}}
    unless @conds;

  foreach my $cond (sort by_condition @conds)
    {
      prog_error ("unknown condition `$cond' for `$var'")
	unless exists $var_value{$var}{$cond};

      if (exists $var_comment{$var} && exists $var_comment{$var}{$cond})
	{
	  $output_vars .= $var_comment{$var}{$cond};
	}

      my $val = $var_value{$var}{$cond};
      my $equals = $var_type{$var}{$cond} eq ':' ? ':=' : '=';
      my $output_var = "$var $equals $val";
      $output_var =~ s/^/make_condition ($cond)/meg;
      $output_vars .= $output_var . "\n";
    }
}


# &variable_pretty_output ($VAR, [@CONDS])
# ----------------------------------------
# Likewise, but pretty, i.e., we *split* the values at spaces.   Use only
# with variables holding filenames.
sub variable_pretty_output ($@)
{
  my ($var, @conds) = @_;

  @conds = keys %{$var_value{$var}}
    unless @conds;

  foreach my $cond (sort by_condition @conds)
    {
      prog_error ("unknown condition `$cond' for `$var'")
	unless exists $var_value{$var}{$cond};

      if (exists $var_comment{$var} && exists $var_comment{$var}{$cond})
	{
	  $output_vars .= $var_comment{$var}{$cond};
	}

      my $val = $var_value{$var}{$cond};
      my $equals = $var_type{$var}{$cond} eq ':' ? ':=' : '=';
      my $make_condition = make_condition ($cond);

      # Suppress escaped new lines.  &pretty_print_internal will
      # add them back, maybe at other places.
      $val =~ s/\\$//mg;

      $output_vars .= pretty_print_internal ("$make_condition$var $equals",
					     "$make_condition\t",
					     split (' ' , $val));
    }
}


# &variable_value_as_list_recursive ($VAR, $COND, $PARENT)
# --------------------------------------------------------
# This is just a wrapper for variable_value_as_list_recursive_worker that
# initializes the global hash `vars_scanned'.  This hash is used to
# avoid infinite recursion.
sub variable_value_as_list_recursive ($$@)
{
    my ($var, $cond, $parent) = @_;
    %vars_scanned = ();
    return &variable_value_as_list_recursive_worker ($var, $cond, $parent);
}


# &define_pretty_variable ($VAR, $COND, @VALUE)
# ---------------------------------------------
# Like define_variable, but the value is a list, and the variable may
# be defined conditionally.  The second argument is the conditional
# under which the value should be defined; this should be the empty
# string to define the variable unconditionally.  The third argument
# is a list holding the values to use for the variable.  The value is
# pretty printed in the output file.
sub define_pretty_variable ($$@)
{
    my ($var, $cond, @value) = @_;

    # Beware that an empty $cond has a different semantics for
    # macro_define and variable_pretty_output.
    $cond ||= 'TRUE';

    if (! variable_defined ($var, $cond))
    {
        macro_define ($var, VAR_AUTOMAKE, '', $cond, "@value", undef);
	variable_pretty_output ($var, $cond || 'TRUE');
	$content_seen{$var} = 1;
    }
}


# define_variable ($VAR, $VALUE)
# ------------------------------
# Define a new user variable VAR to VALUE, but only if not already defined.
sub define_variable ($$)
{
    my ($var, $value) = @_;
    define_pretty_variable ($var, 'TRUE', $value);
}


# Like define_variable, but define a variable to be the configure
# substitution by the same name.
sub define_configure_variable ($)
{
  my ($var) = @_;
  if (! variable_defined ($var, 'TRUE')
      # Explicitly avoid ANSI2KNR -- we AC_SUBST that in
      # protos.m4, but later define it elsewhere.  This is
      # pretty hacky.  We also explicitly avoid AMDEPBACKSLASH:
      # it might be subst'd by `\', which certainly would not be
      # appreciated by Make.
      && ! grep { $_ eq $var } (qw(ANSI2KNR AMDEPBACKSLASH)))
    {
      macro_define ($var, VAR_CONFIGURE, '', 'TRUE',
		    subst $var, $configure_vars{$var});
      variable_pretty_output ($var, 'TRUE');
    }
}


# define_compiler_variable ($LANG)
# --------------------------------
# Define a compiler variable.  We also handle defining the `LT'
# version of the command when using libtool.
sub define_compiler_variable ($)
{
    my ($lang) = @_;

    my ($var, $value) = ($lang->compiler, $lang->compile);
    &define_variable ($var, $value);
    &define_variable ("LT$var", "\$(LIBTOOL) --mode=compile $value")
      if variable_defined ('LIBTOOL');
}


# define_linker_variable ($LANG)
# ------------------------------
# Define linker variables.
sub define_linker_variable ($)
{
    my ($lang) = @_;

    my ($var, $value) = ($lang->lder, $lang->ld);
    # CCLD = $(CC).
    &define_variable ($lang->lder, $lang->ld);
    # CCLINK = $(CCLD) blah blah...
    &define_variable ($lang->linker,
		      ((variable_defined ('LIBTOOL')
			? '$(LIBTOOL) --mode=link ' : '')
		       . $lang->link));
}

################################################################

## ---------------- ##
## Handling rules.  ##
## ---------------- ##

sub register_suffix_rule ($$$)
{
  my ($where, $src, $dest) = @_;

  verb "Sources ending in $src become $dest";
  push @suffixes, $src, $dest;

  # When tranforming sources to objects, Automake uses the
  # %suffix_rules to move from each source extension to
  # `.$(OBJEXT)', not to `.o' or `.obj'.  However some people
  # define suffix rules for `.o' or `.obj', so internally we will
  # consider these extensions equivalent to `.$(OBJEXT)'.  We
  # CANNOT rewrite the target (i.e., automagically replace `.o'
  # and `.obj' by `.$(OBJEXT)' in the output), or warn the user
  # that (s)he'd better use `.$(OBJEXT)', because Automake itself
  # output suffix rules for `.o' or `.obj'...
  $dest = '.$(OBJEXT)' if ($dest eq '.o' || $dest eq '.obj');

  # Reading the comments near the declaration of $suffix_rules might
  # help to understand the update of $suffix_rules that follows...

  # Register $dest as a possible destination from $src.
  # We might have the create the \hash.
  if (exists $suffix_rules->{$src})
    {
      $suffix_rules->{$src}{$dest} = [ $dest, 1 ];
    }
  else
    {
      $suffix_rules->{$src} = { $dest => [ $dest, 1 ] };
    }

  # If we know how to transform $dest in something else, then
  # we know how to transform $src in that "something else".
  if (exists $suffix_rules->{$dest})
    {
      for my $dest2 (keys %{$suffix_rules->{$dest}})
	{
	  my $dist = $suffix_rules->{$dest}{$dest2}[1] + 1;
	  # Overwrite an existing $src->$dest2 path only if
	  # the path via $dest which is shorter.
	  if (! exists $suffix_rules->{$src}{$dest2}
	      || $suffix_rules->{$src}{$dest2}[1] > $dist)
	    {
	      $suffix_rules->{$src}{$dest2} = [ $dest, $dist ];
	    }
	}
    }

  # Similarly, any extension that can be derived into $src
  # can be derived into the same extenstions as $src can.
  my @dest2 = keys %{$suffix_rules->{$src}};
  for my $src2 (keys %$suffix_rules)
    {
      if (exists $suffix_rules->{$src2}{$src})
	{
	  for my $dest2 (@dest2)
	    {
	      my $dist = $suffix_rules->{$src}{$dest2} + 1;
	      # Overwrite an existing $src2->$dest2 path only if
	      # the path via $src is shorter.
	      if (! exists $suffix_rules->{$src2}{$dest2}
		  || $suffix_rules->{$src2}{$dest2}[1] > $dist)
		{
		  $suffix_rules->{$src2}{$dest2} = [ $src, $dist ];
		}
	    }
	}
    }
}

# @CONDS
# rule_define ($TARGET, $SOURCE, $OWNER, $COND, $WHERE)
# -----------------------------------------------------
# Define a new rule.  $TARGET is the rule name.  $SOURCE
# is the filename the rule comes from.  $OWNER is the
# owener of the rule (TARGET_AUTOMAKE or TARGET_USER).
# $COND is the condition string under which the rule is defined.
# $WHERE is where the rule is defined (file name and/or line number).
# Returns a (possibly empty) list of conditions where the rule
# should be defined.
sub rule_define ($$$$$)
{
  my ($target, $source, $owner, $cond, $where) = @_;

  # Don't even think about defining a rule in condition FALSE.
  return () if $cond eq 'FALSE';

  # For now `foo:' will override `foo$(EXEEXT):'.  This is temporary,
  # though, so we emit a warning.
  (my $noexe = $target) =~ s,\$\(EXEEXT\)$,,;
  if ($noexe ne $target
      && exists $targets{$noexe}
      && exists $targets{$noexe}{$cond}
      && $target_name{$noexe}{$cond} ne $target)
    {
      # The no-exeext option enables this feature.
      if (! defined $options{'no-exeext'})
	{
	  msg ('obsolete', $noexe,
	       "deprecated feature: `$noexe' overrides `$noexe\$(EXEEXT)'\n"
	       . "change your target to read `$noexe\$(EXEEXT)'");
	}
      # Don't define.
      return ();
    }

  # For now on, strip off $(EXEEXT) from $target, so we can diagnose
  # a clash if `ctags$(EXEEXT):' is redefined after `ctags:'.
  my $realtarget = $target;
  $target = $noexe;

  # A GNU make-style pattern rule has a single "%" in the target name.
  msg ('portability', $where,
       "`%'-style pattern rules are a GNU make extension")
    if $target =~ /^[^%]*%[^%]*$/;

  # Diagnose target redefinitions.
  if (exists $target_source{$target}{$cond})
    {
      # Sanity checks.
      prog_error ("\$target_source{$target}{$cond} exists, but \$target_owner"
		  . " doesn't.")
	unless exists $target_owner{$target}{$cond};
      prog_error ("\$target_source{$target}{$cond} exists, but \$targets"
		  . " doesn't.")
	unless exists $targets{$target}{$cond};
      prog_error ("\$target_source{$target}{$cond} exists, but \$target_name"
		  . " doesn't.")
	unless exists $target_name{$target}{$cond};

      my $oldowner  = $target_owner{$target}{$cond};

      # Don't mention true conditions in diagnostics.
      my $condmsg = $cond ne 'TRUE' ? " in condition `$cond'" : '';

      if ($owner == TARGET_USER)
	{
	  if ($oldowner eq TARGET_USER)
	    {
	      # Ignore `%'-style pattern rules.  We'd need the
	      # dependencies to detect duplicates, and they are
	      # already diagnosed as unportable by -Wportability.
	      if ($target !~ /^[^%]*%[^%]*$/)
		{
		  ## FIXME: Presently we can't diagnose duplcate user rules
		  ## because we doesn't distinguish rules with commands
		  ## from rules that only add dependencies.  E.g.,
		  ##   .PHONY: foo
		  ##   .PHONY: bar
		  ## is legitimate. (This is phony.test.)

		  # msg ('syntax', $where,
		  #      "redefinition of `$target'$condmsg...");
		  # msg_cond_target ('syntax', $cond, $target,
		  # 		   "... `$target' previously defined here.");
		}
	      # Return so we don't redefine the rule in our tables,
	      # don't check for ambiguous conditional, etc.  The rule
	      # will be output anyway beauce &read_am_file ignore the
	      # return code.
	      return ();
	    }
	  else
	    {
	      # Since we parse the user Makefile.am before reading
	      # the Automake fragments, this condition should never happen.
	      prog_error ("user target `$target' seen after Automake's "
			  . "definition\nfrom `$targets{$target}$condmsg'");
	    }
	}
      else # $owner == TARGET_AUTOMAKE
	{
	  if ($oldowner == TARGET_USER)
	    {
	      # Don't overwrite the user definition of TARGET.
	      return ();
	    }
	  else # $oldowner == TARGET_AUTOMAKE
	    {
	      # Automake should ignore redefinitions of its own
	      # rules if they came from the same file.  This makes
	      # it easier to process a Makefile fragment several times.
	      # Hower it's an error if the target is defined in many
	      # files.  E.g., the user might be using bin_PROGRAMS = ctags
	      # which clashes with our `ctags' rule.
	      # (It would be more accurate if we had a way to compare
	      # the *content* of both rules.  Then $targets_source would
	      # be useless.)
	      my $oldsource = $target_source{$target}{$cond};
	      return () if $source eq $oldsource;

	      msg ('syntax', $where, "redefinition of `$target'$condmsg...");
	      msg_cond_target ('syntax', $cond, $target,
			       "... `$target' previously defined here.");
	      return ();
	    }
	}
      # Never reached.
      prog_error ("Unreachable place reached.");
    }

  # Conditions for which the rule should be defined.
  my @conds = $cond;

  # Check ambiguous conditional definitions.
  my ($message, $ambig_cond) =
    conditional_ambiguous_p ($target, $cond, keys %{$targets{$target}});
  if ($message)			# We have an ambiguty.
    {
      if ($owner == TARGET_USER)
	{
	  # For user rules, just diagnose the ambiguity.
	  msg 'syntax', $where, "$message ...";
	  msg_cond_target ('syntax', $ambig_cond, $target,
			   "... `$target' previously defined here.");
	  return ();
	}
      else
	{
	  # FIXME: for Automake rules, we can't diagnose ambiguities yet.
	  # The point is that Automake doesn't propagate conditionals
	  # everywhere.  For instance &handle_PROGRAMS doesn't care if
	  # bin_PROGRAMS was defined conditionally or not.
	  # On the following input
	  #   if COND1
	  #   foo:
	  #           ...
	  #   else
	  #   bin_PROGRAMS = foo
	  #   endif
	  # &handle_PROGRAMS will attempt to define a `foo:' rule
	  # in condition TRUE (which conflicts with COND1).  Fixing
	  # this in &handle_PROGRAMS and siblings seems hard: you'd
	  # have to explain &file_contents what to do with a
	  # conditional.  So for now we do our best *here*.  If `foo:'
	  # was already defined in condition COND1 and we want to define
	  # it in condition TRUE, then define it only in condition !COND1.
	  # (See cond14.test and cond15.test for some test cases.)
	  my @defined_conds = keys %{$targets{$target}};
	  @conds = ();
	  for my $undefined_cond (invert_conditions(@defined_conds))
	    {
	      push @conds, make_condition ($cond, $undefined_cond);
	    }
	  # No conditions left to define the rule.
	  # Warn, because our workaround is meaningless in this case.
	  if (scalar @conds == 0)
	    {
	      msg 'syntax', $where, "$message ...";
	      msg_cond_target ('syntax', $ambig_cond, $target,
			       "... `$target' previously defined here.");
	      return ();
	    }
	}
    }

  # Finally define this rule.
  for my $c (@conds)
    {
      $targets{$target}{$c} = $where;
      $target_source{$target}{$c} = $source;
      $target_owner{$target}{$c} = $owner;
      $target_name{$target}{$c} = $realtarget;
    }

  # We honor inference rules with multiple targets because many
  # make support this and people use it.  However this is disallowed
  # by POSIX.  We'll print a warning later.
  my $target_count = 0;
  my $inference_rule_count = 0;
  for my $t (split (' ', $target))
    {
      ++$target_count;
      # Check the rule for being a suffix rule. If so, store in a hash.
      # Either it's a rule for two known extensions...
      if ($t =~ /^($KNOWN_EXTENSIONS_PATTERN)($KNOWN_EXTENSIONS_PATTERN)$/
	  # ...or it's a rule with unknown extensions (.i.e, the rule
	  # looks like `.foo.bar:' but `.foo' or `.bar' are not
	  # declared in SUFFIXES and are not known language
	  # extensions).  Automake will complete SUFFIXES from
	  # @suffixes automatically (see handle_footer).
	  || ($t =~ /$SUFFIX_RULE_PATTERN/o && accept_extensions($1)))
	{
	  ++$inference_rule_count;
	  register_suffix_rule ($where, $1, $2);
	}
    }

  # POSIX allow multiple targets befor the colon, but disallow
  # definitions of multiple Inference rules.  It's also
  # disallowed to mix plain targets with inference rules.
  msg ('portability', $where,
       "Inference rules can have only one target before the colon (POSIX).")
    if $inference_rule_count > 0 && $target_count > 1;

  # Return "" instead of TRUE so it can be used with make_paragraphs
  # directly.
  return "" if 1 == @conds && $conds[0] eq 'TRUE';
  return @conds;
}


# See if a target exists.
sub target_defined
{
    my ($target) = @_;
    return exists $targets{$target};
}


################################################################

# &check_trailing_slash ($WHERE, $LINE)
# --------------------------------------
# Return 1 iff $LINE ends with a slash.
# Might modify $LINE.
sub check_trailing_slash ($\$)
{
  my ($where, $line) = @_;

  # Ignore `##' lines.
  return 0 if $$line =~ /$IGNORE_PATTERN/o;

  # Catch and fix a common error.
  msg "syntax", $where, "whitespace following trailing backslash"
    if $$line =~ s/\\\s+\n$/\\\n/;

  return $$line =~ /\\$/;
}


# &append_comments ($VARIABLE, $SPACING, $COMMENT)
# ------------------------------------------------
# Apped $COMMENT to the other comments for $VARIABLE, using
# $SPACING as separator.
sub append_comments ($$$$)
{
    my ($cond, $var, $spacing, $comment) = @_;
    $var_comment{$var}{$cond} .= $spacing
	if (!defined $var_comment{$var}{$cond}
	    || $var_comment{$var}{$cond} !~ /\n$/o);
    $var_comment{$var}{$cond} .= $comment;
}


# &read_am_file ($AMFILE)
# -----------------------
# Read Makefile.am and set up %contents.  Simultaneously copy lines
# from Makefile.am into $output_trailer or $output_vars as
# appropriate.  NOTE we put rules in the trailer section.  We want
# user rules to come after our generated stuff.
sub read_am_file ($)
{
    my ($amfile) = @_;

    my $am_file = new Automake::XFile ("< $amfile");
    verb "reading $amfile";

    my $spacing = '';
    my $comment = '';
    my $blank = 0;
    my $saw_bk = 0;

    use constant IN_VAR_DEF => 0;
    use constant IN_RULE_DEF => 1;
    use constant IN_COMMENT => 2;
    my $prev_state = IN_RULE_DEF;

    while ($_ = $am_file->getline)
    {
	if (/$IGNORE_PATTERN/o)
	{
	    # Merely delete comments beginning with two hashes.
	}
	elsif (/$WHITE_PATTERN/o)
	{
	    error "$amfile:$.", "blank line following trailing backslash"
	      if $saw_bk;
	    # Stick a single white line before the incoming macro or rule.
	    $spacing = "\n";
	    $blank = 1;
	    # Flush all comments seen so far.
	    if ($comment ne '')
	    {
		$output_vars .= $comment;
		$comment = '';
	    }
	}
	elsif (/$COMMENT_PATTERN/o)
	{
	    # Stick comments before the incoming macro or rule.  Make
	    # sure a blank line preceeds first block of comments.
	    $spacing = "\n" unless $blank;
	    $blank = 1;
	    $comment .= $spacing . $_;
	    $spacing = '';
	    $prev_state = IN_COMMENT;
	}
	else
	{
	    last;
	}
	$saw_bk = check_trailing_slash ("$amfile:$.", $_);
    }

    # We save the conditional stack on entry, and then check to make
    # sure it is the same on exit.  This lets us conditonally include
    # other files.
    my @saved_cond_stack = @cond_stack;
    my $cond = conditional_string (@cond_stack);

    my $last_var_name = '';
    my $last_var_type = '';
    my $last_var_value = '';
    # FIXME: shouldn't use $_ in this loop; it is too big.
    while ($_)
    {
        my $here = "$amfile:$.";

	# Make sure the line is \n-terminated.
	chomp;
	$_ .= "\n";

	# Don't look at MAINTAINER_MODE_TRUE here.  That shouldn't be
	# used by users.  @MAINT@ is an anachronism now.
	$_ =~ s/\@MAINT\@//g
	    unless $seen_maint_mode;

	my $new_saw_bk = check_trailing_slash ($here, $_);

	if (/$IGNORE_PATTERN/o)
	{
	    # Merely delete comments beginning with two hashes.
	}
	elsif (/$WHITE_PATTERN/o)
	{
	    # Stick a single white line before the incoming macro or rule.
	    $spacing = "\n";
	    error $here, "blank line following trailing backslash"
	      if $saw_bk;
	}
	elsif (/$COMMENT_PATTERN/o)
	{
	    # Stick comments before the incoming macro or rule.
	    $comment .= $spacing . $_;
	    $spacing = '';
	    error $here, "comment following trailing backslash"
	      if $saw_bk && $comment eq '';
	    $prev_state = IN_COMMENT;
	}
	elsif ($saw_bk)
	{
	    if ($prev_state == IN_RULE_DEF)
	    {
	        $output_trailer .= &make_condition (@cond_stack);
		$output_trailer .= $_;
	    }
	    elsif ($prev_state == IN_COMMENT)
	    {
		# If the line doesn't start with a `#', add it.
		# We do this because a continuated comment like
		#   # A = foo \
		#         bar \
		#         baz
		# is not portable.  BSD make doesn't honor
		# escaped newlines in comments.
		s/^#?/#/;
		$comment .= $spacing . $_;
	    }
	    else # $prev_state == IN_VAR_DEF
	    {
	      $last_var_value .= ' '
		unless $last_var_value =~ /\s$/;
	      $last_var_value .= $_;

	      if (!/\\$/)
		{
		  append_comments ($cond || 'TRUE',
				   $last_var_name, $spacing, $comment);
		  $comment = $spacing = '';
		  macro_define ($last_var_name, VAR_MAKEFILE,
				$last_var_type, $cond,
				$last_var_value, $here)
		    if $cond ne 'FALSE';
		  push (@var_list, $last_var_name);
		}
	    }
	}

	elsif (/$IF_PATTERN/o)
	  {
	    $cond = cond_stack_if ($1, $2, $here);
	  }
	elsif (/$ELSE_PATTERN/o)
	  {
	    $cond = cond_stack_else ($1, $2, $here);
	  }
	elsif (/$ENDIF_PATTERN/o)
	  {
	    $cond = cond_stack_endif ($1, $2, $here);
	  }

	elsif (/$RULE_PATTERN/o)
	{
	    # Found a rule.
	    $prev_state = IN_RULE_DEF;

	    # For now we have to output all definitions of user rules
	    # and can't diagnose duplicates (see the comment in
	    # rule_define). So we go on and ignore the return value.
	    rule_define ($1, $amfile, TARGET_USER, $cond || 'TRUE', $here);

	    check_variable_expansions ($_, $here);

	    $output_trailer .= $comment . $spacing;
            $output_trailer .= &make_condition (@cond_stack);
            $output_trailer .= $_;
	    $comment = $spacing = '';
	}
	elsif (/$ASSIGNMENT_PATTERN/o)
	{
	    # Found a macro definition.
	    $prev_state = IN_VAR_DEF;
	    $last_var_name = $1;
	    $last_var_type = $2;
	    $last_var_value = $3;
	    if ($3 ne '' && substr ($3, -1) eq "\\")
	    {
		# We preserve the `\' because otherwise the long lines
		# that are generated will be truncated by broken
		# `sed's.
		$last_var_value = $3 . "\n";
	    }

	    if (!/\\$/)
	      {
		# Accumulating variables must not be output.
		append_comments ($cond || 'TRUE',
				 $last_var_name, $spacing, $comment);
		$comment = $spacing = '';

		macro_define ($last_var_name, VAR_MAKEFILE,
			      $last_var_type, $cond,
			      $last_var_value, $here)
		  if $cond ne 'FALSE';
		push (@var_list, $last_var_name);
	      }
	}
        elsif (/$INCLUDE_PATTERN/o)
        {
            my $path = $1;

            if ($path =~ s/^\$\(top_srcdir\)\///)
	      {
                push (@include_stack, "\$\(top_srcdir\)/$path");
		# Distribute any included file.

		# Always use the $(top_srcdir) prefix in DIST_COMMON,
		# otherwise OSF make will implicitely copy the included
		# file in the build tree during `make distdir' to satisfy
		# the dependency.
		# (subdircond2.test and subdircond3.test will fail.)
		push_dist_common ("\$\(top_srcdir\)/$path");
	      }
            else
	      {
                $path =~ s/\$\(srcdir\)\///;
                push (@include_stack, "\$\(srcdir\)/$path");
		# Always use the $(srcdir) prefix in DIST_COMMON,
		# otherwise OSF make will implicitely copy the included
		# file in the build tree during `make distdir' to satisfy
		# the dependency.
		# (subdircond2.test and subdircond3.test will fail.)
		push_dist_common ("\$\(srcdir\)/$path");
		$path = $relative_dir . "/" . $path;
	      }
            &read_am_file ($path);
        }
	else
        {
	    # This isn't an error; it is probably a continued rule.
	    # In fact, this is what we assume.
	    $prev_state = IN_RULE_DEF;
	    check_variable_expansions ($_, $here);
	    $output_trailer .= $comment . $spacing;
	    $output_trailer .= &make_condition  (@cond_stack);
	    $output_trailer .= $_;
	    $comment = $spacing = '';
	    error $here, "`#' comment at start of rule is unportable"
	      if $_ =~ /^\t\s*\#/;
	}

	$saw_bk = $new_saw_bk;
        $_ = $am_file->getline;
    }

    $output_trailer .= $comment;

    err_am ("trailing backslash on last line")
      if $saw_bk;

    err_am (@cond_stack ? "unterminated conditionals: @cond_stack"
	    : "too many conditionals closed in include file")
      if "@saved_cond_stack" ne "@cond_stack";
}


# define_standard_variables ()
# ----------------------------
# A helper for read_main_am_file which initializes configure variables
# and variables from header-vars.am.
sub define_standard_variables
{
    my $saved_output_vars = $output_vars;
    my ($comments, undef, $rules) =
      file_contents_internal (1, "$libdir/am/header-vars.am");

    # This will output the definitions in $output_vars, which we don't
    # want...
    foreach my $var (sort keys %configure_vars)
    {
        &define_configure_variable ($var);
        push (@var_list, $var);
    }

    # ... hence, we restore $output_vars.
    $output_vars = $saved_output_vars . $comments . $rules;
}

# Read main am file.
sub read_main_am_file
{
    my ($amfile) = @_;

    # This supports the strange variable tricks we are about to play.
    prog_error (macros_dump () . "variable defined before read_main_am_file")
      if (scalar keys %var_value > 0);

    # Generate copyright header for generated Makefile.in.
    # We do discard the output of predefined variables, handled below.
    $output_vars = ("# $in_file_name generated by automake "
		   . $VERSION . " from $am_file_name.\n");
    $output_vars .= '# ' . subst ('configure_input') . "\n";
    $output_vars .= $gen_copyright;

    # We want to predefine as many variables as possible.  This lets
    # the user set them with `+=' in Makefile.am.  However, we don't
    # want these initial definitions to end up in the output quite
    # yet.  So we just load them, but output them later.
    &define_standard_variables;

    # Read user file, which might override some of our values.
    &read_am_file ($amfile);

    # Output all the Automake variables.  If the user changed one,
    # then it is now marked as VAR_CONFIGURE or VAR_MAKEFILE.
    foreach my $var (uniq @var_list)
    {
      # Some variables, like AMDEPBACKSLASH are in @var_list
      # but don't have a owner.  This is good, because we don't want
      # to output them.
      foreach my $cond (keys %{$var_owner{$var}})
	{
	  variable_output ($var, $cond)
	    if $var_owner{$var}{$cond} == VAR_AUTOMAKE;
	}
    }

    # Now dump the user variables that were defined.  We do it in the same
    # order in which they were defined (skipping duplicates).
    foreach my $var (uniq @var_list)
    {
      foreach my $cond (keys %{$var_owner{$var}})
	{
	  variable_output ($var, $cond)
	    if $var_owner{$var}{$cond} != VAR_AUTOMAKE;
	}
    }
}

################################################################

# $FLATTENED
# &flatten ($STRING)
# ------------------
# Flatten the $STRING and return the result.
sub flatten
{
  $_ = shift;

  s/\\\n//somg;
  s/\s+/ /g;
  s/^ //;
  s/ $//;

  return $_;
}


# @PARAGRAPHS
# &make_paragraphs ($MAKEFILE, [%TRANSFORM])
# ------------------------------------------
# Load a $MAKEFILE, apply the %TRANSFORM, and return it as a list of
# paragraphs.
sub make_paragraphs ($%)
{
    my ($file, %transform) = @_;

    # Complete %transform with global options and make it a Perl
    # $command.
    my $command =
      "s/$IGNORE_PATTERN//gm;"
	. transform (%transform,

		     'CYGNUS'          => $cygnus_mode,
		     'MAINTAINER-MODE'
		     => $seen_maint_mode ? subst ('MAINTAINER_MODE_TRUE') : '',

		     'SHAR'        => $options{'dist-shar'} || 0,
		     'BZIP2'       => $options{'dist-bzip2'} || 0,
		     'ZIP'         => $options{'dist-zip'} || 0,
		     'COMPRESS'    => $options{'dist-tarZ'} || 0,

		     'INSTALL-INFO' => !$options{'no-installinfo'},
		     'INSTALL-MAN'  => !$options{'no-installman'},
		     'CK-NEWS'      => $options{'check-news'} || 0,

		     'SUBDIRS'      => variable_defined ('SUBDIRS'),
		     'TOPDIR'       => backname ($relative_dir),
		     'TOPDIR_P'     => $relative_dir eq '.',
		     'CONFIGURE-AC' => $configure_ac,

		     'BUILD'    => $seen_canonical == AC_CANONICAL_SYSTEM,
		     'HOST'     => $seen_canonical,
		     'TARGET'   => $seen_canonical == AC_CANONICAL_SYSTEM,

		     'LIBTOOL'      => variable_defined ('LIBTOOL'))
	  # We don't need more than two consecutive new-lines.
	  . 's/\n{3,}/\n\n/g';

    # Swallow the file and apply the COMMAND.
    my $fc_file = new Automake::XFile "< $file";
    # Looks stupid?
    verb "reading $file";
    my $saved_dollar_slash = $/;
    undef $/;
    $_ = $fc_file->getline;
    $/ = $saved_dollar_slash;
    eval $command;
    $fc_file->close;
    my $content = $_;

    # Split at unescaped new lines.
    my @lines = split (/(?<!\\)\n/, $content);
    my @res;

    while (defined ($_ = shift @lines))
      {
	my $paragraph = "$_";
	# If we are a rule, eat as long as we start with a tab.
	if (/$RULE_PATTERN/smo)
	  {
	    while (defined ($_ = shift @lines) && $_ =~ /^\t/)
	      {
		$paragraph .= "\n$_";
	      }
	    unshift (@lines, $_);
	  }

	# If we are a comments, eat as much comments as you can.
	elsif (/$COMMENT_PATTERN/smo)
	  {
	    while (defined ($_ = shift @lines)
		   && $_ =~ /$COMMENT_PATTERN/smo)
	      {
		$paragraph .= "\n$_";
	      }
	    unshift (@lines, $_);
	  }

	push @res, $paragraph;
	$paragraph = '';
      }

    return @res;
}



# ($COMMENT, $VARIABLES, $RULES)
# &file_contents_internal ($IS_AM, $FILE, [%TRANSFORM])
# -----------------------------------------------------
# Return contents of a file from $libdir/am, automatically skipping
# macros or rules which are already known. $IS_AM iff the caller is
# reading an Automake file (as opposed to the user's Makefile.am).
sub file_contents_internal ($$%)
{
    my ($is_am, $file, %transform) = @_;

    my $result_vars = '';
    my $result_rules = '';
    my $comment = '';
    my $spacing = '';

    # The following flags are used to track rules spanning across
    # multiple paragraphs.
    my $is_rule = 0;		# 1 if we are processing a rule.
    my $discard_rule = 0;	# 1 if the current rule should not be output.

    # We save the conditional stack on entry, and then check to make
    # sure it is the same on exit.  This lets us conditonally include
    # other files.
    my @saved_cond_stack = @cond_stack;
    my $cond = conditional_string (@cond_stack);

    foreach (make_paragraphs ($file, %transform))
    {
        # Sanity checks.
	error $file, "blank line following trailing backslash:\n$_"
	  if /\\$/;
	error $file, "comment following trailing backslash:\n$_"
	  if /\\#/;

 	if (/^$/)
 	{
	    $is_rule = 0;
 	    # Stick empty line before the incoming macro or rule.
 	    $spacing = "\n";
 	}
 	elsif (/$COMMENT_PATTERN/mso)
 	{
	    $is_rule = 0;
 	    # Stick comments before the incoming macro or rule.
 	    $comment = "$_\n";
	}

	# Handle inclusion of other files.
        elsif (/$INCLUDE_PATTERN/o)
        {
	    if ($cond ne 'FALSE')
	      {
		my $file = ($is_am ? "$libdir/am/" : '') . $1;
		# N-ary `.=' fails.
		my ($com, $vars, $rules)
		  = file_contents_internal ($is_am, $file, %transform);
		$comment .= $com;
		$result_vars .= $vars;
		$result_rules .= $rules;
	      }
        }

        # Handling the conditionals.
        elsif (/$IF_PATTERN/o)
	  {
	    $cond = cond_stack_if ($1, $2, $file);
	  }
	elsif (/$ELSE_PATTERN/o)
	  {
	    $cond = cond_stack_else ($1, $2, $file);
	  }
	elsif (/$ENDIF_PATTERN/o)
	  {
	    $cond = cond_stack_endif ($1, $2, $file);
	  }

        # Handling rules.
 	elsif (/$RULE_PATTERN/mso)
 	{
	  $is_rule = 1;
	  $discard_rule = 0;
	  # Separate relationship from optional actions: the first
	  # `new-line tab" not preceded by backslash (continuation
	  # line).
	  my $paragraph = $_;
	  /^(.*?)(?:(?<!\\)\n(\t.*))?$/s;
	  my ($relationship, $actions) = ($1, $2 || '');

	  # Separate targets from dependencies: the first colon.
	  $relationship =~ /^([^:]+\S+) *: *(.*)$/som;
	  my ($targets, $dependencies) = ($1, $2);
	  # Remove the escaped new lines.
	  # I don't know why, but I have to use a tmp $flat_deps.
	  my $flat_deps = &flatten ($dependencies);
	  my @deps = split (' ', $flat_deps);

	  foreach (split (' ' , $targets))
	    {
	      # FIXME: 1. We are not robust to people defining several targets
	      # at once, only some of them being in %dependencies.  The
	      # actions from the targets in %dependencies are usually generated
	      # from the content of %actions, but if some targets in $targets
	      # are not in %dependencies the ELSE branch will output
	      # a rule for all $targets (i.e. the targets which are both
	      # in %dependencies and $targets will have two rules).

	      # FIXME: 2. The logic here is not able to output a
	      # multi-paragraph rule several time (e.g. for each conditional
	      # it is defined for) because it only knows the first paragraph.

	      # FIXME: 3. We are not robust to people defining a subset
	      # of a previously defined "multiple-target" rule.  E.g.
	      # `foo:' after `foo bar:'.

	      # Output only if not in FALSE.
	      if (defined $dependencies{$_} && $cond ne 'FALSE')
		{
		  &depend ($_, @deps);
		  $actions{$_} .= $actions;
		}
	      else
		{
		  # Free-lance dependency.  Output the rule for all the
		  # targets instead of one by one.
		  my @undefined_conds =
		    rule_define ($targets, $file,
				 $is_am ? TARGET_AUTOMAKE : TARGET_USER,
				 $cond || 'TRUE', $file);
		  for my $undefined_cond (@undefined_conds)
		    {
		      my $condparagraph = $paragraph;
		      $condparagraph =~ s/^/$undefined_cond/gm;
		      $result_rules .= "$spacing$comment$condparagraph\n";
		    }
		  if (scalar @undefined_conds == 0)
		    {
		      # Remember to discard next paragraphs
		      # if they belong to this rule.
		      # (but see also FIXME: #2 above.)
		      $discard_rule = 1;
		    }
		  $comment = $spacing = '';
		  last;
		}
	    }
 	}

 	elsif (/$ASSIGNMENT_PATTERN/mso)
 	{
	    my ($var, $type, $val) = ($1, $2, $3);
 	    error $file, "variable `$var' with trailing backslash"
	      if /\\$/;

	    $is_rule = 0;

	    # Accumulating variables must not be output.
	    append_comments ($cond || 'TRUE', $var, $spacing, $comment);
	    macro_define ($var, $is_am ? VAR_AUTOMAKE : VAR_MAKEFILE,
			  $type, $cond, $val, $file)
	      if $cond ne 'FALSE';
	    push (@var_list, $var);

	    # If the user has set some variables we were in charge
	    # of (which is detected by the first reading of
	    # `header-vars.am'), we must not output them.
	    $result_vars .= "$spacing$comment$_\n"
	      if ($cond ne 'FALSE' && $type ne '+'
		  && exists $var_owner{$var}{$cond || 'TRUE'}
		  && $var_owner{$var}{$cond || 'TRUE'} == VAR_AUTOMAKE);

 	    $comment = $spacing = '';
 	}
 	else
 	{
	    # This isn't an error; it is probably some tokens which
	    # configure is supposed to replace, such as `@SET-MAKE@',
	    # or some part of a rule cut by an if/endif.
  	    if ($cond ne 'FALSE' && ! ($is_rule && $discard_rule))
	      {
		s/^/make_condition (@cond_stack)/gme;
		$result_rules .= "$spacing$comment$_\n";
	      }
 	    $comment = $spacing = '';
 	}
    }

    err_am (@cond_stack ?
	    "unterminated conditionals: @cond_stack" :
	    "too many conditionals closed in include file")
      if "@saved_cond_stack" ne "@cond_stack";

    return ($comment, $result_vars, $result_rules);
}


# $CONTENTS
# &file_contents ($BASENAME, [%TRANSFORM])
# ----------------------------------------
# Return contents of a file from $libdir/am, automatically skipping
# macros or rules which are already known.
sub file_contents ($%)
{
    my ($basename, %transform) = @_;
    my ($comments, $variables, $rules) =
      file_contents_internal (1, "$libdir/am/$basename.am", %transform);
    return "$comments$variables$rules";
}


# $REGEXP
# &transform (%PAIRS)
# -------------------
# Foreach ($TOKEN, $VAL) in %PAIRS produce a replacement expression suitable
# for file_contents which:
#   - replaces %$TOKEN% with $VAL,
#   - enables/disables ?$TOKEN? and ?!$TOKEN?,
#   - replaces %?$TOKEN% with TRUE or FALSE.
sub transform (%)
{
    my (%pairs) = @_;
    my $result = '';

    while (my ($token, $val) = each %pairs)
    {
        $result .= "s/\Q%$token%\E/\Q$val\E/gm;";
	if ($val)
	{
	    $result .= "s/\Q?$token?\E//gm;s/^.*\Q?!$token?\E.*\\n//gm;";
	    $result .= "s/\Q%?$token%\E/TRUE/gm;";
	}
	else
	{
	    $result .= "s/\Q?!$token?\E//gm;s/^.*\Q?$token?\E.*\\n//gm;";
	    $result .= "s/\Q%?$token%\E/FALSE/gm;";
	}
    }

    return $result;
}


# &append_exeext ($MACRO)
# -----------------------
# Macro is an Automake magic macro which primary is PROGRAMS, e.g.
# bin_PROGRAMS.  Make sure these programs have $(EXEEXT) appended.
sub append_exeext ($)
{
  my ($macro) = @_;

  prog_error "append_exeext ($macro)"
    unless $macro =~ /_PROGRAMS$/;

  my @conds = variable_conditions_recursive ($macro);

  my @condvals;
  foreach my $cond (@conds)
    {
      my @one_binlist = ();
      my @condval = variable_value_as_list_recursive ($macro, $cond);
      foreach my $rcurs (@condval)
	{
	  # Skip autoconf substs.  Also skip if the user
	  # already applied $(EXEEXT).
	  if ($rcurs =~ /^\@.*\@$/ || $rcurs =~ /\$\(EXEEXT\)$/)
	    {
	      push (@one_binlist, $rcurs);
	    }
	  else
	    {
	      push (@one_binlist, $rcurs . '$(EXEEXT)');
	    }
	}

      push (@condvals, $cond);
      push (@condvals, "@one_binlist");
    }

  macro_delete ($macro);
  while (@condvals)
    {
      my $cond = shift (@condvals);
      my @val = split (' ', shift (@condvals));
      define_pretty_variable ($macro, $cond, @val);
    }
 }


# @PREFIX
# &am_primary_prefixes ($PRIMARY, $CAN_DIST, @PREFIXES)
# -----------------------------------------------------
# Find all variable prefixes that are used for install directories.  A
# prefix `zar' qualifies iff:
#
# * `zardir' is a variable.
# * `zar_PRIMARY' is a variable.
#
# As a side effect, it looks for misspellings.  It is an error to have
# a variable ending in a "reserved" suffix whose prefix is unknown, eg
# "bni_PROGRAMS".  However, unusual prefixes are allowed if a variable
# of the same name (with "dir" appended) exists.  For instance, if the
# variable "zardir" is defined, then "zar_PROGRAMS" becomes valid.
# This is to provide a little extra flexibility in those cases which
# need it.
sub am_primary_prefixes ($$@)
{
  my ($primary, $can_dist, @prefixes) = @_;

  local $_;
  my %valid = map { $_ => 0 } @prefixes;
  $valid{'EXTRA'} = 0;
  foreach my $varname (keys %var_value)
    {
      # Automake is allowed to define variables that look like primaries
      # but which aren't.  E.g. INSTALL_sh_DATA.
      # Autoconf can also define variables like INSTALL_DATA, so
      # ignore all configure variables (at least those which are not
      # redefined in Makefile.am).
      # FIXME: We should make sure that these variables are not
      # conditionally defined (or else adjust the condition below).
      next
	if (exists $var_owner{$varname}
	    && exists $var_owner{$varname}{'TRUE'}
	    && $var_owner{$varname}{'TRUE'} != VAR_MAKEFILE);

      if ($varname =~ /^(nobase_)?(dist_|nodist_)?(.*)_$primary$/)
	{
	  my ($base, $dist, $X) = ($1 || '', $2 || '', $3 || '');
	  if ($dist ne '' && ! $can_dist)
            {
	      err_var ($varname,
		       "invalid variable `$varname': `dist' is forbidden");
	    }
	  # Standard directories must be explicitely allowed.
	  elsif (! defined $valid{$X} && exists $standard_prefix{$X})
	    {
	      err_var ($varname,
		       "`${X}dir' is not a legitimate directory " .
		       "for `$primary'");
	    }
	  # A not explicitely valid directory is allowed if Xdir is defined.
	  elsif (! defined $valid{$X} &&
		 require_variables_for_macro ($varname, "`$varname' is used",
					      "${X}dir"))
	    {
	      # Nothing to do.  Any error message has been output
	      # by require_variables_for_macro.
	    }
	  else
	    {
	      # Ensure all extended prefixes are actually used.
	      $valid{"$base$dist$X"} = 1;
	    }
	}
    }

  # Return only those which are actually defined.
  return sort grep { variable_defined ($_ . '_' . $primary) } keys %valid;
}


# Handle `where_HOW' variable magic.  Does all lookups, generates
# install code, and possibly generates code to define the primary
# variable.  The first argument is the name of the .am file to munge,
# the second argument is the primary variable (eg HEADERS), and all
# subsequent arguments are possible installation locations.  Returns
# list of all values of all _HOW targets.
#
# FIXME: this should be rewritten to be cleaner.  It should be broken
# up into multiple functions.
#
# Usage is: am_install_var (OPTION..., file, HOW, where...)
sub am_install_var
{
    my (@args) = @_;

    my $do_require = 1;
    my $can_dist = 0;
    my $default_dist = 0;
    while (@args)
    {
	if ($args[0] eq '-noextra')
	{
	    $do_require = 0;
	}
	elsif ($args[0] eq '-candist')
	{
	    $can_dist = 1;
	}
	elsif ($args[0] eq '-defaultdist')
	{
	    $default_dist = 1;
	    $can_dist = 1;
	}
	elsif ($args[0] !~ /^-/)
	{
	    last;
	}
	shift (@args);
    }

    my ($file, $primary, @prefix) = @args;

    # Now that configure substitutions are allowed in where_HOW
    # variables, it is an error to actually define the primary.  We
    # allow `JAVA', as it is customarily used to mean the Java
    # interpreter.  This is but one of several Java hacks.  Similarly,
    # `PYTHON' is customarily used to mean the Python interpreter.
    reject_var $primary, "`$primary' is an anachronism"
      unless $primary eq 'JAVA' || $primary eq 'PYTHON';

    # Get the prefixes which are valid and actually used.
    @prefix = am_primary_prefixes ($primary, $can_dist, @prefix);

    # If a primary includes a configure substitution, then the EXTRA_
    # form is required.  Otherwise we can't properly do our job.
    my $require_extra;

    my @used = ();
    my @result = ();

    # True if the iteration is the first one.  Used for instance to
    # output parts of the associated file only once.
    my $first = 1;
    foreach my $X (@prefix)
    {
	my $nodir_name = $X;
	my $one_name = $X . '_' . $primary;

	my $strip_subdir = 1;
	# If subdir prefix should be preserved, do so.
	if ($nodir_name =~ /^nobase_/)
	  {
	    $strip_subdir = 0;
	    $nodir_name =~ s/^nobase_//;
	  }

	# If files should be distributed, do so.
	my $dist_p = 0;
	if ($can_dist)
	  {
	    $dist_p = (($default_dist && $nodir_name !~ /^nodist_/)
		       || (! $default_dist && $nodir_name =~ /^dist_/));
	    $nodir_name =~ s/^(dist|nodist)_//;
	  }

	# Append actual contents of where_PRIMARY variable to
	# result.
	foreach my $rcurs (&variable_value_as_list_recursive ($one_name, 'all'))
	  {
	    # Skip configure substitutions.  Possibly bogus.
	    if ($rcurs =~ /^\@.*\@$/)
	      {
		if ($nodir_name eq 'EXTRA')
		  {
		    err_var ($one_name,
			     "`$one_name' contains configure substitution, "
			     . "but shouldn't");
		  }
		# Check here to make sure variables defined in
		# configure.ac do not imply that EXTRA_PRIMARY
		# must be defined.
		elsif (! defined $configure_vars{$one_name})
		  {
		    $require_extra = $one_name
		      if $do_require;
		  }

		next;
	      }

	    push (@result, $rcurs);
	  }
	# A blatant hack: we rewrite each _PROGRAMS primary to include
	# EXEEXT.
	append_exeext ($one_name)
	  if $primary eq 'PROGRAMS';
	# "EXTRA" shouldn't be used when generating clean targets,
	# all, or install targets.  We used to warn if EXTRA_FOO was
	# defined uselessly, but this was annoying.
	next
	  if $nodir_name eq 'EXTRA';

	if ($nodir_name eq 'check')
	  {
	    push (@check, '$(' . $one_name . ')');
	  }
	else
	  {
	    push (@used, '$(' . $one_name . ')');
	  }

	# Is this to be installed?
	my $install_p = $nodir_name ne 'noinst' && $nodir_name ne 'check';

	# If so, with install-exec? (or install-data?).
	my $exec_p = ($nodir_name =~ /$EXEC_DIR_PATTERN/o);

	my $check_options_p = $install_p
			      && defined $options{'std-options'};

	# Singular form of $PRIMARY.
	(my $one_primary = $primary) =~ s/S$//;
	$output_rules .= &file_contents ($file,
					 ('FIRST' => $first,

					  'PRIMARY'     => $primary,
					  'ONE_PRIMARY' => $one_primary,
					  'DIR'         => $X,
					  'NDIR'        => $nodir_name,
					  'BASE'        => $strip_subdir,

					  'EXEC'    => $exec_p,
					  'INSTALL' => $install_p,
					  'DIST'    => $dist_p,
					  'CK-OPTS' => $check_options_p));

	$first = 0;
    }

    # The JAVA variable is used as the name of the Java interpreter.
    # The PYTHON variable is used as the name of the Python interpreter.
    if (@used && $primary ne 'JAVA' && $primary ne 'PYTHON')
    {
	# Define it.
	define_pretty_variable ($primary, '', @used);
	$output_vars .= "\n";
    }

    err_var ($require_extra,
	     "`$require_extra' contains configure substitution,\n"
	     . "but `EXTRA_$primary' not defined")
      if ($require_extra && ! variable_defined ('EXTRA_' . $primary));

    # Push here because PRIMARY might be configure time determined.
    push (@all, '$(' . $primary . ')')
	if @used && $primary ne 'JAVA' && $primary ne 'PYTHON';

    # Make the result unique.  This lets the user use conditionals in
    # a natural way, but still lets us program lazily -- we don't have
    # to worry about handling a particular object more than once.
    return uniq (sort @result);
}


################################################################

# Each key in this hash is the name of a directory holding a
# Makefile.in.  These variables are local to `is_make_dir'.
my %make_dirs = ();
my $make_dirs_set = 0;

sub is_make_dir
{
    my ($dir) = @_;
    if (! $make_dirs_set)
    {
	foreach my $iter (@configure_input_files)
	{
	    $make_dirs{dirname ($iter)} = 1;
	}
	# We also want to notice Makefile.in's.
	foreach my $iter (@other_input_files)
	{
	    if ($iter =~ /Makefile\.in$/)
	    {
		$make_dirs{dirname ($iter)} = 1;
	    }
	}
	$make_dirs_set = 1;
    }
    return defined $make_dirs{$dir};
}

################################################################

# This variable is local to the "require file" set of functions.
my @require_file_paths = ();


# &maybe_push_required_file ($DIR, $FILE, $FULLFILE)
# --------------------------------------------------
# See if we want to push this file onto dist_common.  This function
# encodes the rules for deciding when to do so.
sub maybe_push_required_file
{
    my ($dir, $file, $fullfile) = @_;

    if ($dir eq $relative_dir)
    {
	push_dist_common ($file);
	return 1;
    }
    elsif ($relative_dir eq '.' && ! &is_make_dir ($dir))
    {
	# If we are doing the topmost directory, and the file is in a
	# subdir which does not have a Makefile, then we distribute it
	# here.
	push_dist_common ($fullfile);
	return 1;
    }
    return 0;
}


# &require_file_internal ($WHERE, $MYSTRICT, @FILES)
# --------------------------------------------------
# Verify that the file must exist in the current directory.
# $MYSTRICT is the strictness level at which this file becomes required.
#
# Must set require_file_paths before calling this function.
# require_file_paths is set to hold a single directory (the one in
# which the first file was found) before return.
sub require_file_internal ($$@)
{
    my ($where, $mystrict, @files) = @_;

    foreach my $file (@files)
    {
        my $fullfile;
	my $errdir;
	my $errfile;
	my $save_dir;

	my $found_it = 0;
	my $dangling_sym = 0;
	foreach my $dir (@require_file_paths)
	{
	    $fullfile = $dir . "/" . $file;
	    $errdir = $dir unless $errdir;

	    # Use different name for "error filename".  Otherwise on
	    # an error the bad file will be reported as eg
	    # `../../install-sh' when using the default
	    # config_aux_path.
	    $errfile = $errdir . '/' . $file;

	    if (-l $fullfile && ! -f $fullfile)
	    {
		$dangling_sym = 1;
		last;
	    }
	    elsif (-f $fullfile)
	    {
		$found_it = 1;
		maybe_push_required_file ($dir, $file, $fullfile);
		$save_dir = $dir;
		last;
	    }
	}

	# `--force-missing' only has an effect if `--add-missing' is
	# specified.
	if ($found_it && (! $add_missing || ! $force_missing))
	{
	    # Prune the path list.
	    @require_file_paths = $save_dir;
	}
	else
	{
	    # If we've already looked for it, we're done.  You might
	    # wonder why we don't do this before searching for the
	    # file.  If we do that, then something like
	    # AC_OUTPUT(subdir/foo foo) will fail to put foo.in into
	    # DIST_COMMON.
	    if (! $found_it)
	    {
		next if defined $require_file_found{$fullfile};
		$require_file_found{$fullfile} = 1;
	    }

	    if ($strictness >= $mystrict)
	    {
		if ($dangling_sym && $add_missing)
		{
		    unlink ($fullfile);
		}

		my $trailer = '';
		my $suppress = 0;

		# Only install missing files according to our desired
		# strictness level.
		my $message = "required file `$errfile' not found";
		if ($add_missing)
		{
		    if (-f ("$libdir/$file"))
		    {
		        $suppress = 1;

			# Install the missing file.  Symlink if we
			# can, copy if we must.  Note: delete the file
			# first, in case it is a dangling symlink.
			$message = "installing `$errfile'";
			# Windows Perl will hang if we try to delete a
			# file that doesn't exist.
			unlink ($errfile) if -f $errfile;
			if ($symlink_exists && ! $copy_missing)
			{
			    if (! symlink ("$libdir/$file", $errfile))
			    {
				$suppress = 0;
				$trailer = "; error while making link: $!";
			    }
			}
			elsif (system ('cp', "$libdir/$file", $errfile))
			{
			    $suppress = 0;
			    $trailer = "\n    error while copying";
			}
		    }

		    if (! maybe_push_required_file (dirname ($errfile),
                                                    $file, $errfile))
		    {
			if (! $found_it)
			{
			    # We have added the file but could not push it
			    # into DIST_COMMON (probably because this is
			    # an auxiliary file and we are not processing
			    # the top level Makefile). This is unfortunate,
			    # since it means we are using a file which is not
			    # distributed!

			    # Get Automake to be run again: on the second
			    # run the file will be found, and pushed into
			    # the toplevel DIST_COMMON automatically.
			    $automake_needs_to_reprocess_all_files = 1;
			}
		    }

		    # Prune the path list.
		    @require_file_paths = &dirname ($errfile);
		}

		# If --force-missing was specified, and we have
		# actually found the file, then do nothing.
		next
		    if $found_it && $force_missing;

		# If we couldn' install the file, but it is a target in
		# the Makefile, don't print anything.  This allows files
		# like README, AUTHORS, or THANKS to be generated.
		next
		  if !$suppress && target_defined ($file);

		msg ($suppress ? 'note' : 'error', $where, "$message$trailer");
	    }
	}
    }
}

# &require_file ($WHERE, $MYSTRICT, @FILES)
# -----------------------------------------
sub require_file ($$@)
{
    my ($where, $mystrict, @files) = @_;
    @require_file_paths = $relative_dir;
    require_file_internal ($where, $mystrict, @files);
}

# &require_file_with_macro ($COND, $MACRO, $MYSTRICT, @FILES)
# -----------------------------------------------------------
sub require_file_with_macro ($$$@)
{
    my ($cond, $macro, $mystrict, @files) = @_;
    require_file ($var_location{$macro}{$cond}, $mystrict, @files);
}


# &require_conf_file ($WHERE, $MYSTRICT, @FILES)
# ----------------------------------------------
# Looks in configuration path, as specified by AC_CONFIG_AUX_DIR.
sub require_conf_file ($$@)
{
    my ($where, $mystrict, @files) = @_;
    @require_file_paths = @config_aux_path;
    require_file_internal ($where, $mystrict, @files);
    my $dir = $require_file_paths[0];
    @config_aux_path = @require_file_paths;
     # Avoid unsightly '/.'s.
    $config_aux_dir = '$(top_srcdir)' . ($dir eq '.' ? "" : "/$dir");
}


# &require_conf_file_with_macro ($COND, $MACRO, $MYSTRICT, @FILES)
# ----------------------------------------------------------------
sub require_conf_file_with_macro ($$$@)
{
    my ($cond, $macro, $mystrict, @files) = @_;
    require_conf_file ($var_location{$macro}{$cond}, $mystrict, @files);
}

################################################################

# &require_build_directory ($DIRECTORY)
# ------------------------------------
# Emit rules to create $DIRECTORY if needed, and return
# the file that any target requiring this directory should be made
# dependent upon.
sub require_build_directory ($)
{
  my $directory = shift;
  my $dirstamp = "$directory/\$(am__dirstamp)";

  # Don't emit the rule twice.
  if (! defined $directory_map{$directory})
    {
      $directory_map{$directory} = 1;

      # Set a variable for the dirstamp basename.
      define_pretty_variable ('am__dirstamp', 'TRUE',
			      '$(am__leading_dot)dirstamp')
	unless variable_defined ('am__dirstamp');

      # Directory must be removed by `make distclean'.
      $clean_files{$dirstamp} = DIST_CLEAN;

      $output_rules .= ("$dirstamp:\n"
			. "\t\@\$(mkinstalldirs) $directory\n"
			. "\t\@: > $dirstamp\n");
    }

  return $dirstamp;
}

# &require_build_directory_maybe ($FILE)
# --------------------------------------
# If $FILE lies in a subdirectory, emit a rule to create this
# directory and return the file that $FILE should be made
# dependent upon.  Otherwise, just return the empty string.
sub require_build_directory_maybe ($)
{
    my $file = shift;
    my $directory = dirname ($file);

    if ($directory ne '.')
    {
	return require_build_directory ($directory);
    }
    else
    {
	return '';
    }
}

################################################################

# Push a list of files onto dist_common.
sub push_dist_common
{
  prog_error "push_dist_common run after handle_dist"
    if $handle_dist_run;
  macro_define ('DIST_COMMON', VAR_AUTOMAKE, '+', '', "@_", '');
}


# Set strictness.
sub set_strictness
{
  $strictness_name = $_[0];

  # FIXME: 'portability' warnings are currently disabled by default.
  # Eventually we want to turn them on in GNU and GNITS modes, but
  # we don't do this yet in Automake 1.7 to help the 1.6/1.7 transition.
  #
  # Indeed there would be only two ways to get rid of these new warnings:
  #  1. adjusting Makefile.am
  #     This is not always easy (or wanted).  Consider %-rules or
  #     $(function args) variables.
  #  2. using -Wno-portability
  #     This means there is no way to have the same Makefile.am
  #     working both with Automake 1.6 and 1.7 (since 1.6 does not
  #     understand -Wno-portability).
  #
  # In Automake 1.8 (or whatever it is called) we can turn these
  # warnings on, since -Wno-portability will not be an issue for
  # the 1.7/1.8 transition.
  if ($strictness_name eq 'gnu')
    {
      $strictness = GNU;
      setup_channel 'error-gnu', silent => 0;
      setup_channel 'error-gnu/warn', silent => 0, type => 'error';
      setup_channel 'error-gnits', silent => 1;
      # setup_channel 'portability', silent => 0;
      setup_channel 'gnu', silent => 0;
    }
  elsif ($strictness_name eq 'gnits')
    {
      $strictness = GNITS;
      setup_channel 'error-gnu', silent => 0;
      setup_channel 'error-gnu/warn', silent => 0, type => 'error';
      setup_channel 'error-gnits', silent => 0;
      # setup_channel 'portability', silent => 0;
      setup_channel 'gnu', silent => 0;
    }
  elsif ($strictness_name eq 'foreign')
    {
      $strictness = FOREIGN;
      setup_channel 'error-gnu', silent => 1;
      setup_channel 'error-gnu/warn', silent => 0, type => 'warning';
      setup_channel 'error-gnits', silent => 1;
      # setup_channel 'portability', silent => 1;
      setup_channel 'gnu', silent => 1;
    }
  else
    {
      prog_error "level `$strictness_name' not recognized\n";
    }
}


################################################################

# Glob something.  Do this to avoid indentation screwups everywhere we
# want to glob.  Gross!
sub my_glob
{
    my ($pat) = @_;
    return <${pat}>;
}

################################################################

# INTEGER
# require_variables ($WHERE, $REASON, $COND, @VARIABLES)
# ------------------------------------------------------
# Make sure that each supplied variable is defined in $COND.
# Otherwise, issue a warning.  If we know which macro can
# define this variable, hint the user.
# Return the number of undefined variables.
sub require_variables ($$$@)
{
  my ($where, $reason, $cond, @vars) = @_;
  my $res = 0;
  $reason .= ' but ' unless $reason eq '';

 VARIABLE:
  foreach my $var (@vars)
    {
      # Nothing to do if the variable exists.  The $configure_vars test
      # needed for strange variables like AMDEPBACKSLASH or ANSI2KNR
      # that are AC_SUBST'ed but never macro_define'd.
      next VARIABLE
	if ((exists $var_value{$var} && exists $var_value{$var}{$cond})
	    || exists $configure_vars{$var});

      my @undef_cond = variable_not_always_defined_in_cond $var, $cond;
      next VARIABLE
	unless @undef_cond;

      my $text = "$reason`$var' is undefined\n";
      if (@undef_cond && $undef_cond[0] ne 'TRUE')
	{
	  $text .= ("in the following conditions:\n  "
		    . join ("\n  ", @undef_cond));
	}

      ++$res;

      if (exists $am_macro_for_var{$var})
	{
	  $text .= "\nThe usual way to define `$var' is to add "
	    . "`$am_macro_for_var{$var}'\nto `$configure_ac' and run "
	    . "`aclocal' and `autoconf' again.";
	}
      elsif (exists $ac_macro_for_var{$var})
	{
	  $text .= "\nThe usual way to define `$var' is to add "
	    . "`$ac_macro_for_var{$var}'\nto `$configure_ac' and run "
	    . "`autoconf' again.";
	}

      error $where, $text, uniq_scope => US_GLOBAL;
    }
  return $res;
}

# INTEGER
# require_variables_for_macro ($MACRO, $REASON, @VARIABLES)
# ---------------------------------------------------------
# Same as require_variables, but take a macro mame as first argument.
sub require_variables_for_macro ($$@)
{
  my ($macro, $reason, @args) = @_;
  for my $cond (keys %{$var_value{$macro}})
    {
      return require_variables ($var_location{$macro}{$cond}, $reason,
				$cond, @args);
    }
}

# Print usage information.
sub usage ()
{
    print "Usage: $0 [OPTION] ... [Makefile]...

Generate Makefile.in for configure from Makefile.am.

Operation modes:
      --help               print this help, then exit
      --version            print version number, then exit
  -v, --verbose            verbosely list files processed
      --no-force           only update Makefile.in's that are out of date
  -W, --warnings=CATEGORY  report the warnings falling in CATEGORY

Dependency tracking:
  -i, --ignore-deps      disable dependency tracking code
      --include-deps     enable dependency tracking code

Flavors:
      --cygnus           assume program is part of Cygnus-style tree
      --foreign          set strictness to foreign
      --gnits            set strictness to gnits
      --gnu              set strictness to gnu

Library files:
  -a, --add-missing      add missing standard files to package
      --libdir=DIR       directory storing library files
  -c, --copy             with -a, copy missing files (default is symlink)
  -f, --force-missing    force update of standard files

Warning categories include:
  `gnu'           GNU coding standards (default in gnu and gnits modes)
  `obsolete'      obsolete features or constructions
  `portability'   portability issues
  `syntax'        dubious syntactic constructs (default)
  `unsupported'   unsupported or incomplete features (default)
  `all'           all the warnings
  `no-CATEGORY'   turn off warnings in CATEGORY
  `none'          turn off all the warnings
  `error'         treat warnings as errors
";

    my ($last, @lcomm);
    $last = '';
    foreach my $iter (sort ((@common_files, @common_sometimes)))
    {
	push (@lcomm, $iter) unless $iter eq $last;
	$last = $iter;
    }

    my @four;
    print "\nFiles which are automatically distributed, if found:\n";
    format USAGE_FORMAT =
  @<<<<<<<<<<<<<<<<   @<<<<<<<<<<<<<<<<   @<<<<<<<<<<<<<<<<   @<<<<<<<<<<<<<<<<
  $four[0],           $four[1],           $four[2],           $four[3]
.
    $~ = "USAGE_FORMAT";

    my $cols = 4;
    my $rows = int(@lcomm / $cols);
    my $rest = @lcomm % $cols;

    if ($rest)
    {
	$rows++;
    }
    else
    {
	$rest = $cols;
    }

    for (my $y = 0; $y < $rows; $y++)
    {
	@four = ("", "", "", "");
	for (my $x = 0; $x < $cols; $x++)
	{
	    last if $y + 1 == $rows && $x == $rest;

	    my $idx = (($x > $rest)
		       ?  ($rows * $rest + ($rows - 1) * ($x - $rest))
		       : ($rows * $x));

	    $idx += $y;
	    $four[$x] = $lcomm[$idx];
	}
	write;
    }

    print "\nReport bugs to <bug-automake\@gnu.org>.\n";

    # --help always returns 0 per GNU standards.
    exit 0;
}


# &version ()
# -----------
# Print version information
sub version ()
{
  print <<EOF;
automake (GNU $PACKAGE) $VERSION
Written by Tom Tromey <tromey\@redhat.com>.

Copyright 2003 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
EOF
  # --version always returns 0 per GNU standards.
  exit 0;
}

### Setup "GNU" style for perl-mode and cperl-mode.
## Local Variables:
## perl-indent-level: 2
## perl-continued-statement-offset: 2
## perl-continued-brace-offset: 0
## perl-brace-offset: 0
## perl-brace-imaginary-offset: 0
## perl-label-offset: -2
## cperl-indent-level: 2
## cperl-brace-offset: 0
## cperl-continued-brace-offset: 0
## cperl-label-offset: -2
## cperl-extra-newline-before-brace: t
## cperl-merge-trailing-else: nil
## cperl-continued-statement-offset: 2
## End: