latexmk.pl

#!/usr/bin/env perl
# On a UNIX-like system, the above enables latexmk to run independently
#   of the location of the perl executable.  This line relies on the 
#   existence of the program /usr/bin/env
# If there is a problem for any reason, you can replace the first line of 
#   this file by:

#!/usr/bin/perl -w

# with the path of the perl executable adjusted for your system. 


use warnings;

# Delete #??!! when working

# See ?? <===============================

## ?? Issues with clean-up
## List of aux files deleted is those read, not those generated.
## Other files are generated by (pdf)latex; should they be deleted?
## (I have hooks for this).


#=======================================


#??  Force mode doesn't appear to do force (if error in latex file)
#??? Get banner back in.
#??  CORRECT DIAGNOSTICS ON CHANGED FILES IF THEY DIDN'T EXIST BEFORE
#??  Further corrections to deal with disappeared source files for custom dependencies.
#       Message repeatedly appears about remake when source file of cusdep doesn't exist.
#??  logfile w/o fdb file: don't set changed file, perhaps for generated exts.
#    Reconsider
#??  Do proper run-stuff for bibtex, makeindex, cus-deps.  OK I think
#    Parse and correctly find bst and ist files


# ATTEMPT TO ALLOW FILENAMES WITH SPACES:
#    (as of 1 Apr 2006, and then 14 Sep. 2007)

# Problems:
# A.  Quoting filenames will not always work.  
#        a.  Under UNIX, quotes are legal in filenames, so when PERL
#            directly runs a binary, a quoted filename will be treated as
#            as a filename containing a quote character.  But when it calls
#            a shell, the quotes are handled by the shell as quotes.
#        b.  Under MSWin32, quotes are illegal filename characters, and tend
#            to be handled correctly.
#        c.  But under cygwin, results are not so clear (there are many 
#            combinations: native v. cygwin perl, native v cygwin programs
#            NT v. unix scripts, which shell is called.
# B.  TeX doesn't always handle filenames with spaces gracefully.
#        a.  UNIX/LINUX: The version on gluon2 Mar 31, 2006 to Sep. 2007) 
#            doesn't handle them at all.  (TeX treats space as separator.)
#        b.  At least some later versions actually do (Brad Miller e-mail, 
#            Sep. 2007).
#        c.  fptex [[e-TeXk, Version 3.141592-2.1 (Web2c 7.5.2)] does, on 
#            my MSWin at home.  In \input the filename must be in quotes.
#        d.  Bibtex [BibTeX (Web2c 7.5.2) 0.99c on my MSWin system at home,
#            Sep. 2007] does not allow names of bibfiles to have spaces.
# C.  =====> Using the shell for command lines is not safe, since special 
#     characters can cause lots of mayhem.
#     It will therefore be a good idea to sanitize filenames. 
#
# I've sanitized all calls out:
#     a. system and exec use a single argument, which forces
#        use of shell, under all circumstances
#        Thus I can safely use quotes on filenames:  They will be handled by 
#        the shell under UNIX, and simply passed on to the program under MSWin32.
#     b. I reorganized Run, Run_Detached to use single command line
#     c. All calls to Run and Run_Detached have quoted filenames.
#     d. So if a space-free filename with wildcards is given on latexmk's
#        command line, and it globs to space-containing filename(s), that
#        works (fptex on home computer, native NT tex)
#     e. ====> But globbing fails: the glob function takes space as filename 
#        separator.   ====================

#================= TO DO ================
#
# 1.  See ??  ESPECIALLY $MSWin_fudge_break
# 2.  Check fudged conditions in looping and make_files 
# 3.  Should not completely abort after a run that ends in failure from latex
#     Missing input files (including via custom dependency) should be checked for
#     a change in status
#         If sources for missing files from custom dependency 
#             are available, then do a rerun
#         If sources of any kind become available rerun (esp. for pvc)
#             rerun
#         Must parse log_file after unsuccessful run of latex: it may give
#             information about missing files. 
# 4.  Check file of bug reports and requests
# 5.  Rationalize bibtex warnings and errors.  Two almost identical routines.
#         Should 1. Use single routine
#                2. Convert errors to failure only in calling routine
#                3. Save first warning/error.

# ?? Use of generated_exts arrays and hashes needs rationalization

# To do: 
#   Rationalize again handling of include files.
#     Now I use kpsewhich to do searches, if file not found
#        (How do I avoid getting slowed down too much?)
#   Document the assumptions at each stage of processing algorithm.
#   Option to restart previewer automatically, if it dies under -pvc
#   Test for already running previewer gets wrong answer if another
#     process has the viewed file in its command line

$my_name = 'latexmk';
$My_name = 'Latexmk';
$version_num = '4.18';
$version_details = "$My_name, John Collins, 27 July 2010";


use Config;
use File::Copy;
use File::Basename;
use FileHandle;
use File::Find;
use Cwd;            # To be able to change cwd
use Cwd "chdir";    # Ensure $ENV{PWD}  tracks cwd
use Digest::MD5;

#use strict;

# The following variables are assigned once and then used in symbolic 
#     references, so we need to avoid warnings 'name used only once':
use vars qw( $dvi_update_command $ps_update_command $pdf_update_command );

# Translation of signal names to numbers and vv:
%signo = ();
@signame = ();
if ( defined $Config{sig_name} ) {
   $i = 0;
   foreach $name (split(' ', $Config{sig_name})) {
      $signo{$name} = $i;
      $signame[$i] = $name;
      $i++;
   }
}
else {
   warn "Something wrong with the perl configuration: No signals?\n";
}

## Copyright John Collins 1998-2010
##           (username collins at node phys.psu.edu)
##      (and thanks to David Coppit (username david at node coppit.org) 
##           for suggestions) 
## Copyright Evan McLean
##         (modifications up to version 2)
## Copyright 1992 by David J. Musliner and The University of Michigan.
##         (original version)
##
##    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 of the License, 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
##
##
##
##   NEW FEATURES, since v. 2.0:
##     1.  Correct algorithm for deciding how many times to run latex:
##         based on whether source file(s) change between runs
##     2.  Continuous preview works, and can be of ps file or dvi file
##     3.  pdf creation by pdflatex possible
##     4.  Defaults for commands are OS dependent.
##     5.  Parsing of log file instead of source file is used to
##         obtain dependencies, by default.
##
##   Modification log from 23 Jun 2010 onwards in detail
##
##     27 Jul 2010, John Collins  Clean up some comments.  
##                                This is release version for 4.18.
##     23 Jul 2010, John Collins  Only produce "Removing no-longer-needed
##                                  dependent" messages when diagnostics are
##                                  on.
##     23 Jun 2010, John Collins  Add test in log file for natbib's
##                                citation-undefined warning
##     23 Jun 2010, John Collins  V. 4.18
##
##   1998-2010, John Collins.  Many improvements and fixes.
##       See CHANGE-log.txt for full list, and CHANGES for summary
##
##   Modified by Evan McLean (no longer available for support)
##   Original script (RCS version 2.3) called "go" written by David J. Musliner
##
## 2.0 - Final release, no enhancements.  LatexMk is no longer supported
##       by the author.
## 1.9 - Fixed bug that was introduced in 1.8 with path name fix.
##     - Fixed buglet in man page.
## 1.8 - Add not about announcement mailling list above.
##     - Added texput.dvi and texput.aux to files deleted with -c and/or
##       the -C options.
##     - Added landscape mode (-l option and a bunch of RC variables).
##     - Added sensing of "\epsfig{file=...}" forms in dependency generation.
##     - Fixed path names when specified tex file is not in the current
##       directory.
##     - Fixed combined use of -pvc and -s options.
##     - Fixed a bunch of speling errors in the source. :-)
##     - Fixed bugs in xdvi patches in contrib directory.
## 1.7 - Fixed -pvc continuous viewing to reattach to pre-existing
##       process correctly.
##     - Added $pscmd to allow changing process grepping for different
##       systems.
## 1.6 - Fixed buglet in help message
##     - Fixed bugs in detection of input and include files.
## 1.5 - Removed test message I accidentally left in version 1.4
##     - Made dvips use -o option instead of stdout redirection as some
##       people had problems with dvips not going to stdout by default.
##     - Fixed bug in input and include file detection
##     - Fixed dependency resolution process so it detects new .toc file
##       and makeindex files properly.
##     - Added dvi and postscript filtering options -dF and -pF.
##     - Added -v version commmand.
## 1.4 - Fixed bug in -pvc option.
##     - Made "-F" option include non-existant file in the dependency list.
##       (RC variable: $force_include_mode)
##     - Added .lot and .lof files to clean up list of extensions.
##     - Added file "texput.log" to list of files to clean for -c.
##     - LatexMk now handles file names in a similar fashion to latex.
##       The ".tex" extension is no longer enforced.
##     - Added $texfile_search RC variable to look for default files.
##     - Fixed \input and \include so they add ".tex" extension if necessary.
##     - Allow intermixing of file names and options.
##     - Added "-d" and banner options (-bm, -bs, and -bi).
##       (RC variables: $banner, $banner_message, $banner_scale,
##       $banner_intensity, $tmpdir)
##     - Fixed "-r" option to detect an command line syntax errors better.
## 1.3 - Added "-F" option, patch supplied by Patrick van der Smagt.
## 1.2 - Added "-C" option.
##     - Added $clean_ext and $clean_full_ext variables for RC files.
##     - Added custom dependency generation capabilities.
##     - Added command line and variable to specify custom RC file.
##     - Added reading of rc file in current directly.
## 1.1 - Fixed bug where Dependency file generation header is printed
##       rependatively.
##     - Fixed bug where TEXINPUTS path is searched for file that was
##       specified with absolute an pathname.
## 1.0 - Ripped from script by David J. Musliner (RCS version 2.3) called "go"
##     - Fixed a couple of file naming bugs
##        e.g. when calling latex, left the ".tex" extension off the end
##             of the file name which could do some interesting things
##             with some file names.
##     - Redirected output of dvips.  My version of dvips was a filter.
##     - Cleaned up the rc file mumbo jumbo and created a dependency file
##       instead.  Include dependencies are always searched for if a
##       dependency file doesn't exist.  The -i option regenerates the
##       dependency file.
##       Getting rid of the rc file stuff also gave the advantage of
##       not being restricted to one tex file per directory.
##     - Can specify multiple files on the command line or no files
##       on the command line.
##     - Removed lpr options stuff.  I would guess that generally,
##       you always use the same options in which case they can
##       be set up from an rc file with the $lpr variable.
##     - Removed the dviselect stuff.  If I ever get time (or money :-) )
##       I might put it back in if I find myself needing it or people
##       express interest in it.
##     - Made it possible to view dvi or postscript file automatically
##       depending on if -ps option selected.
##     - Made specification of dvi file viewer seperate for -pv and -pvc
##       options.
##-----------------------------------------------------------------------


## Explicit exit codes: 
##             10 = bad command line arguments
##             11 = file specified on command line not found
##                  or other file not found
##             12 = failure in some part of making files
##             13 = error in initialization file
##             20 = probable bug
##             or retcode from called program.


#Line length in log file that indicates wrapping.  
# This number EXCLUDES line-end characters, and is one-based
# It is the parameter max_print_line in the TeX program.  (tex.web)
$log_wrap = 79;

#########################################################################
## Default parsing and file-handling settings

## Array of reg-exps for patterns in log-file for file-not-found
## Each item is the string in a regexp, without the enclosing slashes.
## First parenthesized part is the filename.
## Note the need to quote slashes and single right quotes to make them 
## appear in the regexp.
## Add items by push, e.g.,
##     push @file_not_found, '^No data file found `([^\\\']*)\\\'';
## will give match to line starting "No data file found `filename'"
@file_not_found = (
    '^No file\\s*(.*)\\.$',
    '^\\! LaTeX Error: File `([^\\\']*)\\\' not found\\.',
    '.*?:\\d*: LaTeX Error: File `([^\\\']*)\\\' not found\\.',
    '^LaTeX Warning: File `([^\\\']*)\\\' not found',
    '^Package .* [fF]ile `([^\\\']*)\\\' not found',
    'Error: pdflatex \(file ([^\)]*)\): cannot find image file',
);

## Hash mapping file extension (w/o period, e.g., 'eps') to a single regexp,
#  whose matching by a line in a file with that extension indicates that the 
#  line is to be ignored in the calculation of the hash number (md5 checksum)
#  for the file.  Typically used for ignoring datestamps in testing whether 
#  a file has changed.
#  Add items e.g., by
#     $hash_calc_ignore_pattern{'eps'} = '^%%CreationDate: ';
#  This makes the hash calculation for an eps file ignore lines starting with
#  '%%CreationDate: '
#  ?? Note that a file will be considered changed if 
#       (a) its size changes
#    or (b) its hash changes
#  So it is useful to ignore lines in the hash calculation only if they
#  are of a fixed size (as with a date/time stamp).
%hash_calc_ignore_pattern =();

#########################################################################
## Default document processing programs, and related settings,
## These are mostly the same on all systems.
## Most of these variables represents the external command needed to 
## perform a certain action.  Some represent switches.

## Commands to invoke latex, pdflatex
$latex  = 'latex %O %S';
$pdflatex = 'pdflatex %O %S';
## Switch(es) to make them silent:
$latex_silent_switch  = '-interaction=batchmode';
$pdflatex_silent_switch  = '-interaction=batchmode';

# %input_extensions maps primary_rule_name to pointer to hash of file extensions
#    used for extensionless files specified in the source file by constructs
#    like \input{file}  \includegraphics{file}
#  Could write
#%input_extensions = ( 'latex' => { 'tex' => 1, 'eps' => 1 };, 
#        'pdflatex' => { 'tex' => 1, 'pdf' => 1, 'jpg' => 1, 'png' => 1 }; );
# Instead we'll exercise the user-friendly access routines:
add_input_ext( 'latex', 'tex', 'eps' );
add_input_ext( 'pdflatex', 'tex', 'jpg', 'pdf' );
#show_input_ext( 'latex' ); show_input_ext( 'pdflatex' );

## Command to invoke bibtex
$bibtex  = 'bibtex %O %B';
# Switch(es) to make bibtex silent:
$bibtex_silent_switch  = '-terse';
$bibtex_use = 1;   # Whether to actually run bibtex to update bbl files
                   # 0:  Never run bibtex
                   # 1:  Run bibtex only if the bibfiles exists 
                   #     according to kpsewhich, and the bbl files
                   #     appear to be out-of-date
                   # 2:  Run bibtex when the bbl files are out-of-date
                   # In any event bibtex is only run if the log file
                   #   indicates that the document uses bbl files.

## Command to invoke makeindex
$makeindex  = 'makeindex %O -o %D %S';
# Switch(es) to make makeinex silent:
$makeindex_silent_switch  = '-q';

## Command to convert dvi file to pdf file directly:
$dvipdf  = 'dvipdf %O %S %D';

## Command to convert dvi file to ps file:
$dvips  = 'dvips %O -o %D %S';
## Command to convert dvi file to ps file in landscape format:
$dvips_landscape = 'dvips -tlandscape %O -o %D %S';
# Switch(es) to get dvips to make ps file suitable for conversion to good pdf:
#    (If this is not used, ps file and hence pdf file contains bitmap fonts
#       (type 3), which look horrible under acroread.  An appropriate switch
#       ensures type 1 fonts are generated.  You can put this switch in the 
#       dvips command if you prefer.)
$dvips_pdf_switch = '-P pdf';
# Switch(es) to make dvips silent:
$dvips_silent_switch  = '-q';

## Command to convert ps file to pdf file:
$ps2pdf = 'ps2pdf  %O %S %D';

## Command to search for tex-related files
$kpsewhich = 'kpsewhich %S';


##Printing:
$print_type = 'ps';     # When printing, print the postscript file.
                        # Possible values: 'dvi', 'ps', 'pdf', 'none'

## Which treatment of default extensions and filenames with
##   multiple extensions is used, for given filename on
##   tex/latex's command line?  See sub find_basename for the
##   possibilities. 
## Current tex's treat extensions like UNIX teTeX:
$extension_treatment = 'unix';

$dvi_update_signal = undef;
$ps_update_signal = undef;
$pdf_update_signal = undef;

$dvi_update_command = undef;
$ps_update_command = undef;
$pdf_update_command = undef;

$new_viewer_always = 0;     # If 1, always open a new viewer in pvc mode.
                            # If 0, only open a new viewer if no previous
                            #     viewer for the same file is detected.

$quote_filenames = 1;       # Quote filenames in external commands



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

################################################################
##  Special variables for system-dependent fudges, etc.
$log_file_binary = 0;   # Whether to treat log file as binary
                        # Normally not, since the log file SHOULD be pure text.
                        # But Miktex 2.7 sometimes puts binary characters
                        #    in it.  (Typically in construct \OML ... after
                        #    overfull box with mathmode.)
                        # Sometimes there is ctrl/Z, which is not only non-text, 
                        #    but is end-of-file marker for MS-Win in text mode.  

$MSWin_fudge_break = 1; # Give special treatment to ctrl/C and ctrl/break
                        #    in -pvc mode under MSWin
                        # Under MSWin32 (at least with perl 5.8 and WinXP)
                        #   when latemk is running another program, and the 
                        #   user gives ctrl/C or ctrl/break, to stop the 
                        #   daughter program, not only does it reach
                        #   the daughter, but also latexmk/perl, so
                        #   latexmk is stopped also.  In -pvc mode,
                        #   this is not normally desired.  So when the
                        #   $MSWin_fudge_break variable is set,
                        #   latexmk arranges to ignore ctrl/C and
                        #   ctrl/break during processing of files;
                        #   only the daughter programs receive them.
                        # This fudge is not applied in other
                        #   situations, since then having latexmk also
                        #   stopping because of the ctrl/C or
                        #   ctrl/break signal is desirable.
                        # The fudge is not needed under UNIX (at least
                        #   with Perl 5.005 on Solaris 8).  Only the
                        #   daughter programs receive the signal.  In
                        #   fact the inverse would be useful: In
                        #   normal processing, as opposed to -pvc, if
                        #   force mode (-f) is set, a ctrl/C is
                        #   received by a daughter program does not
                        #   also stop latexmk.  Under tcsh, we get
                        #   back to a command prompt, while latexmk
                        #   keeps running in the background!


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


# System-dependent overrides:
if ( $^O eq "MSWin32" ) {
# Pure MSWindows configuration
    ## Configuration parameters:

    ## Use first existing case for $tmpdir:
    $tmpdir = $ENV{TMPDIR} || $ENV{TEMP} || '.';
    $log_file_binary = 1;   # Protect against ctrl/Z in log file from
                            # Miktex 2.7.

    ## List of possibilities for the system-wide initialization file.  
    ## The first one found (if any) is used.
    @rc_system_files = ( 'C:/latexmk/LatexMk' );

    $search_path_separator = ';';  # Separator of elements in search_path

    # For a pdf-file, "start x.pdf" starts the pdf viewer associated with
    #   pdf files, so no program name is needed:
    $pdf_previewer = 'start %O %S';
    $ps_previewer  = 'start %O %S';
    $ps_previewer_landscape  = $ps_previewer;
    $dvi_previewer  = 'start %O %S';
    $dvi_previewer_landscape = "$dvi_previewer";
    # Viewer update methods: 
    #    0 => auto update: viewer watches file (e.g., gv)
    #    1 => manual update: user must do something: e.g., click on window.
    #         (e.g., ghostview, MSWIN previewers, acroread under UNIX)
    #    2 => send signal.  Number of signal in $dvi_update_signal,
    #                         $ps_update_signal, $pdf_update_signal
    #    3 => viewer can't update, because it locks the file and the file 
    #         cannot be updated.  (acroread under MSWIN)
    #    4 => run a command to force the update.  The commands are 
    #         specified by the variables $dvi_update_command, 
    #         $ps_update_command, $pdf_update_command
    $dvi_update_method = 1;
    $ps_update_method = 1;
    $pdf_update_method = 3; # acroread locks the pdf file
    # Use NONE as flag that I am not implementing some commands:
    $lpr =
        'NONE $lpr variable is not configured to allow printing of ps files';
    $lpr_dvi =
        'NONE $lpr_dvi variable is not configured to allow printing of dvi files';
    $lpr_pdf =
        'NONE $lpr_pdf variable is not configured to allow printing of pdf files';
    # The $pscmd below holds a command to list running processes.  It
    # is used to find the process ID of the viewer looking at the
    # current output file.  The output of the command must include the
    # process number and the command line of the processes, since the
    # relevant process is identified by the name of file to be viewed.
    # Its use is not essential.
    $pscmd = 
        'NONE $pscmd variable is not configured to detect running processes';
    $pid_position = -1;     # offset of PID in output of pscmd.  
                            # Negative means I cannot use ps
}
elsif ( $^O eq "cygwin" ) {
    # The problem is a mixed MSWin32 and UNIX environment. 
    # Perl decides the OS is cygwin in two situations:
    # 1. When latexmk is run from a cygwin shell under a cygwin
    #    environment.  Perl behaves in a UNIX way.  This is OK, since
    #    the user is presumably expecting UNIXy behavior.  
    # 2. When CYGWIN exectuables are in the path, but latexmk is run
    #    from a native NT shell.  Presumably the user is expecting NT
    #    behavior. But perl behaves more UNIXy.  This causes some
    #    clashes. 
    # The issues to handle are:
    # 1.  Perl sees both MSWin32 and cygwin filenames.  This is 
    #     normally only an advantage.
    # 2.  Perl uses a UNIX shell in the system command
    #     This is a nasty problem: under native NT, there is a
    #     start command that knows about NT file associations, so that
    #     we can do, e.g., (under native NT) system("start file.pdf");
    #     But this won't work when perl has decided the OS is cygwin,
    #     even if it is invoked from a native NT command line.  An
    #     NT command processor must be used to deal with this.
    # 3.  External executables can be native NT (which only know
    #     NT-style file names) or cygwin executables (which normally
    #     know both cygwin UNIX-style file names and NT file names,
    #     but not always; some do not know about drive names, for
    #     example).
    #     Cygwin executables for tex and latex may only know cygwin
    #     filenames. 
    # 4.  The BIBINPUTS and TEXINPUTS environment variables may be
    #     UNIX-style or MSWin-style depending on whether native NT or
    #     cygwin executables are used.  They are therefore parsed
    #     differently.  Here is the clash:
    #        a. If a user is running under an NT shell, is using a
    #           native NT installation of tex (e.g., fptex or miktex),
    #           but has the cygwin executables in the path, then perl
    #           detects the OS as cygwin, but the user needs NT
    #           behavior from latexmk.
    #        b. If a user is running under an UNIX shell in a cygwin
    #           environment, and is using the cygwin installation of
    #           tex, then perl detects the OS as cygwin, and the user
    #           needs UNIX behavior from latexmk.
    #     Latexmk has no way of detecting the difference.  The two
    #     situations may even arise for the same user on the same
    #     computer simply by changing the order of directories in the
    #     path environment variable


    ## Configuration parameters: We'll assume native NT executables.
    ## The user should override if they are not.

    # This may fail: perl converts MSWin temp directory name to cygwin
    # format. Names containing this string cannot be handled by native
    # NT executables.
    $tmpdir = $ENV{TMPDIR} || $ENV{TEMP} || '.';

    ## List of possibilities for the system-wide initialization file.  
    ## The first one found (if any) is used.
    ## We could stay with MSWin files here, since cygwin perl understands them
    ## @rc_system_files = ( 'C:/latexmk/LatexMk' );
    ## But they are deprecated in v. 1.7.  So use the UNIX version, prefixed
    ##   with a cygwin equivalent of the MSWin location
    @rc_system_files = 
     ( '/cygdrive/c/latexmk/LatexMk', 
       '/opt/local/share/latexmk/LatexMk', 
       '/usr/local/share/latexmk/LatexMk',
       '/usr/local/lib/latexmk/LatexMk' );

    $search_path_separator = ';';  # Separator of elements in search_path
    # This is tricky.  The search_path_separator depends on the kind
    # of executable: native NT v. cygwin.  
    # So the user will have to override this.

    # We will assume that files can be viewed by native NT programs.
    #  Then we must fix the start command/directive, so that the
    #  NT-native start command of a cmd.exe is used.
    # For a pdf-file, "start x.pdf" starts the pdf viewer associated with
    #   pdf files, so no program name is needed:
    $start_NT = "cmd /c start \"\"";
    $pdf_previewer = "$start_NT %O %S";
    $ps_previewer  = "$start_NT %O %S";
    $ps_previewer_landscape  = $ps_previewer;
    $dvi_previewer  = "$start_NT %O %S";
    $dvi_previewer_landscape = $dvi_previewer;
    # Viewer update methods: 
    #    0 => auto update: viewer watches file (e.g., gv)
    #    1 => manual update: user must do something: e.g., click on window.
    #         (e.g., ghostview, MSWIN previewers, acroread under UNIX)
    #    2 => send signal.  Number of signal in $dvi_update_signal,
    #                         $ps_update_signal, $pdf_update_signal
    #    3 => viewer can't update, because it locks the file and the file 
    #         cannot be updated.  (acroread under MSWIN)
    $dvi_update_method = 1;
    $ps_update_method = 1;
    $pdf_update_method = 3; # acroread locks the pdf file
    # Use NONE as flag that I am not implementing some commands:
    $lpr =
        'NONE $lpr variable is not configured to allow printing of ps files';
    $lpr_dvi =
        'NONE $lpr_dvi variable is not configured to allow printing of dvi files';
    $lpr_pdf =
        'NONE $lpr_pdf variable is not configured to allow printing of pdf files';
    # The $pscmd below holds a command to list running processes.  It
    # is used to find the process ID of the viewer looking at the
    # current output file.  The output of the command must include the
    # process number and the command line of the processes, since the
    # relevant process is identified by the name of file to be viewed.
    # Its use is not essential.
    # When the OS is detected as cygwin, there are two possibilities:
    #    a.  Latexmk was run from an NT prompt, but cygwin is in the
    #        path. Then the cygwin ps command will not see commands
    #        started from latexmk.  So we cannot use it.
    #    b.  Latexmk was started within a cygwin environment.  Then
    #        the ps command works as we need.
    # Only the user, not latemk knows which, so we default to not
    # using the ps command.  The user can override this in a
    # configuration file. 
    $pscmd = 
        'NONE $pscmd variable is not configured to detect running processes';
    $pid_position = -1;     # offset of PID in output of pscmd.  
                            # Negative means I cannot use ps
}
else {
    # Assume anything else is UNIX or clone

    ## Configuration parameters:


    ## Use first existing case for $tmpdir:
    $tmpdir = $ENV{TMPDIR} || '/tmp';

    ## List of possibilities for the system-wide initialization file.  
    ## The first one found (if any) is used.
    ## Normally on a UNIX it will be in a subdirectory of /opt/local/share or
    ## /usr/local/share, depending on the local conventions.
    ## /usr/local/lib/latexmk/LatexMk is put in the list for
    ## compatibility with older versions of latexmk.
    @rc_system_files = 
     ( '/opt/local/share/latexmk/LatexMk', 
       '/usr/local/share/latexmk/LatexMk',
       '/usr/local/lib/latexmk/LatexMk' );

    $search_path_separator = ':';  # Separator of elements in search_path

    $dvi_update_signal = $signo{USR1} 
         if ( defined $signo{USR1} ); # Suitable for xdvi
    $ps_update_signal = $signo{HUP} 
         if ( defined $signo{HUP} );  # Suitable for gv
    $pdf_update_signal = $signo{HUP} 
         if ( defined $signo{HUP} );  # Suitable for gv
    ## default document processing programs.
    # Viewer update methods: 
    #    0 => auto update: viewer watches file (e.g., gv)
    #    1 => manual update: user must do something: e.g., click on window.
    #         (e.g., ghostview, MSWIN previewers, acroread under UNIX)
    #    2 => send signal.  Number of signal in $dvi_update_signal,
    #                         $ps_update_signal, $pdf_update_signal
    #    3 => viewer can't update, because it locks the file and the file 
    #         cannot be updated.  (acroread under MSWIN)
    #    4 => Run command to update.  Command in $dvi_update_command, 
    #    $ps_update_command, $pdf_update_command.
    $dvi_previewer  = 'start xdvi %O %S';
    $dvi_previewer_landscape = 'start xdvi -paper usr %O %S';
    if ( defined $dvi_update_signal ) { 
        $dvi_update_method = 2;  # xdvi responds to signal to update
    } else {
        $dvi_update_method = 1;  
    }
#    if ( defined $ps_update_signal ) { 
#        $ps_update_method = 2;  # gv responds to signal to update
#        $ps_previewer  = 'start gv -nowatch';
#        $ps_previewer_landscape  = 'start gv -swap -nowatch';
#    } else {
#        $ps_update_method = 0;  # gv -watch watches the ps file
#        $ps_previewer  = 'start gv -watch';
#        $ps_previewer_landscape  = 'start gv -swap -watch';
#    }
    # Turn off the fancy options for gv.  Regular gv likes -watch etc
    #   GNU gv likes --watch etc.  User must configure
    $ps_update_method = 0;  # gv -watch watches the ps file
    $ps_previewer  = 'start gv %O %S';
    $ps_previewer_landscape  = 'start gv -swap %O %S';
    $pdf_previewer = 'start acroread %O %S';
    $pdf_update_method = 1;  # acroread under unix needs manual update
    $lpr = 'lpr %O %S';         # Assume lpr command prints postscript files correctly
    $lpr_dvi =
        'NONE $lpr_dvi variable is not configured to allow printing of dvi files';
    $lpr_pdf =
        'NONE $lpr_pdf variable is not configured to allow printing of pdf files';
    # The $pscmd below holds a command to list running processes.  It
    # is used to find the process ID of the viewer looking at the
    # current output file.  The output of the command must include the
    # process number and the command line of the processes, since the
    # relevant process is identified by the name of file to be viewed.
    # Uses:
    #   1.  In preview_continuous mode, to save running a previewer
    #       when one is already running on the relevant file.
    #   2.  With xdvi in preview_continuous mode, xdvi must be
    #       signalled to make it read a new dvi file.
    #
    # The following works on Solaris, LINUX, HP-UX, IRIX
    # Use -f to get full listing, including command line arguments.
    # Use -u $ENV{CMD} to get all processes started by current user (not just
    #   those associated with current terminal), but none of other users' 
    #   processes. 
    $pscmd = "ps -f -u $ENV{USER}"; 
    $pid_position = 1; # offset of PID in output of pscmd; first item is 0.  
    if ( $^O eq "linux" ) {
        # Ps on Redhat (at least v. 7.2) appears to truncate its output
        #    at 80 cols, so that a long command string is truncated.
        # Fix this with the --width option.  This option works under 
        #    other versions of linux even if not necessary (at least 
        #    for SUSE 7.2). 
        # However the option is not available under other UNIX-type 
        #    systems, e.g., Solaris 8.
        $pscmd = "ps --width 200 -f -u $ENV{USER}"; 
    }
    elsif ( $^O eq "darwin" ) {
        # OS-X on Macintosh
        $lpr_pdf  = 'lpr %O %S';  
        $pscmd = "ps -ww -u $ENV{USER}"; 
    }
}

## default parameters
$max_repeat = 5;        # Maximum times I repeat latex.  Normally
                        # 3 would be sufficient: 1st run generates aux file,
                        # 2nd run picks up aux file, and maybe toc, lof which 
                        # contain out-of-date information, e.g., wrong page
                        # references in toc, lof and index, and unresolved
                        # references in the middle of lines.  But the 
                        # formatting is more-or-less correct.  On the 3rd
                        # run, the page refs etc in toc, lof, etc are about
                        # correct, but some slight formatting changes may
                        # occur, which mess up page numbers in the toc and lof,
                        # Hence a 4th run is conceivably necessary. 
                        # At least one document class (JHEP.cls) works
			# in such a way that a 4th run is needed.  
                        # We allow an extra run for safety for a
			# maximum of 5. Needing further runs is
			# usually an indication of a problem; further
			# runs may not resolve the problem, and
			# instead could cause an infinite loop.
$clean_ext = "";        # space separated extensions of files that are
                        # to be deleted when doing cleanup, beyond
                        # standard set
$clean_full_ext = "";   # space separated extensions of files that are
                        # to be deleted when doing cleanup_full, beyond
                        # standard set and those in $clean_ext
@cus_dep_list = ();     # Custom dependency list
@default_files = ( '*.tex' );   # Array of LaTeX files to process when 
                        # no files are specified on the command line.
                        # Wildcards allowed
                        # Best used for project specific files.
@default_excluded_files = ( );   
                        # Array of LaTeX files to exclude when using
                        # @default_files, i.e., when no files are specified
                        # on the command line.
                        # Wildcards allowed
                        # Best used for project specific files.
$texfile_search = "";   # Specification for extra files to search for
                        # when no files are specified on the command line
                        # and the @default_files variable is empty.
                        # Space separated, and wildcards allowed.
                        # These files are IN ADDITION to *.tex in current 
                        # directory. 
                        # This variable is obsolete, and only in here for
                        # backward compatibility.

$fdb_ext = 'fdb_latexmk'; # Extension for the file for latexmk's
			  # file-database
                          # Make it long to avoid possible collisions.
$fdb_ver = 2;             # Version number for kind of fdb_file.

$jobname = '';          # Jobname: as with current tex, etc indicates
                        # basename of generated files.
                        # Defined so that --jobname=STRING on latexmk's
                        # command line has same effect as with current
                        # tex, etc.  (If $jobname is non-empty, then
                        # the --jobname=... option is used on tex.)


## default flag settings.
$silent = 0;            # silence latex's messages?
$landscape_mode = 0;	# default to portrait mode

# The following two arrays contain lists of extensions (without
# period) for files that are read in during a (pdf)LaTeX run but that
# are generated automatically from the previous run, as opposed to
# being user generated files (directly or indirectly from a custom
# dependency).  These files get two kinds of special treatment:
#     1.  In clean up, where depending on the kind of clean up, some
#         or all of these generated files are deleted.
#         (Note that special treatment is given to aux files.)
#     2.  In analyzing the results of a run of (pdf)LaTeX, to
#         determine if another run is needed.  With an error free run,
#         a rerun should be provoked by a change in any source file,
#         whether a user file or a generated file.  But with a run
#         that ends in an error, only a change in a user file during
#         the run (which might correct the error) should provoke a
#         rerun, but a change in a generated file should not.
# These arrays can be user-configured.

@generated_exts = ( 'aux', 'idx', 'ind', 'lof', 'lot', 'out', 'toc' );
     # N.B. 'out' is generated by hyperref package

# Which kinds of file do I have requests to make?
# If no requests at all are made, then I will make dvi file
# If particular requests are made then other files may also have to be
# made.  E.g., ps file requires a dvi file
$dvi_mode = 0;          # No dvi file requested
$postscript_mode = 0;           # No postscript file requested
$pdf_mode = 0;          # No pdf file requested to be made by pdflatex
                        # Possible values: 
                        #     0 don't create pdf file
                        #     1 to create pdf file by pdflatex
                        #     2 to create pdf file by ps2pdf
                        #     3 to create pdf file by dvipdf
$view = 'default';      # Default preview is of highest of dvi, ps, pdf
$sleep_time = 2;	# time to sleep b/w checks for file changes in -pvc mode
$banner = 0;            # Non-zero if we have a banner to insert
$banner_scale = 220;    # Original default scale
$banner_intensity = 0.95;  # Darkness of the banner message
$banner_message = 'DRAFT'; # Original default message
$do_cd = 0;     # Do not do cd to directory of source file.
                #   Thus behave like latex.
$dependents_list = 0;   # Whether to display list(s) of dependencies
@dir_stack = (); # Stack of pushed directories.
$cleanup_mode = 0;      # No cleanup of nonessential LaTex-related files.
                        # $cleanup_mode = 0: no cleanup
                        # $cleanup_mode = 1: full cleanup 
                        # $cleanup_mode = 2: cleanup except for dvi,
                        #                    dviF, pdf, ps, & psF 
$cleanup_fdb  = 0;      # No removal of file for latexmk's file-database
$cleanup_only = 0;      # When doing cleanup, do not go-on to making files
$cleanup_includes_generated = 0; 
                        # Determines whether cleanup deletes files generated by
                        #    (pdf)latex (found from \openout lines in log file).
$diagnostics = 0;
$dvi_filter = '';	# DVI filter command
$ps_filter = '';	# Postscript filter command

$force_mode = 0;        # =1 to force processing past errors
$go_mode = 0;           # =1 to force processing regardless of time-stamps
                        # =2 full clean-up first
$preview_mode = 0;
$preview_continuous_mode  = 0;
$printout_mode = 0;     # Don't print the file

# Do we make view file in temporary then move to final destination?
#  (To avoid premature updating by viewer).
$always_view_file_via_temporary = 0;      # Set to 1 if  viewed file is always
                                   #    made through a temporary.
$pvc_view_file_via_temporary = 1;  # Set to 1 if only in -pvc mode is viewed 
                                   #    file made through a temporary.

# State variables initialized here:

$updated = 0;           # Flags when something has been remade
                        # Used to allow convenient user message in -pvc mode
$waiting = 0;           # Flags whether we are in loop waiting for an event
                        # Used to avoid unnecessary repeated o/p in wait loop

# Used for some results of parsing log file:
$reference_changed = 0;
$bad_reference = 0;
$bad_citation = 0;


# Set search paths for includes.
# Set them early so that they can be overridden
$BIBINPUTS = $ENV{'BIBINPUTS'};
if (!$BIBINPUTS) { $BIBINPUTS = '.'; }
#?? OBSOLETE
$TEXINPUTS = $ENV{'TEXINPUTS'};
if (!$TEXINPUTS) { $TEXINPUTS = '.'; }

# Convert search paths to arrays:
# If any of the paths end in '//' then recursively search the
# directory.  After these operations, @BIBINPUTS  should
# have all the directories that need to be searched

@BIBINPUTS = find_dirs1 ($BIBINPUTS);


######################################################################
######################################################################
#
#  ???  UPDATE THE FOLLOWING!!
#
# We will need to determine whether source files for runs of various
# programs are out of date.  In a normal situation, this is done by
# asking whether the times of the source files are later than the
# destination files.  But this won't work for us, since a common
# situation is that a file is written on one run of latex, for
# example, and read back in on the next run (e.g., an .aux file).
# Some situations of this kind are standard in latex generally; others
# occur with particular macro packages or with particular
# postprocessors. 
#
# The correct criterion for whether a source is out-of-date is
# therefore NOT that its modification time is later than the
# destination file, but whether the contents of the source file have
# changed since the last successful run.  This also handles the case
# that the user undoes some changes to a source file by replacing the
# source file by reverting to an earlier version, which may well have
# an older time stamp.  Since a direct comparison of old and new files
# would involve storage and access of a large number of backup files,
# we instead use the md5 signature of the files.  (Previous versions
# of latexmk used the backup file method, but restricted to the case
# of .aux and .idx files, sufficient for most, but not all,
# situations.)
#
# We will have a database of (time, size, md5) for the relevant
# files. If the time and size of a file haven't changed, then the file
# is assumed not to have changed; this saves us from having to
# determine its md5 signature, which would involve reading the whole 
# file, which is naturally time-consuming, especially if network file
# access to a server is needed, and many files are involved, when most
# of them don't change.  It is of course possible to change a file
# without changing its size, but then to adjust its timestamp 
# to what it was previously; this requires a certain amount of
# perversity.  We can safely assume that if the user edits a file or
# changes its contents, then the file's timestamp changes.  The
# interesting case is that the timestamp does change, because the file
# has actually been written to, but that the contents do not change;
# it is for this that we use the md5 signature.  However, since
# computing the md5 signature involves reading the whole file, which
# may be large, we should avoid computing it more than necessary. 
#
# So we get the following structure:
#
#     1.  For each relevant run (latex, pdflatex, each instance of a
#         custom dependency) we have a database of the state of the
#         source files that were last used by the run.
#     2.  On an initial startup, the database for a primary tex file
#         is read that was created by a previous run of latex or
#         pdflatex, if this exists.  
#     3.  If the file doesn't exist, then the criterion for
#         out-of-dateness for an initial run is that it goes by file
#         timestamps, as in previous versions of latexmk, with due
#         (dis)regard to those files that are known to be generated by
#         latex and re-read on the next run.
#     4.  Immediately before a run, the database is updated to
#         represent the current conditions of the run's source files.
#     5.  After the run, it is determined whether any of the source
#         files have changed.  This covers both files written by the
#         run, which are therefore in a dependency loop, and files that
#         the user may have updated during the run.  (The last often
#         happens when latex takes a long time, for a big document,
#         and the user makes edits before latex has finished.  This is
#         particularly prevalent when latexmk is used with
#         preview-continuous mode.)
#     6.  In the case of latex or pdflatex, the custom dependencies
#         must also be checked and redone if out-of-date.
#     7.  If any source files have changed, the run is redone,
#         starting at step 1.
#     8.  There is naturally a limit on the number of reruns, to avoid
#         infinite loops from bugs and from pathological or unforeseen
#         conditions. 
#     9.  After the run is done, the run's file database is updated.
#         (By hypothesis, the sizes and md5s are correct, if the run
#         is successful.)
#    10.  To allow reuse of data from previous runs, the file database
#         is written to a file after every complete set of passes
#         through latex or pdflatex.  (Note that there is separate
#         information for latex and pdflatex; the necessary
#         information won't coincide: Out-of-dateness for the files
#         for each program concerns the properties of the files when
#         the other program was run, and the set of source files could
#         be different, e.g., for graphics files.)  
#
# We therefore maintain the following data structures.:
#
#     a.  For each run (latex, pdflatex, each custom dependency) a
#         database is maintained.  This is a hash from filenames to a
#         reference to an array:  [time, size, md5].  The semantics of
#         the database is that it represents the state of the source
#         files used in the run.  During a run it represents the state
#         immediately before the run; after a run, with all reruns, it
#         represents the state of the files used, modified by having
#         the latest timestamps for generated files.
#     b.  There is a global database for all files, which represents
#         the current state.  This saves having to recompute the md5
#         signatures of a changed file used in more than one run
#         (e.g., latex and pdflatex).
#     c.  Each of latex and pdflatex has a list of the relevant custom
#         dependencies. 
#
# In all the following a fdb-hash is a hash of the form:
#                      filename -> [time, size, md5] 
# If a file is found to disappear, its entry is removed from the hash.
# In returns from fdb access routines, a size entry of -1 indicates a
# non-existent file.


# List of known rules.  Rule types: primary, 
#     external (calls program), internal (calls routine), cusdep.

%possible_primaries = ( 'latex'  => 'primary',  'pdflatex'  => 'primary' );
%primaries = ();    # Hash of rules for primary part of make.  Keys are 
                    # currently 'latex', 'pdflatex' or both.  Value is
                    # currently irrelevant.  Use hash for ease of lookup
   # Make remove this later, if use makeB

# Hashes, whose keys give names of particular kinds of rule.  We use
# hashes for ease of lookup.
%possible_one_time = ( 'view' => 1, 'print' => 1, 'update_view' => 1,  );
%requested_filerules = ();  # Hash for rules corresponding to requested files.  
                    # The keys are the rulenames and the value is 
                    # currently irrelevant.
%one_time = ();     # Hash for requested one-time-only rules, currently
                    # possible values 'print' and 'view'.  


%rule_db = ();      # Database of all rules:
                    # Hash: rulename -> [array of rule data]
                    # Rule data:
                    #   0: [ cmd_type, ext_cmd, int_cmd, test_kind, 
                    #       source, dest, base, out_of_date,
		    #       out_of_date_user, time_of_last_run,
                    #       last_result, last_message
                    #      ]
                    # where 
                    #     cmd_type is 'primary', 'external', or 'cusdep'
                    #     ext_cmd is string for associated external command
                    #       with substitutions (%D for destination, %S
		    #       for source, %B for base of current rule,
		    #       %R for base of primary tex file, %T for
		    #       texfile name, and %O for options.
                    #     int_cmd specifies any internal command to be
		    #       used to implement the application of the
		    #       rule.  If this is present, it overrides
		    #       the external command, and it is the
		    #       responsibility of the perl subroutine
		    #       specified in intcmd to execute the
		    #       external command if this is appropriate.
		    #       This variable intcmd is a reference to an array,  
                    #       $$intcmd[0] = internal routine
                    #       $$intcmd[1...] = its arguments (if any)
                    #     test_kind specifies method of determining
                    #       whether a file is out-of-date:
                    #         0 for never
                    #         1 for usual: whether there is a source
		    #              file change 
                    #         2 for dest earlier than source
                    #         3 for method 2 at first run, 1 thereafter
                    #              (used when don't have file data from
		    #              previous run).
                    #     source = name of primary source file, if any
                    #     dest   = name of primary destination file,
		    #              if any
                    #     base   = base name, if any, of files for
		    #              this rule
                    #     out_of_date = 1 if it has been detected that
		    #                     this rule needs to be run
		    #                     (typically because a source
		    #                     file has changed).
                    #                   0 otherwise
                    #     out_of_date_user is like out_of_date, except
                    #         that the detection of out-of-dateness
                    #         has been made from a change of a
		    #         putative user file, i.e., one that is
		    #         not a generated file (e.g., aux). This
		    #         kind of out-of-dateness should provoke a
		    #         rerun where or not there was an error
		    #         during a run of (pdf)LaTeX.  Normally,
		    #         if there is an error, one should wait
		    #         for the user to correct the error.  But
		    #         it is possible the error condition is
		    #         already corrected during the run, e.g.,
		    #         by the user changing a source file in
		    #         response to an error message. 
                    #     time_of_last_run = time that this rule was
		    #              last applied.  (In standard units
		    #              from perl, to be directly compared
		    #              with file modification times.)
                    #     changed flags whether special changes have been made
                    #          that require file-existence status to be ignored
                    #     last_result is 
                    #                 -1 if no run has been made,
                    #                  0 if the last run was successful
                    #                  1 if last run was successful, but
                    #                    failed to create an output file
                    #                  2 if last run failed
                    #     last_message is error message for last run
                    #   1: {Hash sourcefile -> [source-file data] }
                    # Source-file data array: 
                    #   0: time
                    #   1: size
                    #   2: md5
                    #   3: name of rule to make this file
                    #   4: whether the file is of the kind made by epstopdf.sty 
                    #      during a primary run.  It will have been read during
                    #      the run, so that even though the file changes during
                    #      a primary run, there is no need to trigger another 
                    #      run because of this.
                    #  Size and md5 correspond to the values at the last run.
                    #  But time may be updated to correspond to the time
                    #  for the file, if the file is otherwise unchanged.
                    #  This saves excessive md5 calculations, which would
                    #  otherwise be done everytime the file is checked, 
                    #  in the following situation:
                    #     When the file has been rewritten after a run
                    #     has started (commonly aux, bbl files etc),
                    #     but the actual file contents haven't
                    #     changed.  Then because the filetime has
                    #     changed, on every file-change check latexmk
                    #     would normally redo the md5 calculation to
                    #     test for actual changes.  Once one such
                    #     check is done, and the contents are
                    #     unchanged, later checks are superfluous, and
                    #     can be avoided by changing the file's time
                    #     in the source-file list.

%fdb_current = ();  # Fdb-hash for all files used.


#==================================================
## Read rc files:

# User's home directory
$HOME = '';
if (exists $ENV{'HOME'} ) {
    $HOME = $ENV{'HOME'};
}
elsif (exists $ENV{'USERPROFILE'} ) {
    $HOME = $ENV{'USERPROFILE'};
}



sub read_first_rc_file_in_list {
    foreach my $rc_file ( @_ ) {
        #print "===Testing for rc file \"$rc_file\" ...\n";
        if ( -e $rc_file ) {
            #print "===Reading rc file \"$rc_file\" ...\n";
            process_rc_file( $rc_file );
            return;
        }
    }
}

# Read system rc file:
read_first_rc_file_in_list( @rc_system_files );
# Read user rc file.
read_first_rc_file_in_list( "$HOME/.latexmkrc" );
# Read rc file in current directory.
read_first_rc_file_in_list( "latexmkrc", ".latexmkrc" );

#==================================================

if ($bibtex_use > 1) {
    push @generated_exts, 'bbl';
}

#show_array ("BIBINPUTS", @BIBINPUTS); die;

## Process command line args.
@command_line_file_list = ();
$bad_options = 0;

#print "Command line arguments:\n"; for ($i = 0; $i <= $#ARGV; $i++ ) {  print "$i: '$ARGV[$i]'\n"; }

while ($_ = $ARGV[0])
{
  # Make -- and - equivalent at beginning of option:
  s/^--/-/;
  shift;
  if (/^-bibtex$/) { $bibtex_use = 2; }
  elsif (/^-bibtex-$/) { $bibtex_use = 0; }
  elsif (/^-nobibtex$/) { $bibtex_use = 0; }
  elsif (/^-bibtex-cond$/) { $bibtex_use = 1; }
  elsif (/^-c$/)        { $cleanup_mode = 2; $cleanup_fdb = 1; $cleanup_only = 1; }
  elsif (/^-C$/ || /^-CA$/ ) { $cleanup_mode = 1; $cleanup_fdb = 1; $cleanup_only = 1; }
  elsif (/^-CF$/)    { $cleanup_fdb = 1; }
  elsif (/^-cd$/)    { $do_cd = 1; }
  elsif (/^-cd-$/)   { $do_cd = 0; }
  elsif (/^-commands$/) { &print_commands; exit; }
  elsif (/^-d$/)     { $banner = 1; }
  elsif (/^-dependents$/) { $dependents_list = 1; }
  elsif (/^-nodependents$/ || /^-dependents-$/) { $dependents_list = 0; }
  elsif (/^-dvi$/)   { $dvi_mode = 1; }
  elsif (/^-dvi-$/)  { $dvi_mode = 0; }
  elsif (/^-f$/)     { $force_mode = 1; }
  elsif (/^-f-$/)    { $force_mode = 0; }
  elsif (/^-g$/)     { $go_mode = 1; }
  elsif (/^-g-$/)    { $go_mode = 0; }
  elsif (/^-gg$/)    { 
     $go_mode = 2; $cleanup_mode = 1; $cleanup_fdb = 1; $cleanup_only = 0; 
  }
  elsif ( /^-h$/ || /^-help$/ )   { &print_help; exit;}
  elsif (/^-diagnostics/) { $diagnostics = 1; }
  elsif (/^-jobname=(.*)$/) {
      $jobname = $1;
  }
  elsif (/^-l$/)     { $landscape_mode = 1; }
  elsif (/^-l-$/)    { $landscape_mode = 0; }
  elsif (/^-new-viewer$/) {
                       $new_viewer_always = 1; 
  }
  elsif (/^-new-viewer-$/) {
                       $new_viewer_always = 0; 
  }
  elsif (/^-p$/)     { $printout_mode = 1; 
                       $preview_continuous_mode = 0; # to avoid conflicts
                       $preview_mode = 0;  
                     }
  elsif (/^-p-$/)    { $printout_mode = 0; }
  elsif (/^-pdfdvi$/){ $pdf_mode = 3; }
  elsif (/^-pdfps$/) { $pdf_mode = 2; }
  elsif (/^-pdf$/)   { $pdf_mode = 1; }
  elsif (/^-pdf-$/)  { $pdf_mode = 0; }
  elsif (/^-print=(.*)$/) {
      $value = $1;
      if ( $value =~ /^dvi$|^ps$|^pdf$/ ) {
          $print_type = $value;
          $printout_mode = 1;
      }
      else {
          &exit_help("$My_name: unknown print type '$value' in option '$_'");
      }
  }
  elsif (/^-ps$/)    { $postscript_mode = 1; }
  elsif (/^-ps-$/)   { $postscript_mode = 0; }
  elsif (/^-pv$/)    { $preview_mode = 1; 
                       $preview_continuous_mode = 0; # to avoid conflicts
                       $printout_mode = 0; 
                     }
  elsif (/^-pv-$/)   { $preview_mode = 0; }
  elsif (/^-pvc$/)   { $preview_continuous_mode = 1;
                       $force_mode = 0;    # So that errors do not cause loops
                       $preview_mode = 0;  # to avoid conflicts
                       $printout_mode = 0; 
                     }
  elsif (/^-pvc-$/)  { $preview_continuous_mode = 0; }
  elsif (/^-silent$/ || /^-quiet$/ ){ $silent = 1; }
  elsif (/^-v$/ || /^-version$/)   { 
      print "\n$version_details. Version $version_num\n"; 
      exit; 
  }
  elsif (/^-verbose$/)  { $silent = 0; }
  elsif (/^-view=default$/) { $view = "default";}
  elsif (/^-view=dvi$/)     { $view = "dvi";}
  elsif (/^-view=none$/)    { $view = "none";}
  elsif (/^-view=ps$/)      { $view = "ps";}
  elsif (/^-view=pdf$/)     { $view = "pdf"; }
  elsif (/^-e$/) {  
     if ( $#ARGV < 0 ) {
        &exit_help( "No code to execute specified after -e switch"); 
     }
     execute_code_string( $ARGV[0] );
     shift;
  }
  elsif (/^-r$/) {  
     if ( $ARGV[0] eq '' ) {
        &exit_help( "No RC file specified after -r switch"); 
     }
     if ( -e $ARGV[0] ) {
	process_rc_file( $ARGV[0] );
     } 
     else {
	$! = 11;
	die "$My_name: RC file [$ARGV[0]] does not exist\n"; 
     }
     shift; 
  }
  elsif (/^-bm$/) {
     if ( $ARGV[0] eq '' ) {
	&exit_help( "No message specified after -bm switch");
     }
     $banner = 1; $banner_message = $ARGV[0];
     shift; 
  }
  elsif (/^-bi$/) {
     if ( $ARGV[0] eq '' ) {
	&exit_help( "No intensity specified after -bi switch");
     }
     $banner_intensity = $ARGV[0];
     shift; 
  }
  elsif (/^-bs$/) {
     if ( $ARGV[0] eq '' ) {
	&exit_help( "No scale specified after -bs switch");
     }
     $banner_scale = $ARGV[0];
     shift; 
  }
  elsif (/^-dF$/) {
     if ( $ARGV[0] eq '' ) {
	&exit_help( "No dvi filter specified after -dF switch");
     }
     $dvi_filter = $ARGV[0];
     shift; 
  }
  elsif (/^-pF$/) {
     if ( $ARGV[0] eq '' ) {
        &exit_help( "No ps filter specified after -pF switch");
     }
     $ps_filter = $ARGV[0];
     shift; 
  }
  elsif (/^-/) {
     warn "$My_name: $_ bad option\n"; 
     $bad_options++;
  }
  else {
     push @command_line_file_list, $_ ; 
  }
}

if ( $bad_options > 0 ) {
    &exit_help( "Bad options specified" );
}

warn "$My_name: This is $version_details, version: $version_num.\n",
     "**** Report bugs etc to John Collins <collins at phys.psu.edu>. ****\n"
   unless $silent;

# For backward compatibility, convert $texfile_search to @default_files
# Since $texfile_search is initialized to "", a nonzero value indicates
# that an initialization file has set it.
if ( $texfile_search ne "" ) {
    @default_files = split / /, "*.tex $texfile_search";
}

#printA "A: Command line file list:\n";
#for ($i = 0; $i <= $#command_line_file_list; $i++ ) {  print "$i: '$command_line_file_list[$i]'\n"; }

#Glob the filenames command line if the script was not invoked under a 
#   UNIX-like environment.
#   Cases: (1) MS/MSwin native    Glob
#                      (OS detected as MSWin32)
#          (2) MS/MSwin cygwin    Glob [because we do not know whether
#                  the cmd interpreter is UNIXy (and does glob) or is
#                  native MS-Win (and does not glob).]
#                      (OS detected as cygwin)
#          (3) UNIX               Don't glob (cmd interpreter does it)
#                      (Currently, I assume this is everything else)
if ( ($^O eq "MSWin32") || ($^O eq "cygwin") ) {
    # Preserve ordering of files
    @file_list = glob_list1(@command_line_file_list);
#print "A1:File list:\n";
#for ($i = 0; $i <= $#file_list; $i++ ) {  print "$i: '$file_list[$i]'\n"; }
}
else {
    @file_list = @command_line_file_list;
#print "A2:File list:\n";
#for ($i = 0; $i <= $#file_list; $i++ ) {  print "$i: '$file_list[$i]'\n"; }
}
@file_list = uniq1( @file_list );


# Check we haven't selected mutually exclusive modes.
# Note that -c overides all other options, but doesn't cause
# an error if they are selected.
if (($printout_mode && ( $preview_mode || $preview_continuous_mode ))
    || ( $preview_mode && $preview_continuous_mode ))
{
  # Each of the options -p, -pv, -pvc turns the other off.
  # So the only reason to arrive here is an incorrect inititalization
  #   file, or a bug.
  &exit_help( "Conflicting options (print, preview, preview_continuous) selected");
}

if ( @command_line_file_list ) {   
    # At least one file specified on command line (before possible globbing).
    if ( !@file_list ) {
        &exit_help( "Wildcards in file names didn't match any files");
    }
}
else {
    # No files specified on command line, try and find some
    # Evaluate in order specified.  The user may have some special
    #   for wanting processing in a particular order, especially
    #   if there are no wild cards.
    # Preserve ordering of files
    my @file_list1 = uniq1( glob_list1(@default_files) );
    my @excluded_file_list = uniq1( glob_list1(@default_excluded_files) );
    # Make hash of excluded files, for easy checking:
    my %excl = ();
    foreach my $file (@excluded_file_list) {
	$excl{$file} = '';
    }
    foreach my $file (@file_list1) {
	push( @file_list, $file)  unless ( exists $excl{$file} );
    }    
    if ( !@file_list ) {
	&exit_help( "No file name specified, and I couldn't find any");
    }
}

$num_files = $#file_list + 1;
$num_specified = $#command_line_file_list + 1;

#print "Command line file list:\n";
#for ($i = 0; $i <= $#command_line_file_list; $i++ ) {  print "$i: '$command_line_file_list[$i]'\n"; }
#print "File list:\n";
#for ($i = 0; $i <= $#file_list; $i++ ) {  print "$i: '$file_list[$i]'\n"; }


# If selected a preview-continuous mode, make sure exactly one filename was specified
if ($preview_continuous_mode && ($num_files != 1) ) {
    if ($num_specified > 1) {
        &exit_help( 
          "Need to specify exactly one filename for ".
              "preview-continuous mode\n".
          "    but $num_specified were specified"
        );
    }
    elsif ($num_specified == 1) {
        &exit_help( 
          "Need to specify exactly one filename for ".
              "preview-continuous mode\n".
          "    but wildcarding produced $num_files files"
        );
    }
    else {
        &exit_help( 
          "Need to specify exactly one filename for ".
              "preview-continuous mode.\n".
          "    Since none were specified on the command line, I looked for \n".
          "    files in '@default_files'.\n".
          "    But I found $num_files files, not 1."
        );
    }
}

# If selected jobname, can only apply that to one file:
if ( ($jobname ne '') && ($num_files > 1) ) {
    &exit_help( 
          "Need to specify at most one filename if ".
          "jobname specified, \n".
          "    but $num_files were found (after defaults and wildcarding)."
        );
}


# Normalize the commands, to have place-holders for source, dest etc:
&fix_cmds;

# If landscape mode, change dvips processor, and the previewers:
if ( $landscape_mode )
{
  $dvips = $dvips_landscape;
  $dvi_previewer = $dvi_previewer_landscape;
  $ps_previewer = $ps_previewer_landscape;
}

if ( $silent ) { 
    add_option( \$latex, " $latex_silent_switch" ); 
    add_option( \$pdflatex, " $pdflatex_silent_switch" ); 
    add_option( \$bibtex, " $bibtex_silent_switch" ); 
    add_option( \$makeindex, " $makeindex_silent_switch" ); 
    add_option( \$dvips, " $dvips_silent_switch" ); 
}

if ( $jobname ne '' ) { 
    $jobstring = "--jobname=$jobname";
    add_option( \$latex, " $jobstring" ); 
    add_option( \$pdflatex, " $jobstring" ); 
}

# Which kind of file do we preview?
if ( $view eq "default" ) {
    # If default viewer requested, use "highest" of dvi, ps and pdf
    #    that was requested by user.  
    # No explicit request means view dvi.
    $view = "dvi";
    if ( $postscript_mode ) { $view = "ps"; }
    if ( $pdf_mode ) { $view = "pdf"; }
}

# Make sure we make the kind of file we want to view:
if ($view eq 'dvi') { $dvi_mode = 1; }
if ($view eq 'ps') { $postscript_mode = 1; }
if ( ($view eq 'pdf') && ($pdf_mode == 0) ) { 
    $pdf_mode = 1; 
}

# Make sure that we make something if all requests are turned off
if ( ! ( $dvi_mode || $pdf_mode || $postscript_mode || $printout_mode) ) {
    print "No specific requests made, so default to dvi by latex\n";
    $dvi_mode = 1;
}

# Set new-style requested rules:
if ( $dvi_mode ) { $requested_filerules{'latex'} = 1; }
if ( $pdf_mode == 1 ) { $requested_filerules{'pdflatex'} = 1; }
elsif ( $pdf_mode == 2 ) { 
   $requested_filerules{'latex'} = 1;
   $requested_filerules{'dvips'} = 1;
   $requested_filerules{'ps2pdf'} = 1; 
}
elsif ( $pdf_mode == 3 ) { 
   $requested_filerules{'latex'} = 1;
   $requested_filerules{'dvipdf'} = 1; 
}
if ( $postscript_mode ) { 
   $requested_filerules{'latex'} = 1; 
   $requested_filerules{'dvips'} = 1; 
}
if ( $printout_mode ) { $one_time{'print'} = 1; }
if ( $preview_continuous_mode || $preview_mode ) { $one_time{'view'} = 1; }
if ( length($dvi_filter) != 0 ) { $requested_filerules{'dvi_filter'} = 1; }
if ( length($ps_filter) != 0 )  { $requested_filerules{'ps_filter'} = 1; }
if ( $banner ) { $requested_filerules{'dvips'} = 1; }


if ( $pdf_mode == 2 ) {
    # We generate pdf from ps.  Make sure we have the correct kind of ps.
    add_option( \$dvips, " $dvips_pdf_switch" );
}

# Note sleep has granularity of 1 second.
# Sleep periods 0 < $sleep_time < 1 give zero delay,
#    which is probably not what the user intended.
# Sleep periods less than zero give infinite delay
if ( $sleep_time < 0 ) {
     warn "$My_name: Correcting negative sleep_time to 1 sec.\n";
     $sleep_time = 1;
}
elsif ( ($sleep_time < 1) && ( $sleep_time != 0 ) ) {
     warn "$My_name: Correcting nonzero sleep_time of less than 1 sec to 1 sec.\n";
     $sleep_time = 1;
}
elsif ( $sleep_time == 0 ) {
     warn "$My_name: sleep_time was configured to zero.\n",
    "    Do you really want to do this?  It will give 100% CPU usage.\n";
}

# Make convenient forms for lookup.
# Extensions always have period.

# Convert @generated_exts to a hash for ease of look up and deletion
# Keep extension without period!
%generated_exts_all = ();
foreach (@generated_exts ) {
    $generated_exts_all{$_} = 1;
}

$quell_uptodate_msgs = $silent; 
   # Whether to quell informational messages when files are uptodate
   # Will turn off in -pvc mode

# Process for each file.
# The value of $bibtex_mode set in an initialization file may get
# overridden, during file processing, so save it:
#?? Unneeded now: $save_bibtex_mode = $bibtex_mode;

$failure_count = 0;
@failed_primaries = ();

FILE:
foreach $filename ( @file_list )
{
    # Global variables for making of current file:
    $updated = 0;
    $failure = 0;        # Set nonzero to indicate failure at some point of 
                         # a make.  Use value as exit code if I exit.
    $failure_msg = '';   # Indicate reason for failure

    if ( $do_cd ) {
       ($filename, $path) = fileparse( $filename );
       warn "$My_name: Changing directory to '$path'\n";
       pushd( $path );
    }
    else {
	$path = '';
    }


    ## remove extension from filename if was given.
    if ( &find_basename($filename, $root_filename, $texfile_name) )
    {
	if ( $force_mode ) {
	   warn "$My_name: Could not find file [$texfile_name]\n";
	}
	else {
            &ifcd_popd;
	    &exit_msg1( "Could not find file [$texfile_name]",
			11);
	}
    }
    if ($jobname ne '' ) {
        $root_filename = $jobname;
    }

    # Initialize basic dependency information:

    # For use under error conditions:
    @default_includes = ($texfile_name, "$root_filename.aux");  

    $fdb_file = "$root_filename.$fdb_ext";
    
    if ( $cleanup_mode > 0 ) {
        # Use parse_logB to get names of generated files.
        # It returns its results in the following variables:
        local %dependents = ();    # Maps files to status
        local @bbl_files = ();
        local %idx_files = ();     # Maps idx_file to (ind_file, base)
        local %generated_log = (); # Lists generated files
        local $primary_out = '';   # Actual output file (dvi or pdf)
        local %conversions = ();   # (pdf)latex-performed conversions.
                     # Maps output file created and read by (pdf)latex
                     #    to source file of conversion.
        print "$My_name: Examining log file for generated files...\n";
        &parse_logB;

        my @index_bibtex_generated = ();
        my @aux_files = ();
        my %other_generated = %generated_log;

        rdb_read_generatedB( $fdb_file, \@index_bibtex_generated, \@aux_files, 
                             \%other_generated );
        if ( ($go_mode == 2) && !$silent ) {
            warn "$My_name: Removing all generated files\n" unless $silent;
	}
        if ($bibtex_use < 2) { 
           delete $generated_exts_all{'bbl'}; 
        }
        my %index_bibtex_generated = ();
        my %aux_files = ();
        foreach (@index_bibtex_generated) {
	    $index_bibtex_generated{$_} = 1
	     unless ( /\.bbl$/ && ($bibtex_use < 2) );
            delete( $other_generated{$_} );
	}
        foreach (@aux_files) {
            $aux_files{$_} = 1;
            delete( $other_generated{$_} );
	}
        if ($diagnostics) {
            show_array( "For deletion:\n"
                       ." Generated (from makeindex and bibtex):", 
                        keys %index_bibtex_generated );
            show_array( " Aux files:", keys %aux_files );
            show_array( "Other generated files:\n"
                       ." (only deleted if \$cleanup_includes_generated is set): ",
                        keys %other_generated );
	}
        # Add to the generated files, some log file and some backup
        #    files used in previous versions of latexmk
        &cleanup1( $fdb_ext, 'blg', 'ilg', 'log', 'aux.bak', 'idx.bak', 
                   split(' ',$clean_ext), 
                   keys %generated_exts_all 
                 );
        unlink( 'texput.log', "texput.aux", 
                keys %index_bibtex_generated, 
                keys %aux_files );
        if ($cleanup_includes_generated) {
	    unlink( keys %other_generated );
	}
        if ( $cleanup_mode == 1 ) { 
            &cleanup1( 'dvi', 'dviF', 'ps', 'psF', 'pdf',
                       split(' ', $clean_full_ext)
                     );
        }
    }
    if ($cleanup_fdb) { unlink $fdb_file; }
    if ($cleanup_only) { next FILE; }

    # Initialize file and rule databases.
    %rule_list = ();
    &rdb_make_rule_list;
    &rdb_set_rules(\%rule_list);


#??? The following are not needed if use makeB.  
#    ?? They may be set too early?
# Arrays and hashes for picking out accessible rules.
# Distinguish rules for making files and others
    @accessible_all = sort ( &rdb_accessible( keys %requested_filerules, keys %one_time ));
    %accessible_filerules = ();
    foreach (@accessible_all) {
        unless ( /view/ || /print/ ) { $accessible_filerules{$_} = 1; }
    }
    @accessible_filerules = sort  keys %accessible_filerules;

#    show_array ( "=======All rules used", @accessible_all );
#    show_array ( "=======Requested file rules", sort keys %requested_filerules );
#    show_array ( "=======Rules for files", @accessible_filerules );

    if ( $diagnostics ) {
       print "$My_name: Rules after start up for '$texfile_name'\n";
       rdb_show();
    }

    %primaries = ();
    foreach (@accessible_all) {
        if ( ($_ eq 'latex') || ($_ eq 'pdflatex') ) { $primaries{$_} = 1; }
    }

    $have_fdb = 0;
    if ( (! -e $fdb_file) && (! -e "$root_filename.aux") ) {
        # No aux and no fdb file => set up trivial aux file 
        #    and corresponding fdb_file.  Arrange them to provoke one run 
        #    as minimum, but no more if actual aux file is trivial.
        #    (Useful on big files without cross references.)
        &set_trivial_aux_fdb;
    }

    if ( -e $fdb_file ) {
        $rdb_errors = rdb_read( $fdb_file );
        $have_fdb = ($rdb_errors == 0);
    }
    if (!$have_fdb) { 
        # We didn't get a valid set of data on files used in
        # previous run.  So use filetime criterion for make
        # instead of change from previous run, until we have
        # done our own make.
	rdb_recurseA( [keys %possible_primaries],
		      sub{ if ( $$Ptest_kind == 1 ) { $$Ptest_kind = 3;} }
        );
        if ( -e "$root_filename.log" ) {
	    rdb_for_some( [keys %possible_primaries], \&rdb_set_from_logB );
	}
    }
    foreach $rule ( rdb_accessible( uniq1( keys %requested_filerules  )  ) ){
        # For all source files of all accessible rules, 
        #    if the file data are not already set (e.g., from fdb_latexmk
        #    file, set them from disk.
        rdb_one_rule ($rule, undef, 
                      sub{ if ( $$Ptime == 0) { &rdb_update1; } }
	);
    }

    if ($go_mode) {
        # Force everything to be remade.
	rdb_recurseA( [keys %requested_filerules], sub{$$Pout_of_date=1;}  );
    }


    if ( $diagnostics ) {
       print "$My_name: Rules after initialization\n";
       rdb_show();
    }

    #************************************************************

    if ( $preview_continuous_mode ) { 
        &make_preview_continuousB; 
        # Will probably exit by ctrl/C and never arrive here.
        next FILE;
    }


## Handling of failures:
##    Variable $failure is set to indicate a failure, with information
##       put in $failure_msg.  
##    These variables should be set to 0 and '' at any point at which it
##       should be assumed that no failures have occurred.
##    When after a routine is called it is found that $failure is set, then
##       processing should normally be aborted, e.g., by return.
##    Then there is a cascade of returns back to the outermost level whose 
##       responsibility is to handle the error.
##    Exception: An outer level routine may reset $failure and $failure_msg
##       after initial processing, when the error condition may get 
##       ameliorated later.
    #Initialize failure flags now.
    $failure = 0;
    $failure_msg = '';
    $failure = rdb_makeB( keys %requested_filerules );
    if ($failure > 0) { next FILE; }
    rdb_for_some( [keys %one_time], \&rdb_run1 );
} # end FILE
continue {
    if ($dependents_list) { rdb_list(); }
    # Handle any errors
    $error_message_count = rdb_show_rule_errors();
    if ( ($error_message_count == 0) || ($failure > 0) ) {
        if ( $failure_msg ) {
            #Remove trailing space
            $failure_msg =~ s/\s*$//;
            warn "$My_name: Did not finish processing file:\n   $failure_msg\n";
            $failure = 1;
        }
    }
    if ( ($failure > 0) || ($error_message_count > 0) ) {
        $failure_count ++;
        push @failed_primaries, $filename;
    }
    &ifcd_popd;
}
# If we get here without going through the continue section:
if ( $do_cd && ($#dir_stack > -1) ) {
   # Just in case we did an abnormal exit from the loop
   warn "$My_name: Potential bug: dir_stack not yet unwound, undoing all directory changes now\n";
   &finish_dir_stack;
}

if ($failure_count > 0) {
    if ( $#file_list > 0 ) {
        # Error occured, but multiple files were processed, so
        #     user may not have seen all the error messages
        warn "\n------------\n";
        show_array( 
           "$My_name: Some operations failed, for the following tex file(s)", 
           @failed_primaries);
    }
    if ( !$force_mode ) {
      warn "$My_name: Use the -f option to force complete processing.\n";
    }
    exit 12;
}



# end MAIN PROGRAM
#############################################################

sub fix_cmds {
   # If commands do not have placeholders for %S etc, put them in
    foreach ($latex, $pdflatex, $lpr, $lpr_dvi, $lpr_pdf,
             $pdf_previewer, $ps_previewer, $ps_previewer_landscape,
             $dvi_previewer, $dvi_previewer_landscape,
             $kpsewhich
    ) {
        # Source only
        if ( $_ && ! /%/ ) { $_ .= " %O %S"; }
    }
    foreach ($bibtex) {
        # Base only
        if ( $_ && ! /%/ ) { $_ .= " %O %B"; }
    }
    foreach ($dvipdf, $ps2pdf) {
        # Source and dest without flag for destination
        if ( $_ && ! /%/ ) { $_ .= " %O %S %D"; }
    }
    foreach ($dvips, $makeindex) {
        # Source and dest with -o dest before source
        if ( $_ && ! /%/ ) { $_ .= " %O -o %D %S"; }
    }
    foreach ($dvi_filter, $ps_filter) {
        # Source and dest, but as filters
        if ( $_ && ! /%/ ) { $_ .= " %O <%S >%D"; }
    }
} #END fix_cmds

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

sub add_option {
    # Call add_option( \$cmd, $opt )
    # Add option to command
    if ( ${$_[0]} !~ /%/ ) { &fix_cmds; }
    ${$_[0]} =~ s/%O/$_[1] %O/;
} #END add_option

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

sub rdb_make_rule_list {
# Substitutions: %S = source, %D = dest, %B = this rule's base
#                %T = texfile, %R = root = base for latex.

    # Defaults for dvi, ps, and pdf files
    # Use local, not my, so these variables can be referenced
    local $dvi_final = "%R.dvi";
    local $ps_final  = "%R.ps";
    local $pdf_final = "%R.pdf";
    if ( length($dvi_filter) > 0) {
        $dvi_final = "%R.dviF";
    }
    if ( length($ps_filter) > 0) {
        $ps_final = "%R.psF";
    }

    my $print_file = '';
    my $print_cmd = '';
    if ( $print_type eq 'dvi' ) {
	$print_file = $dvi_final;
	$print_cmd = $lpr_dvi;
    }
    elsif ( $print_type eq 'pdf' ) {
	$print_file = $pdf_final;
	$print_cmd = $lpr_pdf;
    }
    elsif ( $print_type eq 'ps' ) {
	$print_file = $ps_final;
	$print_cmd = $lpr;
    }

    my $view_file = '';
    my $viewer = '';
    my $viewer_update_method = 0;
    my $viewer_update_signal = undef;
    my $viewer_update_command = undef;

    if ( ($view eq 'dvi') || ($view eq 'pdf') || ($view eq 'ps') ) { 
        $view_file = ${$view.'_final'};
        $viewer = ${$view.'_previewer'};
        $viewer_update_method = ${$view.'_update_method'};
        $viewer_update_signal = ${$view.'_update_signal'};
        if (defined ${$view.'_update_command'}) {
           $viewer_update_command = ${$view.'_update_command'};
        }
    }
    # Specification of internal command for viewer update:
    my $PA_update = ['do_update_view', $viewer_update_method, $viewer_update_signal, 0, 1];

# For test_kind: Use file contents for latex and friends, but file time for the others.
# This is because, especially for dvi file, the contents of the file may contain
#    a pointer to a file to be included, not the contents of the file! 
    %rule_list = (
        'latex'    => [ 'primary',  "$latex",     '',            "%T",        "%B.dvi",  "%R", 1 ],
        'pdflatex' => [ 'primary',  "$pdflatex",  '',            "%T",        "%B.pdf",  "%R", 1 ],
        'dvipdf'   => [ 'external', "$dvipdf",    'do_viewfile', $dvi_final,  "%B.pdf",  "%R", 2 ],
        'dvips'    => [ 'external', "$dvips",     'do_viewfile', $dvi_final,  "%B.ps",   "%R", 2 ],
        'dvifilter'=> [ 'external', $dvi_filter,  'do_viewfile', "%B.dvi",    "%B.dviF", "%R", 2 ],
        'ps2pdf'   => [ 'external', "$ps2pdf",    'do_viewfile', $ps_final,   "%B.pdf",  "%R", 2 ],
        'psfilter' => [ 'external', $ps_filter,   'do_viewfile', "%B.ps",     "%B.psF",  "%R", 2 ],
        'print'    => [ 'external', "$print_cmd", 'if_source',   $print_file, "",        "",   2 ],
        'update_view' => [ 'external', $viewer_update_command, $PA_update,
                               $view_file,  "",        "",   2 ],
        'view'     => [ 'external', "$viewer",    'if_source',   $view_file,  "",        "",   2 ],
    );
    %source_list = ();
    foreach my $rule (keys %rule_list) {
        $source_list{$rule} = [];
        my $PAsources = $source_list{$rule};
        my ( $cmd_type, $cmd, $source, $dest, $root ) = @{$rule_list{$rule}};
        if ($source) {
	    push @$PAsources, [ $rule, $source, '' ];
	}
    }

# Ensure we only have one way to make pdf file, and that it is appropriate:
    if ($pdf_mode == 1) { delete $rule_list{'dvipdf'}; delete $rule_list{'ps2pdf'}; }
    elsif ($pdf_mode == 2) { delete $rule_list{'dvipdf'}; delete $rule_list{'pdflatex'}; }
    else { delete $rule_list{'pdflatex'}; delete $rule_list{'ps2pdf'}; }

} # END rdb_make_rule_list 

#************************************************************

sub rdb_set_rules {
    # Call rdb_set_rules( \%rule_list, ...)
    # Set up rule database from definitions

    # Map of files to rules that MAKE them:
    %rule_db = ();

    foreach my $Prule_list (@_) {
	foreach my $rule ( keys %$Prule_list) {
	    my ( $cmd_type, $ext_cmd, $int_cmd, $source, $dest, $base, $test_kind ) = @{$$Prule_list{$rule}};
	    my $needs_making = 0;
	    # Substitute in the filename variables, since we will use
	    # those for determining filenames.  But delay expanding $cmd 
	    # until run time, in case of changes.
	    foreach ($base, $source, $dest ) { 
		s/%R/$root_filename/;
	    }
	    foreach ($source, $dest ) { 
		s/%B/$base/;
		s/%T/$texfile_name/;
	    }
    #        print "$rule: $cmd_type, EC='$ext_cmd', IC='$int_cmd', $test_kind,\n",
    #              "    S='$source', D='$dest', B='$base' $needs_making\n";
	    rdb_create_rule( $rule, $cmd_type, $ext_cmd, $int_cmd, $test_kind, 
			     $source, $dest, $base,
			     $needs_making, undef, 1 );
# !! ?? Last line was
#			     $needs_making, undef, ($test_kind==1) );
	}
    } # End arguments of subroutine
    &rdb_make_links;
} # END rdb_set_rules

#************************************************************

sub rdb_make_links {
# ?? Problem if there are multiple rules for getting a file.  Notably pdf.
#    Which one to choose?
    # Create $from_rule if there's a suitable rule.
    # Map files to rules:
    local %from_rules = ();
    rdb_for_all( sub{ if($$Pdest){$from_rules{$$Pdest} = $rule;} } );
#??    foreach (sort keys %from_rules) {print "D='$_' F='$from_rules{$_}\n";}
    rdb_for_all( 
        0,
        sub{ 
            if ( exists $from_rules{$file} ) { $$Pfrom_rule = $from_rules{$file}; }
#??            print "$rule: $file, $$Pfrom_rule\n";
	}
    );
    rdb_for_all( 
        0,
        sub{ 
            if ( exists $from_rules{$file} ) { 
                $$Pfrom_rule = $from_rules{$file}; 
            }
            if ( $$Pfrom_rule && (! rdb_rule_exists( $$Pfrom_rule ) ) ) {
                $$Pfrom_rule = '';
	    }
#??            print "$rule: $file, $$Pfrom_rule\n";
	}
    );
} # END rdb_make_links

#************************************************************

sub set_trivial_aux_fdb {
    # 1. Write aux file EXACTLY as would be written if the tex file
    #    had no cross references, etc. I.e., a minimal .aux file. 
    # 2. Write a corresponding fdb file
    # 3. Provoke a run of (pdf)latex (actually of all primaries). 

    local $aux_file = "$root_filename.aux";
    open( aux_file, '>', $aux_file )
        or die "Cannot write file '$aux_file'\n";
    print aux_file "\\relax \n";
    close(aux_file);

    foreach my $rule (keys %primaries ) { 
        rdb_ensure_file( $rule, $texfile_name );
        rdb_ensure_file( $rule, $aux_file );
        rdb_one_rule(  $rule,  
                       sub{ $$Pout_of_date = 1; }
                    );
    }
    &rdb_write( $fdb_file );
} #END set_trivial_aux_fdb

#************************************************************
#### Particular actions
#************************************************************
#************************************************************

sub do_cusdep {
    # Unconditional application of custom-dependency
    # except that rule is not applied if the source file source 
    # does not exist, and an error is returned if the dest is not made.
    #
    # Assumes rule context for the custom-dependency, and that my first 
    # argument is the name of the subroutine to apply
    my $func_name = $_[0];
    my $return = 0;
    if ( !-e $$Psource ) {
        # Source does not exist.  Users of this rule will need to turn
        # it off when custom dependencies are reset
	if ( !$silent ) {
## ??? Was commented out.  1 Sep. 2008 restored, for cusdep no-file-exists issue
            warn "$My_name: In trying to apply custom-dependency rule\n",
            "  to make '$$Pdest' from '$$Psource'\n",
            "  the source file has disappeared since the last run\n";
	}
        # Treat as successful
    }
    elsif ( !$func_name ) {
        warn "$My_name: Possible misconfiguration or bug:\n",
        "  In trying to apply custom-dependency rule\n",
        "  to make '$$Pdest' from '$$Psource'\n",
        "  the function name is blank.\n";
    }
    elsif ( ! defined &$func_name ) {
        warn "$My_name: Misconfiguration or bug,",
        " in trying to apply custom-dependency rule\n",
        "  to make '$$Pdest' from '$$Psource'\n",
        "  function name '$func_name' does not exists.\n";
    }
    else {
	my $cusdep_ret = &$func_name( $$Pbase );
        if ( defined $cusdep_ret && ($cusdep_ret != 0) ) {
	    $return = $cusdep_ret;
	    if ($return) {
                warn "Rule '$rule', function '$func_name'\n",
                     "   failed with return code = $return\n";
	    }
	}
        elsif ( !-e $$Pdest ) {
            # Destination non-existent, but routine failed to give an error
            warn "$My_name: In running custom-dependency rule\n",
            "  to make '$$Pdest' from '$$Psource'\n",
            "  function '$func_name' did not make the destination.\n";
	    $return = -1;
	}
    }
    return $return;
}  # END do_cusdep

#************************************************************

sub do_viewfile {
    # Unconditionally make file for viewing, going through temporary file if
    # Assumes rule context

    my $return = 0;
    my ($base, $path, $ext) = fileparseA( $$Pdest );
    if ( &view_file_via_temporary ) {
        my $tmpfile = tempfile1( "${root_filename}_tmp", $ext );
        $return = &rdb_ext_cmd1( '', '', $tmpfile );
        move( $tmpfile, $$Pdest );
    }
    else {
        $return = &rdb_ext_cmd;
    }
    return $return;
} #END do_viewfile

#************************************************************

sub do_update_view {
    # Update viewer
    # Assumes rule context
    # Arguments: (method, signal, viewer_process)

    my $return = 0;

    # Although the process is passed as an argument, we'll need to update it.
    # So (FUDGE??) bypass the standard interface for the process.
    # We might as well do this for all the arguments.
    my $viewer_update_method = ${$PAint_cmd}[1];
    my $viewer_update_signal = ${$PAint_cmd}[2];
    my $Pviewer_process             = \${$PAint_cmd}[3];
    my $Pneed_to_get_viewer_process = \${$PAint_cmd}[4];
    
    if ($viewer_update_method == 2) {
	if ($$Pneed_to_get_viewer_process) {
            $$Pviewer_process = &find_process_id( $$Psource );
            if ($$Pviewer_process != 0) {
                $$Pneed_to_get_viewer_process = 0;
	    }
	}
        if ($$Pviewer_process == 0) {
            print "$My_name: need to signal viewer for file '$$Psource', but didn't get \n",
                  "   process ID for some reason, e.g., no viewer, bad configuration, bug\n"
                if $diagnostics ;             
        }
        elsif ( defined $viewer_update_signal) {
            print "$My_name: signalling viewer, process ID $$Pviewer_process\n"
                if $diagnostics ;
	    kill $viewer_update_signal, $$Pviewer_process;
	}
        else {
            warn "$My_name: viewer is supposed to be sent a signal\n",
                 "  but no signal is defined.  Misconfiguration or bug?\n";
            $return = 1;
        }
    }
    elsif ($viewer_update_method == 4) {
        if (defined $$Pext_cmd) {
            $return = &rdb_ext_cmd;
	}
        else {
            warn "$My_name: viewer is supposed to be updated by running a command,\n",
                 "  but no command is defined.  Misconfiguration or bug?\n";
        }
    }
    return $return;
} #END do_update_view

#************************************************************

sub if_source {
    # Unconditionally apply rule if source file exists.
    # Assumes rule context
    if ( -e $$Psource ) {
        return &rdb_ext_cmd;
    }
    else {
	return -1;
    }
} #END if_source

#************************************************************
#### Subroutines
#************************************************************
#************************************************************

# Finds the basename of the root file
# Arguments:
#  1 - Filename to breakdown
#  2 - Where to place base file
#  3 - Where to place tex file
#  Returns non-zero if tex file does not exist
#
# The rules for determining this depend on the implementation of TeX.
# The variable $extension_treatment determines which rules are used.

sub find_basename
#?? Need to use kpsewhich, if possible
{
  local($given_name, $base_name, $ext, $path, $tex_name);
  $given_name = $_[0];
  if ( "$extension_treatment" eq "miktex_old" ) {
       # Miktex v. 1.20d: 
       #   1. If the filename has an extension, then use it.
       #   2. Else append ".tex".
       #   3. The basename is obtained from the filename by
       #      removing the path component, and the extension, if it
       #      exists.  If a filename has a multiple extension, then
       #      all parts of the extension are removed. 
       #   4. The names of generated files (log, aux) are obtained by
       #      appending .log, .aux, etc to the basename.  Note that
       #      these are all in the CURRENT directory, and the drive/path
       #      part of the originally given filename is ignored.
       #
       #   Thus when the given filename is "\tmp\a.b.c", the tex
       #   filename is the same, and the basename is "a".

       ($base_name, $path, $ext) = fileparse( $given_name, '\..*' );
       if ( "$ext" eq "") { $tex_name = "$given_name.tex"; }
       else { $tex_name = $given_name; }
       $_[1] = $base_name;
       $_[2] = $tex_name;
  }
  elsif ( "$extension_treatment" eq "unix" ) {
       # unix (at least web2c 7.3.1) => 
       #   1. If filename.tex exists, use it, 
       #   2. else if filename exists, use it.
       #   3. The base filename is obtained by deleting the path
       #      component and, if an extension exists, the last
       #      component of the extension, even if the extension is
       #      null.  (A name ending in "." has a null extension.)
       #   4. The names of generated files (log, aux) are obtained by
       #      appending .log, .aux, etc to the basename.  Note that
       #      these are all in the CURRENT directory, and the drive/path
       #      part of the originally given filename is ignored.
       #
       #   Thus when the given filename is "/tmp/a.b.c", there are two
       #   cases: 
       #      a.  /tmp/a.b.c.tex exists.  Then this is the tex file,
       #          and the basename is "a.b.c".
       #      b.  /tmp/a.b.c.tex does not exist.  Then the tex file is
       #          "/tmp/a.b.c", and the basename is "a.b".

      if ( -e "$given_name.tex" ) {
         $tex_name = "$given_name.tex";
      }
      else {
         $tex_name = "$given_name";
      }
      ($base_name, $path, $ext) = fileparse( $tex_name, '\.[^\.]*' );
      $_[1] = $base_name;
      $_[2] = $tex_name;
  }
  else {
     die "$My_name: Incorrect configuration gives \$extension_treatment=",
         "'$extension_treatment'\n";
  }
   if ($diagnostics) {
      print "Given='$given_name', tex='$tex_name', base='$base_name'\n";
  }
  return ! -e $tex_name;
} #END find_basename

#************************************************************

sub make_preview_continuousB {
    local @changed = ();
    local @disappeared = ();
    local @no_dest = ();       # Non-existent destination files
    local @rules_never_run = ();
    local @rules_to_apply = ();

    local $failure = 0;
    local $runs = 0;
    local %rules_applied = ();
    local $updated = 0;

    # What to make?
    my @targets = keys %requested_filerules;

    $quell_uptodate_msgs = 1;

    local $view_file = '';
    rdb_one_rule( 'view', sub{ $view_file = $$Psource; } );
  
    if ( ($view eq 'dvi') || ($view eq 'pdf') || ($view eq 'ps') ) { 
        warn "Viewing $view\n";
    }
    elsif ( $view eq 'none' ) {
        warn "Not using a previewer\n";
        $view_file = '';
    }
    else {
        warn "$My_name:  BUG: Invalid preview method '$view'\n";
        exit 20;
    }

    my $viewer_running = 0;    # No viewer known to be running yet
    # Get information from update_view rule
    local $viewer_update_method = 0;
    # Pointers so we can update the following:
    local $Pviewer_process = undef;    
    local $Pneed_to_get_viewer_process = undef;
    rdb_one_rule( 'update_view', 
                  sub{ $viewer_update_method = $$PAint_cmd[1]; 
                       $Pviewer_process = \$$PAint_cmd[3]; 
                       $Pneed_to_get_viewer_process = \$$PAint_cmd[4]; 
                     } 
                );
    # Note that we don't get the previewer process number from the program
    # that starts it; that might only be a script to get things set up and the 
    # actual previewer could be (and sometimes IS) another process.

    if ( ($view_file ne '') && (-e $view_file) && !$new_viewer_always ) {
        # Is a viewer already running?
        #    (We'll save starting up another viewer.)
        $$Pviewer_process = &find_process_id( $view_file );
        if ( $$Pviewer_process ) {
            warn "$My_name: Previewer is already running\n" 
              if !$silent;
            $viewer_running = 1;
            $$Pneed_to_get_viewer_process = 0;
        }
    }

    # Loop forever, rebuilding .dvi and .ps as necessary.
    # Set $first_time to flag first run (to save unnecessary diagnostics)
CHANGE:
    for (my $first_time = 1; 1; $first_time = 0 ) {
        my %rules_to_watch = %requested_filerules;
        $updated = 0;
        $failure = 0;
        $failure_msg = '';
        if ( $MSWin_fudge_break && ($^O eq "MSWin32") ) {
            # Fudge under MSWin32 ONLY, to stop perl/latexmk from
            #   catching ctrl/C and ctrl/break, and let it only reach
            #   downstream programs. See comments at first definition of
            #   $MSWin_fudge_break.
            $SIG{BREAK} = $SIG{INT} = 'IGNORE';
        }
        $failure = rdb_makeB( @targets );

##     warn "=========Viewer PID = $$Pviewer_process; updated=$updated\n";

        if ( $MSWin_fudge_break && ($^O eq "MSWin32") ) {
            $SIG{BREAK} = $SIG{INT} = 'DEFAULT';
        }
        # Start viewer if needed.
        if ( ($failure > 0) && (! $force_mode) ) {
            # No viewer yet
	}
        elsif ( ($view_file ne '') && (-e $view_file) && $updated && $viewer_running ) {
            # A viewer is running.  Explicitly get it to update screen if we have to do it:
            rdb_one_rule( 'update_view', \&rdb_run1 );
        }
        elsif ( ($view_file ne '') && (-e $view_file) && !$viewer_running ) {
            # Start the viewer
	    if ( !$silent ) {
                if ($new_viewer_always) {
                    warn "$My_name: starting previewer for '$view_file'\n",
                         "------------\n";
  	        }
                else {
                    warn "$My_name: I have not found a previewer that ",
                         "is already running. \n",
                         "   So I will start it for '$view_file'\n",
                         "------------\n";
 	       }
	    }
            local $retcode = 0;
            rdb_one_rule( 'view', sub { $retcode = &rdb_run1;} );
            if ( $retcode != 0 ) {
                if ($force_mode) {
                    warn "$My_name: I could not run previewer\n";
                }
                else {
                    &exit_msg1( "I could not run previewer", $retcode);
                }
            }
            else {
                $viewer_running = 1;
                $$Pneed_to_get_viewer_process = 1;
	    } # end analyze result of trying to run viewer
        } # end start viewer
        if ( $failure > 0 ) {
            if ( !$failure_msg ) {
	        $failure_msg = 'Failure to make the files correctly';
            }
            @pre_primary = ();   # Array of rules
            @post_primary = ();  # Array of rules
            @one_time = ();      # Array of rules
            &rdb_classify_rules( \%possible_primaries, keys %requested_filerules );
            # There will be files changed during the run that are irrelevant.
            # We need to wait for the user to change the files.

            # So set the GENERATED files from (pdf)latex as up-to-date:
            rdb_for_some( [keys %current_primaries], \&rdb_update_gen_files );

            # And don't watch for changes for post_primary rules (ps and pdf 
            # from dvi, etc haven't been run after an error in (pdf)latex, so
            # are out-of-date by filetime criterion, but they should not be run
            # until after another (pdf)latex run:
            foreach (@post_primary) { delete $rules_to_watch{$_}; }

            $failure_msg =~ s/\s*$//;  #Remove trailing space
            warn "$My_name: $failure_msg\n",
    "    ==> You will need to change a source file before I do another run <==\n";
	}
        rdb_show_rule_errors();
        if ( $first_time || $updated || $failure ) {
            print "\n=== Watching for updated files. Use ctrl/C to stop ...\n";
        }
        $waiting = 1; if ($diagnostics) { warn "WAITING\n"; }
        &catch_break;
        $have_break = 0;
  WAIT: while (1) {
           sleep( $sleep_time );
	   if ($have_break) { last WAIT; }
           if ( rdb_new_changes(keys %rules_to_watch) ) { 
  	       if (!$silent) {
                   warn "$My_name: Need to remake files.\n";
 	           &rdb_diagnose_changes( '  ' );
	       }
               last WAIT;
	   }
        # Does this do this job????
           local $new_files = 0;
           rdb_for_some( [keys %current_primaries], sub{ $new_files += &rdb_find_new_filesB } );
           if ($new_files > 0) {
	       warn "$My_name: New file(s) found.\n";
               last WAIT; 
	   }
	   if ($have_break) { last WAIT; }
     } # end WAIT:
     &default_break;
     if ($have_break) { 
          print "$My_name: User typed ctrl/C or ctrl/break.  I'll stop.\n";
          exit;
     }
     $waiting = 0; if ($diagnostics) { warn "NOT       WAITING\n"; }
  } #end infinite_loop CHANGE:
} #END sub make_preview_continuousB

#************************************************************

sub process_rc_file {
    # Usage process_rc_file( filename )
    # Run rc_file whose name is given in first argument
    #    Exit with code 11 if file could not be read.  
    #      (In general this is not QUITE the right error)
    #    Exit with code 13 if there is a syntax error or other problem.
    # ???Should I leave the exiting to the caller (perhaps as an option)?
    #     But I can always catch it with an eval if necessary.
    #     That confuses ctrl/C and ctrl/break handling.
    my $rc_file = $_[0];
    warn "$My_name: Executing PERL code in file '$rc_file'...\n" 
        if  $diagnostics;
    do( $rc_file );
    # The return value from the do is not useful, since it is the value of 
    #    the last expression evaluated, which could be anything.
    # The correct test of errors is on the values of $! and $@.

# This is not entirely correct.  On gluon2:
#      rc_file does open of file, and $! has error, apparently innocuous
#      See ~/proposal/06/latexmkrc-effect

    my $OK = 1;
    if ( $! ) {
        # Get both numeric error and its string, by forcing numeric and 
        #   string contexts:
        my $err_no = $!+0;
        my $err_string = "$!";
        warn "$My_name: Initialization file '$rc_file' could not be read,\n",
             "   or it gave some other problem. Error code \$! = $err_no.\n",
             "   Error string = '$err_string'\n";
	$! = 256;
        $OK = 0;
    }
    if ( $@ ) {
	$! = 256;
        # Indent the error message to make it easier to locate
        my $indented = prefix( $@, "    " );
        $@ = "";
        warn "$My_name: Initialization file '$rc_file' gave an error:\n",
            "$indented";
        $OK = 0;
    }
    if ( ! $OK ) { 
        die "$My_name: Stopping because of problem with rc file\n"; 
    }
} #END process_rc_file

#************************************************************

sub execute_code_string {
    # Usage execute_code_string( string_of_code )
    # Run the PERL code contained in first argument
    #    Exit with code 13 if there is a syntax error or other problem.
    # ???Should I leave the exiting to the caller (perhaps as an option)?
    #     But I can always catch it with an eval if necessary.
    #     That confuses ctrl/C and ctrl/break handling.
    my $code = $_[0];
    warn "$My_name: Executing initialization code specified by -e:\n",
         "   '$code'...\n" 
        if  $diagnostics;
    eval $code;
    # The return value from the eval is not useful, since it is the value of 
    #    the last expression evaluated, which could be anything.
    # The correct test of errors is on the values of $! and $@.

    if ( $@ ) {
	$! = 256;
        my $message = $@;
        $@ = "";
        $message =~ s/\s*$//;
        die "$My_name: ",
            "Stopping because executing following code from command line\n",
            "    $code\n",
            "gave an error:\n",
            "    $message\n";
    }
} #END execute_code_string

#************************************************************

sub cleanup1 {
    # Usage: cleanup1( exts_without_period, ... )
    foreach (@_) { unlink("$root_filename.$_"); }
} #END cleanup1

#************************************************************
#************************************************************
#************************************************************

#   Error handling routines, warning routines, help

#************************************************************

sub die_trace {
    # Call: die_trace( message );
    &traceback;   # argument(s) passed unchanged
    die "\n";
} #END die_trace

#************************************************************

sub traceback {
    # Call: &traceback 
    # or traceback( message,  )
    my $msg = shift;
    if ($msg) { warn "$msg\n"; }
    warn "Traceback:\n";
    my $i=0;     # Start with immediate caller
    while ( my ($pack, $file, $line, $func) = caller($i++) ) {
        if ($func eq 'die_trace') { next; }
        warn "   $func called from line $line\n";
    }
} #END traceback

#************************************************************

sub exit_msg1
{
  # exit_msg1( error_message, retcode [, action])
  #    1. display error message
  #    2. if action set, then restore aux file
  #    3. exit with retcode
  warn "\n------------\n";
  warn "$My_name: $_[0].\n";
  warn "-- Use the -f option to force complete processing.\n";

  my $retcode = $_[1];
  if ($retcode >= 256) {
     # Retcode is the kind returned by system from an external command
     # which is 256 * command's_retcode
     $retcode /= 256;
  }
  exit $retcode;
} #END exit_msg1

#************************************************************

sub warn_running {
   # Message about running program:
    if ( $silent ) {
        warn "$My_name: @_\n";
    }
    else {
        warn "------------\n@_\n------------\n";
    }
} #END warn_running

#************************************************************

sub exit_help
# Exit giving diagnostic from arguments and how to get help.
{
    warn "\n$My_name: @_\n",
         "Use\n",
         "   $my_name -help\nto get usage information\n";
    exit 10;
} #END exit_help


#************************************************************

sub print_help
{
  print
  "$My_name $version_num: Automatic LaTeX document generation routine\n\n",
  "Usage: $my_name [latexmk_options] [filename ...]\n\n",
  "  Latexmk_options:\n",
  "   -bibtex         - use bibtex when needed (default)\n",
  "   -bibtex-        - never use bibtex\n",
  "   -bibtex-cond    - use bibtex when needed, but only if the bib files exist\n",
  "   -bm <message> - Print message across the page when converting to postscript\n",
  "   -bi <intensity> - Set contrast or intensity of banner\n",
  "   -bs <scale> - Set scale for banner\n",
  "   -commands  - list commands used by $my_name for processing files\n",
  "   -c     - clean up (remove) all nonessential files, except\n",
  "            dvi, ps and pdf files.\n",
  "            This and the other clean-ups are instead of a regular make.\n",
  "   -C     - clean up (remove) all nonessential files\n",
  "            including aux, dep, dvi, postscript and pdf files\n",
  "            and file of database of file information\n",
  "   -CA     - clean up (remove) absolutely ALL nonessential files.\n",
  "            Now equivalent to -C option.\n",
  "   -CF     - Remove file of database of file information before doing \n",
  "            other actions\n",
  "   -cd    - Change to directory of source file when processing it\n",
  "   -cd-   - Do NOT change to directory of source file when processing it\n",
  "   -dependents   - Show list of dependent files after processing\n",
  "   -dependents-  - Do not show list of dependent files after processing\n",
  "   -dF <filter> - Filter to apply to dvi file\n",
  "   -dvi   - generate dvi\n",
  "   -dvi-  - turn off required dvi\n",
  "   -e <code> - Execute specified PERL code\n",
  "   -f     - force continued processing past errors\n",
  "   -f-    - turn off forced continuing processing past errors\n",
  "   -gg    - Super go mode: clean out generated files (-CA), and then\n",
  "            process files regardless of file timestamps\n",
  "   -g     - process regardless of file timestamps\n",
  "   -g-    - Turn off -g\n",
  "   -h     - print help\n",
  "   -help - print help\n",
  "   -jobname=STRING - set basename of output file(s) to STRING.\n",
  "            (Like --jobname=STRING on command line for many current\n",
  "            implementations of latex/pdflatex.)\n",
  "   -l     - force landscape mode\n",
  "   -l-    - turn off -l\n",
  "   -new-viewer    - in -pvc mode, always start a new viewer\n",
  "   -new-viewer-   - in -pvc mode, start a new viewer only if needed\n",
  "   -nobibtex      - never use bibtex\n",
  "   -nodependents  - Do not show list of dependent files after processing\n",
  "   -pdf   - generate pdf by pdflatex\n",
  "   -pdfdvi - generate pdf by dvipdf\n",
  "   -pdfps - generate pdf by ps2pdf\n",
  "   -pdf-  - turn off pdf\n",
  "   -ps    - generate postscript\n",
  "   -ps-   - turn off postscript\n",
  "   -pF <filter> - Filter to apply to postscript file\n",
  "   -p     - print document after generating postscript.\n",
  "            (Can also .dvi or .pdf files -- see documentation)\n",
  "   -print=dvi     - when file is to be printed, print the dvi file\n",
  "   -print=ps      - when file is to be printed, print the ps file (default)\n",
  "   -print=pdf     - when file is to be printed, print the pdf file\n",
  "   -pv    - preview document.  (Side effect turn off continuous preview)\n",
  "   -pv-   - turn off preview mode\n",
  "   -pvc   - preview document and continuously update.  (This also turns\n",
  "                on force mode, so errors do not cause $my_name to stop.)\n",
  "            (Side effect: turn off ordinary preview mode.)\n",
  "   -pvc-  - turn off -pvc\n",
  "   -r <file> - Read custom RC file\n",
  "   -silent  - silence progress messages from called programs\n",
  "   -v     - display program version\n",
  "   -verbose - display usual progress messages from called programs\n",
  "   -version      - display program version\n",
  "   -view=default - viewer is default (dvi, ps, pdf)\n",
  "   -view=dvi     - viewer is for dvi\n",
  "   -view=none    - no viewer is used\n",
  "   -view=ps      - viewer is for ps\n",
  "   -view=pdf     - viewer is for pdf\n",
  "   filename = the root filename of LaTeX document\n",
  "\n",
  "-p, -pv and -pvc are mutually exclusive\n",
  "-h, -c and -C overides all other options.\n",
  "-pv and -pvc require one and only one filename specified\n",
  "All options can be introduced by '-' or '--'.  (E.g., --help or -help.)\n",
  "Contents of RC file specified by -r overrides options specified\n",
  "  before the -r option on the command line\n";

} #END print_help

#************************************************************
sub print_commands
{
  warn "Commands used by $my_name:\n",
       "   To run latex, I use \"$latex\"\n",
       "   To run pdflatex, I use \"$pdflatex\"\n",
       "   To run bibtex, I use \"$bibtex\"\n",
       "   To run makeindex, I use \"$makeindex\"\n",
       "   To make a ps file from a dvi file, I use \"$dvips\"\n",
       "   To make a ps file from a dvi file with landscape format, ",
           "I use \"$dvips_landscape\"\n",
       "   To make a pdf file from a dvi file, I use \"$dvipdf\"\n",
       "   To make a pdf file from a ps file, I use \"$ps2pdf\"\n",
       "   To view a pdf file, I use \"$pdf_previewer\"\n",
       "   To view a ps file, I use \"$ps_previewer\"\n",
       "   To view a ps file in landscape format, ",
            "I use \"$ps_previewer_landscape\"\n",
       "   To view a dvi file, I use \"$dvi_previewer\"\n",
       "   To view a dvi file in landscape format, ",
            "I use \"$dvi_previewer_landscape\"\n",
       "   To print a ps file, I use \"$lpr\"\n",
       "   To print a dvi file, I use \"$lpr_dvi\"\n",
       "   To print a pdf file, I use \"$lpr_pdf\"\n",
       "   To find running processes, I use \"$pscmd\", \n",
       "      and the process number is at position $pid_position\n";
   warn "Notes:\n",
        "  Command starting with \"start\" is run detached\n",
        "  Command that is just \"start\" without any other command, is\n",
        "     used under MS-Windows to run the command the operating system\n",
        "     has associated with the relevant file.\n",
        "  Command starting with \"NONE\" is not used at all\n";
} #END print_commands

#************************************************************

sub view_file_via_temporary {
    return $always_view_file_via_temporary 
           || ($pvc_view_file_via_temporary && $preview_continuous_mode);
} #END view_file_via_temporary

#************************************************************
#### Tex-related utilities


sub check_bibtex_log {
    # Check for bibtex warnings:
    # Usage: check_bibtex_log( base_of_bibtex_run )
    # return 0: OK, 1: bibtex warnings, 2: bibtex errors, 
    #        3: could not open .blg file.

    my $base = $_[0];
    my $log_name = "$base.blg";
    my $log_file = new FileHandle;
    open( $log_file, "<$log_name" )
      or return 3;
    my $have_warning = 0;
    my $have_error = 0;
    while (<$log_file>) {
        if (/Warning--/) { 
            #print "Bibtex warning: $_"; 
            $have_warning = 1;
        }
        if (/error message/) { 
            #print "Bibtex error: $_"; 
            $have_error = 1;
        }
    }
    close $log_file;
    if ($have_error) {return 2;}
    if ($have_warning) {return 1;}
    return 0;
} #END check_bibtex_log

#**************************************************

sub clean_file_name{
    # Convert filename found in log file to true filename.
    # Used normally only by parse_logB, below
    # 1. For names of form 
    #    `"string".ext', which arises e.g., from \jobname.bbl:
    #    when the base filename contains spaces, \jobname has quotes.
    #    and from \includegraphics with basename specified.
    # 2. Or "string.ext" from \includegraphcs with basename and ext specified.
    my $filename = $_[0];
    $filename =~ s/^\"([^\"]*)\"(.*)$/$1$2/;
    return $filename;
}
# ------------------------------

sub parse_logB {
# Scan log file for: dependent files
#    reference_changed, bad_reference, bad_citation
# Return value: 1 if success, 0 if no log file.
# Set global variables:
#   %dependents: maps definite dependents to code:
#      0 = from missing-file line
#            May have no extension
#            May be missing path
#      1 = from 'File: ... Graphic file (type ...)' line
#            no path.  Should exist, but may need a search, by kpsewhich.
#      2 = from regular '(...' coding for input file, 
#            Has NO path, which it would do if LaTeX file
#            Highly likely to be mis-parsed line
#      3 = ditto, but has a path character ('/').  
#            Should be LaTeX file that exists.
#            If it doesn't exist, we have probably a mis-parsed line.
#            There's no need to do a search.
#      4 = definitive, which in this subroutine is only done:
#             for default dependents, 
#             and for files that exist and are source of conversion
#                reported by epstopdf et al.
#      5 = Had a missing file line.  Now the file exists.
#      6 = File was written during run.  (Overrides 5)
#      7 = File was created during run to be read in.  (Overrides 5 and 6)
#          (e.g., by epstopdf)
# Treat the following specially, since they have special rules
#   @bbl_files to list of .bbl files.
#   %idx_files to map from .idx files to .ind files.
# %generated_log: keys give set of files written by (pdf)latex (e.g., aux, idx)
#   as determined by \openout = ... lines in log file.
# Also set
#   $reference_changed, $bad_reference, $bad_citation
# Trivial or default values if log file does not exist/cannot be opened

# Give a quick way of looking up custom-dependency extensions
    my %cusdep_from = ();
    my %cusdep_to = ();
    foreach ( @cus_dep_list ) {
        my ($fromext, $toext) = split;
	$cusdep_from{$fromext} = $cusdep_from{".$fromext"} = $_;
	$cusdep_to{$toext} = $cusdep_to{".$toext"} = $_;
    }
#    print "==== Cusdep from-exts:"; foreach (keys %cusdep_from) {print " '$_'";} print "\n";
#    print "==== Cusdep to-exts:"; foreach (keys %cusdep_to) {print " '$_'";} print "\n";

    # Returned info:
    %dependents = ();
    foreach (@default_includes) { $dependents{$_} = 4; }
    @bbl_files = ();
    %idx_files = ();    # Maps idx_file to (ind_file, base)
    %generated_log = ();
    %conversions = ();
    # $primary_out is actual output file (dvi or pdf)
    # It is initialized before the call to this routine, to ensure
    # a sensible default in case of misparsing

    $reference_changed = 0;
    $bad_reference = 0;
    $bad_citation = 0;

    my $log_name = "$root_filename.log";
    my $log_file = new FileHandle;
    if ( ! open( $log_file, "<$log_name" ) ) {
        return 0;
    }
    if ($log_file_binary) { binmode $log_file; }
# Collect lines of log file
    my @lines = ();
    while(<$log_file>) { 
        # Could use chomp here, but that fails if there is a mismatch
        #    between the end-of-line sequence used by latex and that
        #    used by perl.  (Notably a problem with MSWin latex and
        #    cygwin perl!)
        s/[\n\r]*$//;
        # Handle wrapped lines:
        # They are lines brutally broken at exactly $log_wrap chars 
        #    excluding line-end.  Sometimes a line $log_wrap chars 
        #    long is an ordinary line, sometimes it is part of a line
        #    that was wrapped.  To handle all cases, I keep both
        #    options open by putting the line into @lines before
        #    and after appending the next line:  
        my $len = length($_);
        while ( ($len == $log_wrap) && !eof($log_file) ) {
            push @lines, $_;
            my $extra = <$log_file>;
            $extra =~ s/[\n\r]*$//;
            $len = length($extra);
            $_ .= $extra;
        }
        push @lines, $_;
    }
    close $log_file;

    push @lines, "";   # Blank line to terminate.  So multiline blocks 
              # are always terminated by non-block line, rather than eof.
    
    my $line = 0;
    # For parsing multiple line blocks of info
    my $current_pkg = "";   # non-empty string for package name, if in 
                            # middle of parsing multi-line block of form:
                            #       Package name ....
                            #       (name) ...
                            #       ...
    my $block_type = "";         # Specify information in such a block
    my $delegated_source = "";   # If it is a file conversion, specify source
    my $delegated_output = "";   #    and output file.  (Don't put in
                                 #    data structure until block is ended.)
    my %new_conversions = ();
    my @retries = ();
LINE:
    while( ($line <= $#lines) || ($#retries > -1) ) { 
        if ($#retries > -1) {
	    $_ = pop @retries;
	}
	else {
            $_ = $lines[$line];
            $line ++;
	}
        if ( /^! pdfTeX warning/ || /^pdfTeX warning/ ) {
            # This kind of warning is produced by some versions of pdftex
            # or produced by my reparse of warnings from other
            # versions.
            next;
	}
        elsif ( /^(.+)(pdfTeX warning.*)$/ ) {
            # Line contains a pdfTeX warnings that may have been
            # inserted directly after other material without an
            # intervening new line.  I think pdfTeX always inserts a
            # newline after the warning.  (From examination of source
            # code.) 
     	    push @retries, $1;
            # But continue parsing the original line, in case it was a
            # misparse, e.g., of a filename ending in 'pdfTeX';
	}
        if ( $line == 1 ){
	    if ( /^This is / ) {
	        # First line OK
                next LINE;
            } else {
                warn "$My_name: Error on first line of '$log_name'.  ".
                    "This is apparently not a TeX log file.\n";
                $failure = 1;
                $failure_msg = "Log file '$log_name' appears to have wrong format.";
                return 0;
	    }
	}
        if ( $block_type ) {
            # In middle of parsing block
            if ( /^\($current_pkg\)/ ) {
                # Block continues
                if ( ($block_type eq 'conversion') 
                     && /^\($current_pkg\)\s+Output file: <([^>]+)>/ ) 
                {
		    $delegated_output = $1;
                }
		next LINE;
	    }
            # Block has ended.
            if ($block_type eq 'conversion') {
                 $new_conversions{$delegated_source} =  $delegated_output;
	    }
            $current_pkg = $block_type 
                 = $delegated_source = $delegated_output = "";
            # Then process current line
	}
        # Check for changed references, bad references and bad citations:
        if (/Rerun to get/) { 
            warn "$My_name: References changed.\n";
            $reference_changed = 1;
        } 
	if (/LaTeX Warning: (Reference[^\001]*undefined)./) { 
	    warn "$My_name: $1 \n";
	    $bad_reference = 1;
	} 
	if (/LaTeX Warning: (Citation[^\001]*undefined)./) {
	    warn "$My_name: $1 \n";
	    $bad_citation = 1;
	}
        if (/Package natbib Warning: (Citation[^\001]*undefined)./) {
            warn "$My_name: $1 \n";
            $bad_citation = 1;
	}
	if ( /^Document Class: / ) {
	    # Class sign-on line
	    next LINE;
	}
	if ( /^\(Font\)/ ) {
	    # Font info line
	    next LINE;
	}
	if (/^No pages of output\./) {
            $primary_out = ''; 
	    warn "$My_name: Log file says no output from latex\n";
	    next LINE;
	}
	if ( /^Output written on\s+(.*)\s+\(\d+\s+page/ ) {
            $primary_out = $1;
	    warn "$My_name: Log file says output to '$1'\n"
	       unless $silent;
	    next LINE;
	}
	if ( /^Overfull / 
	     || /^Underfull / 
             || /^or enter new name\. \(Default extension: .*\)/ 
             || /^\*\*\* \(cannot \\read from terminal in nonstop modes\)/
           ) {
	    # Latex error/warning, etc.
	    next LINE;
	}
        if ( /^\\openout\d+\s*=\s*\`([^\']+)\'\.$/ ) {
            $generated_log{$1} = 1;
	    next LINE;
	}
        # Test for conversion produced by package:
	if ( /^Package (\S+) Info: Source file: <([^>]+)>/ ) {
	    # Info. produced by epstopdf (and possibly others) 
            #    about file conversion
            $current_pkg = $1;
            $delegated_source = $2;
            $block_type = 'conversion';
	    next LINE;
	}
#    Test for writing of index file.  The precise format of the message 
#    depends on which package (makeidx.sty , multind.sty or index.sty) and 
#    which version writes the message.
	if ( /Writing index file (.*)$/ ) {
            my $idx_file = '';
	    if ( /^Writing index file (.*)$/ ) {
                # From makeidx.sty or multind.sty
                $idx_file = $1;
	    }
            elsif ( /^index\.sty> Writing index file (.*)$/ ) {
                # From old versions of index.sty
                $idx_file = $1;
	    }
            elsif ( /^Package \S* Info: Writing index file (.*) on input line/ ) {
                # From new versions of index.sty
                $idx_file = $1;                
	    }
            else {
		warn "$My_name: Message indicates index file was written\n",
		     "  ==> but I do not know how to understand it: <==\n",
   		     "  '$_'\n";
                next LINE;
	    }
	    # Typically, there is trailing space, not part of filename:
	    $idx_file =~ s/\s*$//;
	    $idx_file = clean_file_name($idx_file);
	    my ($idx_base, $idx_path, $idx_ext) = fileparseA( $idx_file );
	    $idx_base = $idx_path.$idx_base;
            $idx_file = $idx_base.$idx_ext;
	    if ( $idx_ext eq '.idx' ) {
		warn "$My_name: Index file '$idx_file' was written\n"
		  unless $silent;
		$idx_files{$idx_file} = [ "$idx_base.ind", $idx_base ];
	    }
	    elsif ( exists $cusdep_from{$idx_ext} ) {
                if ( !$silent ) {
        	    warn "$My_name: Index file '$idx_file' was written\n";
                    warn "   Cusdep '$cusdep_from{$idx_ext}' should be used\n";
                }
                # No action needed here
	    }
	    else {
		warn "$My_name: Index file '$idx_file' written\n",
		     "  ==> but it has an extension I do not know how to handle <==\n";
	    }

	    next LINE;
	}
	if ( /^No file (.*?\.bbl)./ ) {
	    # Notice that the
	    my $bbl_file = clean_file_name($1);
	    warn "$My_name: Non-existent bbl file '$bbl_file'\n $_\n";
    	    $dependents{$bbl_file} = 0;
	    push @bbl_files, $bbl_file;
	    next LINE;
	}
	foreach my $pattern (@file_not_found) {
	    if ( /$pattern/ ) {
		my $file = clean_file_name($1);
		warn "$My_name: Missing input file: '$file' from line\n  '$_'\n"
		    unless $silent;
		$dependents{$file} = 0;
		next LINE;
	    }
	}
	if ( /^File: ([^\s\[]*) Graphic file \(type / ) {
	    # First line of message from includegraphics/x
	    $dependents{$1} = 1;
	    next LINE;
	}
	# Now test for generic lines to ignore, only after special cases!
	if ( /^File: / ) {
	   # Package sign-on line. Includegraphics/x also produces a line 
	   # with this signature, but I've already handled it.
	   next LINE;
	}
	if ( /^Package: / ) {
	    # Package sign-on line
	    next LINE;
	}
	if (/^\! LaTeX Error: / ) {
	    next LINE;
	}
   INCLUDE_CANDIDATE:
        while ( /\((.*$)/ ) {
        # Filename found by
        # '(', then filename, then terminator.
        # Terminators: obvious candidates: ')':  end of reading file
        #                                  '(':  beginning of next file
        #                                  ' ':  space is an obvious separator
        #                                  ' [': start of page: latex
        #                                        and pdflatex put a
        #                                        space before the '['
        #                                  '[':  start of config file
        #                                        in pdflatex, after
        #                                        basefilename.
        #                                  '{':  some kind of grouping
        # Problem: 
        #   All or almost all special characters are allowed in
        #   filenames under some OS, notably UNIX.  Luckily most cases
        #   are rare, if only because the special characters need
        #   escaping.  BUT 2 important cases are characters that are
        #   natural punctuation
        #   Under MSWin, spaces are common (e.g., "C:\Program Files")
        #   Under VAX/VMS, '[' delimits directory names.  This is
        #   tricky to handle.  But I think few users use this OS
        #   anymore.
        #
        # Solution: use ' [', but not '[' as first try at delimiter.
	# Then if candidate filename is of form 'name1[name2]', then
	#   try splitting it.  If 'name1' and/or 'name2' exists, put
	#   it/them in list, else just put 'name1[name2]' in list.
	# So form of filename is now:
	#  '(', 
	# then any number of characters that are NOT ')', '(', or '{'
	#   (these form the filename);
	# then ' [', or ' (', or ')', or end-of-string.
	# That fails for pdflatex
	# In log file:
	#   '(' => start of reading of file, followed by filename
	#   ')' => end of reading of file
	#   '[' => start of page (normally preceeded by space)
	# Remember: 
	#    filename (on VAX/VMS) may include '[' and ']' (directory
	#             separators) 
	#    filenames (on MS-Win) commonly include space.
        #    filenames on UNIX can included space.
        #    Miktex quotes filenames
        #    But web2c doesn't.  Then 
        #       (string  message
        #    is ambiguous: is the filename "string" or "string message".
        #    Allow both as candidates, since user filenames with spaces 
        #    are rare.  System filenames with spaces are common, but
        #    they are normally followed by a newline rather than messages. 

	# First step: replace $_ by whole of line after the '('
	#             Thus $_ is putative filename followed by other stuff.
            $_ = $1; 
            # Array of new candidate include files; sometimes more than one.
            my @new_includes = ();
            my $quoted = 0;
            if ( /^\"([^\(^\)^\"]+)\"/ ) {
               # Quoted file name, as from MikTeX
		$quoted = 1;
            }
            elsif ( /^([^\(^\)]*?)\s+[\[\{\<]/ ) {
                # Terminator: space then '[' or '{' or '<'
                # Use *? in condition: to pick up first ' [' (etc) 
                # as terminator
            }
            elsif ( /^([^\(^\)]*)\s+(?=\()/ ) {
                # Terminator is ' (', but '(' isn't in matched string,
                # so we keep the '(' ready for the next match
            }
            elsif  ( /^([^\(^\)]*)(\))/ ) {
                # Terminator is ')'
            }
	    else {
                #Terminator is end-of-string
            }
            $_ = $';       # Put $_ equal to the unmatched tail of string '
            my $include_candidate = $1;
            $include_candidate =~ s/\s*$//;   # Remove trailing space.
            if ( !$quoted && ($include_candidate =~ /(\S+)\s/ ) ){
                # Non-space-containing filename-candidate
                # followed by space followed by message
                # (Common)
                push @new_includes, $1;
	    }
            if ( $include_candidate eq "[]" ) {
                # Part of overfull hbox message
                next INCLUDE_CANDIDATE;
            }
            if ( $include_candidate =~ /^\\/ ) {
                # Part of font message
                next INCLUDE_CANDIDATE;
            }
            # Remove quotes around filename, as for MikTeX.  I've already
            # treated this as a special case.  For safety check here:
            $include_candidate =~ s/^\"(.*)\"$/$1/;

            push @new_includes, $include_candidate;
            if ( $include_candidate =~ /^(.+)\[([^\]]+)\]$/ ) {
                # Construct of form 'file1[file2]', as produced by pdflatex
                if ( -e $1 ) {
                    # If the first component exists, we probably have the
                    #   pdflatex form
                    push @new_includes, $1, $2;
	        }
                else {
                    # We have something else.
                    # So leave the original candidate in the list
	        }
	    }
	INCLUDE_NAME:
            foreach my $include_name (@new_includes) {
	        my ($base, $path, $ext) = fileparseB( $include_name );
                if ( ($path eq './') || ($path eq '.\\') ) {
                    $include_name = $base.$ext;
		}
                if ( $include_name !~ m'[/|\\]' ) {
                    # Filename does not include a path character
                    # High potential for misparsed line
		    $dependents{$include_name} = 2;
		} else {
		    $dependents{$include_name} = 3;
		}
		if ( $ext eq '.bbl' ) {
  		    warn "$My_name: Found input bbl file '$include_name'\n"
		       unless $silent;
  		    push @bbl_files, $include_name;
		}
	    } # INCLUDE_NAME
        } # INCLUDE_CANDIDATE
    } # LINE

    # Default includes are always definitive:
    foreach (@default_includes) { $dependents{$_} = 4; }

    ###print "New parse: \n";
    ###foreach (sort keys %dependents) { print "  '$_': $dependents{$_}\n"; }

    my @misparsed = ();
    my @missing = ();
    my @not_found = ();
CANDIDATE:
    foreach my $candidate (keys %dependents) {
        my $code = $dependents{$candidate};
        if ( -d $candidate ) {
            #  If $candidate is directory, it was presumably found from a 
            #     mis-parse, so remove it from the list.  (Misparse can 
            #     arise, for example from a mismatch of latexmk's $log_wrap
            #     value and texmf.cnf value of max_print_line.)
            delete $dependents{$candidate};
        }
        elsif ( -e $candidate ) {
            if ( exists $generated_log{$candidate} ){
	        $dependents{$candidate} = 6;
	    }
            elsif ($code == 0) {
	        $dependents{$candidate} = 5;
	    }
	    else {
		$dependents{$candidate} = 4;
	    }
	}
	elsif ($code == 1) {
            # Graphics file that is supposed to have been read.
            # Candidate name is as given in source file, not as path
            #   to actual file.
            # We have already tested that file doesn't exist, as given.
            #   so use kpsewhich.  
            # If the file still is not found, assume non-existent;
            my @kpse_result = kpsewhich( $candidate );
            if ($#kpse_result > -1) {
                $dependents{$kpse_result[0]} = 4;
		delete $dependents{$candidate};
		next CANDIDATE;
	    }
	    else {
		push @not_found, $candidate;
	    }
	}
        elsif ($code == 2) {
            # Candidate is from '(...' construct in log file, for input file
            #    which should include pathname if valid input file.
            # Name does not have pathname-characteristic character (hence
            #    $code==2.
            # We get here if candidate file does not exist with given name
            # Almost surely result of a misparsed line in log file.
	    delete $dependents{$candidate};
            push @misparse, $candidate;
	}
        elsif ($code == 3) {
            # Candidate is from '(...' construct in log file, for input file
            #    which should include pathname if valid input file.
            # Name does have pathname-characteristic character (hence
            #    $code==3.
            # But we get here only if candidate file does not exist with 
            # given name.  
            # Almost surely result of a misparsed line in log file.
            # But with lower probability than $code == 2
	    delete $dependents{$candidate};
            push @misparse, $candidate;
	}
	elsif ($code == 0) {
            my ($base, $path, $ext) = fileparseA($candidate);
            $ext =~ s/^\.//;
            if ( ($ext eq '') && (-e "$path$base.tex") ) {
                # I don't think the old version was correct.
                # If the missing-file report was of a bare
                #    extensionless file, and a corresponding .tex file
                #    exists, then the missing file does not correspond
                #    to the missing file, unless the .tex file was
                #    created during the run.  
                # OLD $dependents{"$path$base.tex"} = 4;
		# OLD delete $dependents{$candidate};
                # NEW:
		$dependents{"$path$base.tex"} = 4;
	    }
	    push @missing, $candidate;
	}
    }
CANDIDATE_PAIR:
    foreach my $delegated_source (keys %new_conversions) {
	my $delegated_output = $new_conversions{$delegated_source};
        my $rule = "Delegated $delegated_source, $delegated_output";
        # N.B. $delegated_source eq '' means the output file
        #      was created without a named input file.
        foreach my $candidate ($delegated_source, $delegated_output) {
            if (! -e $candidate ) {
                # The file might be somewhere that can be found
                #   in the search path of kpathsea:
                my @kpse_result = kpsewhich( $candidate,);
                if ($#kpse_result > -1) {
                    $candidate = $kpse_result[0];
	        }
	    }
	}
        if ( ( (-e $delegated_source) || ($delegated_source eq '') )
              && (-e $delegated_output) )
        {
	    $conversions{$delegated_output} = $delegated_source;
	    $dependents{$delegated_output} = 7;
            if ($delegated_source) {
                $dependents{$delegated_source} = 4;
	    }
	}
	elsif (!$silent) {
            print "Logfile claimed conversion from '$delegated_source' ",
        	  "to '$delegated_output'.  But:\n";
            if (! -e $delegated_output) {
		print  "   Output file does not exist\n";
	    }
            if ( ($delegated_source ne '') && (! -e $delegated_source) ) {
		print  "   Input file does not exist\n";
	    }
	}
    }

    
    if ( $diagnostics ) {
        @misparse = uniqs( @misparse );
        @missing = uniqs( @missing );
        @not_found = uniqs( @not_found );
        my @dependents = sort( keys %dependents );

        my $dependents = $#dependents + 1;
        my $misparse = $#misparse + 1;
        my $missing = $#missing + 1;
        my $not_found = $#not_found + 1;
        my $exist = $dependents - $not_found - $missing;
        my $bbl = $#bbl_files + 1;

        print "$dependents dependent files detected, of which ",
              "$exist exist, $not_found were not found,\n",
	      "   and $missing appear not to exist.\n";
        print "Dependents:\n";
        foreach (@dependents) { 
            print "   '$_' "; 
            if ( $dependents{$_} == 6 ) { print " written by (pdf)latex";}
	    print "\n";
            if ( $dependents{$_} == 7 ) { print " converted by (pdf)latex";}
	    print "\n";
        }
        if ($not_found > 0) {
	    print "Not found:\n";
	    foreach (@not_found) { print "   $_\n"; }
	}
        if ($missing > 0) {
	    print "Not existent:\n";
	    foreach (@missing) { print "   $_\n"; }
	}
        if ( $bbl > 0 ) {
            print "Input bbl files:\n";
            foreach (@bbl_files) { print "   $_\n"; }
        }

        if ( $misparse > 0 ) {
	    print "$misparse\n";
            print "Apparent input files apparently from misunderstood lines in .log file:\n";
            foreach ( @misparse ) { print "   $_\n"; }
        }
    }
    return 1;
} #END parse_logB

#************************************************************

sub parse_aux {
    #Usage: parse_aux( $aux_file, \@new_bib_files, \@new_aux_files )
    # Parse aux_file (recursively) for bib files.  
    # If can't open aux file, then
    #    Return 0 and leave @new_bib_files empty
    # Else set @new_bib_files from information in the aux files
    #    And:
    #    Return 1 if no problems
    #    Return 2 with @new_bib_files empty if there are no \bibdata
    #      lines. 
    #    Return 3 if I couldn't locate all the bib_files
    # Set @new_aux_files to aux files parsed

    my $aux_file = $_[0];
    local $Pbib_files = $_[1];
    local $Paux_files = $_[2];
   
    @$Pbib_files = ();
    @$Paux_files = ();

    parse_aux1( $aux_file );
    if ($#{$Paux_files} < 0) {
       return 0;
    }
    @$Pbib_files = uniqs( @$Pbib_files );

    if ( $#{$Pbib_files} == -1 ) {
        warn "$My_name: No .bib files listed in .aux file '$aux_file' \n",
        return 2;
    }
    my $bibret = &find_file_list1( $Pbib_files, $Pbib_files,
                                  '.bib', \@BIBINPUTS );
    @$Pbib_files = uniqs( @$Pbib_files );
    if ($bibret == 0) {
        warn "$My_name: Found bibliography file(s) [@$Pbib_files]\n" 
        unless $silent;
    }
    else {
        warn "$My_name: Failed to find one or more bibliography files ",
             "in [@$Pbib_files]\n";
        if ($force_mode) {
            warn "==== Force_mode is on, so I will continue.  ",
                 "But there may be problems ===\n";
        }
        else {
            #$failure = -1;
            #$failure_msg = 'Failed to find one or more bib files';
            #warn "$My_name: Failed to find one or more bib files\n";
        }
        return 3;
    }
    return 1;
} #END parse_aux

#************************************************************

sub parse_aux1
# Parse single aux file for bib files.  
# Usage: &parse_aux1( aux_file_name )
#   Append newly found bib_filenames in @$Pbib_files, already 
#        initialized/in use.
#   Append aux_file_name to @$Paux_files if aux file opened
#   Recursively check \@input aux files
#   Return 1 if success in opening $aux_file_name and parsing it
#   Return 0 if fail to open it
{
   my $aux_file = $_[0];
   my $aux_fh = new FileHandle;
   if (! open($aux_fh, $aux_file) ) { 
       warn "$My_name: Couldn't find aux file '$aux_file'\n";
       return 0; 
   }
   push @$Paux_files, $aux_file;
AUX_LINE:
   while (<$aux_fh>) {
      if ( /^\\bibdata\{(.*)\}/ ) { 
          # \\bibdata{comma_separated_list_of_bib_file_names}
          # (Without the '.bib' extension)
          push( @$Pbib_files, split /,/, $1 ); 
      }
      elsif ( /^\\\@input\{(.*)\}/ ) { 
          # \\@input{next_aux_file_name}
	  &parse_aux1( $1 );
      }
   }
   close($aux_fh);
   return 1;
} #END parse_aux1

#************************************************************

#************************************************************
#************************************************************
#************************************************************

#   Manipulations of main file database:

#************************************************************

sub fdb_get {
    # Call: fdb_get(filename [, runtime])
    # Returns an array (time, size, md5) for the current state of the
    #    named file.
    # The optional argument runtime is the runtime of some command
    #    associated with the use of the file --- see below.
    # For non-existent file, deletes its entry in fdb_current, 
    #    and returns (0,-1,0)  
    # As an optimization, the md5 value is taken from the cache in 
    #    fdb_current, if the time and size stamp indicate that the 
    #    file has not changed.
    # The md5 value is recalculated if
    #    the current filetime differs from the cached value: 
    #               file has been written
    #    the current filesize differs from the cached value: 
    #               file has definitely changed
    # But the file can also be rewritten without change in filetime when 
    #    file processing happens within the 1-second granularity of the 
    #    timestamp (notably for aux files from latex on a short source file).
    # The only case that concerns us is when the file is an input to a program
    #    at some runtime t, the file is rewritten later by the same or another
    #    program, with timestamp t, and when the initial file also has 
    #    timestamp t.
    # A test is applied for this situation if the runtime argument is 
    #    supplied and is nonzero.

    my ($file, $run_time) = @_;
    if ( ! defined $run_time ) { $run_time = 0;}
    my ($new_time, $new_size) = get_time_size($file);
    my @nofile =  (0,-1,0);     # What we use for initializing
				# a new entry in fdb or flagging
				# non-existent file
    if ( $new_size < 0 ) {
        delete $fdb_current{$file};
        return @nofile;
    }
    my $recalculate_md5 = 0;
    if ( ! exists $fdb_current{$file} ) {
        # Ensure we have a record.  
        $fdb_current{$file} = [@nofile];
        $recalculate_md5 = 1;
    }
    my $file_data = $fdb_current{$file};
    my ( $time, $size, $md5 ) = @$file_data;
#warn "--------- Getting MD5: $file: (N,O,R) = $new_time, $time, $run_time\n";
#warn "--------- $file: (OT,OS,OM) = @$file_data\n";
    if ( ($new_time != $time) || ($new_size != $size) 
         || ( $run_time && ($run_time == $time ) )
       ) {
        # Only force recalculation of md5 if time or size changed
        # Or if the time equals runtime, so that the file may
        #   have changed within the 1-second granularity of the time
        # Else we assume file is really unchanged.
        $recalculate_md5 = 1;
    }
    if ($recalculate_md5) {
#warn "--------- RECALC MD5: $rule $file: (N,O,R) = $new_time, $time, $run_time\n";
#warn "  ------- $file: (OT,OS,OM) = @$file_data\n";
#&traceback;
        @$file_data = ( $new_time, $new_size, get_checksum_md5( $file ) );
#warn "  ------- $file: (NT,NS,NM) = @$file_data\n";
    }
    return @$file_data;;
} #END fdb_get

#************************************************************

sub fdb_set {
    # Call: fdb_set(filename, $time, $size, $md5 )
    # Set data in file data cache, i.e., %fdb_current
    my ($file, $time, $size, $md5 ) = @_;
    if ( ! exists $fdb_current{$file} ) {
        $fdb_current{$file} = [0, -1, 0];
    }
    @{$fdb_current{$file}} = ( $time, $size, $md5 );
} #END fdb_set

#************************************************************

sub fdb_show {
    # Displays contents of fdb
    foreach my $file ( sort keys %fdb_current ) {
        print "'$file': @{$fdb_current{$file}}\n";
    }
} #END fdb_show

#************************************************************
#************************************************************
#************************************************************

# Routines for manipulating rule database

#************************************************************

sub rdb_read {
    # Call: rdb_read( $in_name  )
    # Sets rule database from saved file, in format written by rdb_write.
    # Returns -1 if file could not be read else number of errors.
    # Thus return value on success is 0
    my $in_name = $_[0];
    my $in_handle = new FileHandle;
    $in_handle->open( $in_name, '<' )
       or return ();
    my $errors = 0;
    my $state = 0;   # Outside a section 
    my $rule = '';
    my $run_time = 0;
    my $source = '';
    my $dest = '';
    my $base = '';
    local %new_sources = ();  # Hash: rule => { file=>[ time, size, md5, fromrule ] }
    my $new_source = undef;   # Reference to hash of sources for current rule
LINE:
    while ( <$in_handle> ) {
        # Remove leading and trailing white space.
        s/^\s*//;
        s/\s*$//;
        # Ignore blank lines and comments
        if ( /^$/ || /^#/ || /^%/ ) { next LINE;}
        if ( /^\[\"([^\"]+)\"\]/ ) {
            # Start of section
	    $rule = $1;
#??             print "--- Starting rule '$rule'\n";
            my $tail = $'; #'  Single quote in comment tricks the parser in
                           # emacs from misparsing an isolated single quote
            $run_time = 0;
            $source = $dest = $base = '';
            if ( $tail =~ /^\s*(\S+)\s*$/ ) {
                $run_time = $1;
	    }
            elsif ( $tail =~ /^\s*(\S+)\s+\"([^\"]*)\"\s+\"([^\"]*)\"\s+\"([^\"]*)\"\s*$/ ) {
                $run_time = $1;
                $source = $2;
                $dest = $3;
                $base = $4;
	    }
            if ( rdb_rule_exists( $rule ) ) {
                rdb_one_rule( $rule, 
                              sub{ $$Ptest_kind = 1; 
                                   $$Prun_time = $run_time;
                                   #??if ($source) { $$Psource = $source; }
                                   #??if ($dest) { $$Pdest = $dest; }
                                   #??if ($base) { $$Pbase = $base; }
                                 }                                      
                             );
	    }
            elsif ($rule =~ /^cusdep\s+(\S+)\s+(\S+)\s+(.+)$/ ) {
                # Create custom dependency
                my $fromext = $1;
                my $toext = $2;
                my $base = $3;
                $source = "$base.$fromext";
                $dest =   "$base.$toext";
                my $PAnew_cmd = ['do_cusdep', ''];
                foreach my $dep ( @cus_dep_list ) {
                    my ($tryfromext,$trytoext,$must,$func_name) = split(' ',$dep);
                    if ( ($tryfromext eq $fromext) && ($trytoext eq $toext) ) {
                       $$PAnew_cmd[1] = $func_name;
                    }
                }
                # Set source file as non-existent.  
                # If it existed on last run, it will be in later 
                #    lines of the fdb file
                rdb_create_rule( $rule, 'cusdep', '', $PAnew_cmd, 1, 
                                 $source, $dest, $base, 0, $run_time, 1 );
	    }
            elsif ( $rule =~ /^(makeindex|bibtex)\s*(.*)$/ ) {
                my $rule_generic = $1;
                if ( ! $source ) {
                    # If fdb_file was old-style (v. 1)
                    $source = $2;
                    my $path = '';
                    my $ext = '';
                    ($base, $path, $ext) = fileparseA( $source );
                    $base = $path.$base;
                    if ($rule_generic eq 'makeindex') {
                        $dest = "$base.ind";
		    }
                    elsif ($rule_generic eq 'bibtex') {
                        $dest = "$base.bbl";
                        $source = "$base.aux";
		    }
	        }
  	        warn "$My_name: File-database '$in_name': setting rule '$rule'\n"
                   if $diagnostics;
                my $cmd_type = 'external';
                my $ext_cmd = ${$rule_generic};
		warn "  Rule kind = '$rule_generic'; ext_cmd = '$ext_cmd';\n",
		     "  source = '$source'; dest = '$dest'; base = '$base';\n"
                   if $diagnostics;
                # Set source file as non-existent.  
                # If it existed on last run, it will be in later 
                #    lines of the fdb file
                rdb_create_rule( $rule, $cmd_type, $ext_cmd, '', 1, 
                                 $source, $dest, $base, 0, $run_time, 1);
	    }
            else {
  	        warn "$My_name: In file-database '$in_name' rule '$rule'\n",
                     "   is not in use in this session\n"
                if $diagnostics;
 	        $new_source = undef;
	        $state = 3;
	        next LINE;
	    }
            $new_source = $new_sources{$rule} = {};
	    $state = 1;  #Reading a section
	}
	elsif ( /^\"([^\"]*)\"\s+(\S+)\s+(\S+)\s+(\S+)\s+\"([^\"]*)\"/ ) {
            # Source file line
            if ($state == 3) {
                # The rule is not being currently used.
                next LINE;
            }
	    my $file = $1;
	    my $time = $2;
	    my $size = $3;
	    my $md5 = $4;
            my $from_rule = $5;
#??            print "  --- File '$file'\n";
            if ($state != 1) {
                warn "$My_name: In file-database '$in_name' ",
  		     "line $. is outside a section:\n   '$_'\n";
		$errors++;
		next LINE;
	    }
            # Set file in database.  But ensure we don't do an unnecessary 
            #    fdb_get, which can trigger a new MD5 calculation, which is
            #    lengthy for a big file.  Ininitially flagging the file
            #    as non-existent solves the problem:
	    rdb_ensure_file( $rule, $file, undef, 1 ); 
            rdb_set_file1( $rule, $file, $time, $size, $md5 );
            fdb_set( $file, $time, $size, $md5 );
            # Save the rest of the data, especially the from_fule until we know all 
            #   the rules, otherwise the from_rule may not exist.
            # Also we'll have a better chance of looping through files.
	    ${$new_source}{$file} = [ $time, $size, $md5, $from_rule ];
	}
	elsif ($state == 0) {
            # Outside a section.  Nothing to do.
	}
	else {
	    warn "$My_name: In file-database '$in_name' ",
                 "line $. is of wrong format:\n   '$_'\n";
	    $errors++;
	    next LINE;
	}
    }
    undef $in_handle;
    # Set cus dependencies.
    &rdb_set_dependentsA( keys %rule_db );

#?? Check from_rules exist.

    return $errors;
}  # END rdb_read

#************************************************************

sub rdb_read_generatedB {
    # Call: rdb_read_generatedB( $in_name, \@generated1, \@aux_files.
    #                            \%other_generated )
    # From rule database in saved file, in format written by rdb_write,
    # finds the non-trivial generated files that are to be deleted by a cleanup. 
    # On entry:
    #     Keys of %other_generated: any detected generated files, from \openout 
    #       lines log file, usually.
    # At exit:
    #    Array @generated1: files associated with makeindex and bibtex
    #    Array @aux_files: aux files
    #    Added to keys of %other_generated: any files that are destinations of rules
    #       whose source is in %other_generated.  (One stage only.)
    my ($in_name, $Pgenerated1, $Paux_files, $Pother_generated) = @_;
    @$Pgenerated1 = ();
    @$Paux_files = ();

    my $in_handle = new FileHandle;
    $in_handle->open( $in_name, '<' )
       or return ();
    my $rule = '';
    my $run_time = 0;
    my $source = '';
    my $dest = '';
    my $base = '';
    my $ext = '';
    my $path = '';
    my $state = 0;   # Outside a section 
    my @other_additions = ();
LINE:
    while ( <$in_handle> ) {
        # Remove leading and trailing white space.
        s/^\s*//;
        s/\s*$//;
        # Ignore blank lines and comments
        if ( /^$/ || /^#/ || /^%/ ) { next LINE;}
        if ( /^\[\"([^\"]+)\"\]/ ) {
            # Start of section
	    $rule = $1;
            my $tail = $'; #'  Single quote in comment tricks the parser in
                           # emacs from misparsing an isolated single quote
            $run_time = 0;
            $source = $dest = $base = '';
            if ( $tail =~ /^\s*(\S+)\s+\"([^\"]*)\"\s+\"([^\"]*)\"\s+\"([^\"]*)\"\s*$/ ) {
                $source = $2;
                $dest = $3;
                $base = $4;
	    }
            else { next LINE; }
            if ( $rule =~ /^makeindex/ ) {
                push @$Pgenerated1, $source, $dest, "$base.ilg";
	    }
            elsif ( $rule =~ /^bibtex/ ) {
                push @$Pgenerated1, $dest, "$base.blg";
                push @$Paux_files, $source;
	    }
            elsif ( exists ${$Pother_generated}{$source} ) {
                push @other_additions, $dest;
	    }
	    $state = 1;  #Reading a section
	}
	elsif ( /^\"([^\"]*)\"\s+(\S+)\s+(\S+)\s+(\S+)\s+\"([^\"]*)\"/ ) {
            # Source file line
            if ($state == 3) {
                # The rule is not being currently used.
                next LINE;
            }
	    my $file = $1;
            ($base, $path, $ext) = fileparseA( $file );
            if ( $ext eq '.aux' ) { push @$Paux_files, $file; }
	}
	elsif ($state == 0) {
            # Outside a section.  Nothing to do.
	}
	else {
	    warn "$My_name: In file-database '$in_name' ",
                 "line $. is of wrong format:\n   '$_'\n";
	    next LINE;
	}
    } # LINE
    undef $in_handle;
    foreach (@other_additions) {
	${$Pother_generated}{$_} = 1;
    }
}  # END rdb_read_generatedB

#************************************************************

sub rdb_write {
    # Call: rdb_write( $out_name )
    # Writes to the given file name the database of file and rule data
    #   for all rules needed to make final output
    # !!?? Previously was:
    # OLD Writes to the given file name the database of file and rule data
    # OLD   accessible from the primary rules.
    # Returns 1 on success, 0 if file couldn't be opened.
    local $out_name = $_[0];
    local $out_handle = new FileHandle;
    if ( ($out_name eq "") || ($out_name eq "-") ) {
        # Open STDOUT
        $out_handle->open( '>-' );
    }
    else {
       $out_handle->open( $out_name, '>' );
    }
    if (!$out_handle) { return 0; }

    local %current_primaries = ();   # Hash whose keys are primary rules 
                # needed, i.e., known latex-like rules which trigger
                # circular dependencies
    local @pre_primary = ();   # Array of rules
    local @post_primary = ();  # Array of rules
    local @one_time = ();      # Array of rules
    &rdb_classify_rules( \%possible_primaries, keys %requested_filerules );

    print $out_handle "# Fdb version $fdb_ver\n";
# !!??   Rules or rules accessible from primary
#    my @rules = rdb_accessible( uniq1( keys %possible_primaries )  ) ;
    my @rules = rdb_accessible( uniq1( keys %possible_primaries, keys %requested_filerules  )  ) ;
    # Separate call to sort.  Otherwise rdb_accessible seems to get wrong argument.
    @rules = sort( @rules );
    rdb_for_some(
       \@rules,
       sub { 
           # Omit data on a unused and never-run primary rule:
	   if ( ($$Prun_time == 0) 
                && exists( $possible_primaries{$rule} )
                && ! exists( $current_primaries{$rule} )
              )
           { 
               return; 
           }
           print $out_handle "[\"$rule\"] $$Prun_time \"$$Psource\" \"$$Pdest\" \"$$Pbase\" \n"; 
           rdb_do_files(
             sub { print $out_handle "  \"$file\" $$Ptime $$Psize $$Pmd5 \"$$Pfrom_rule\"\n"; }
	   );
       }
    );
    undef $out_handle;
    return 1;
} #END rdb_write

#************************************************************

sub rdb_set_from_logB {
    # Assume rule context.  
    # This is intended to be applied only for a primary (LaTeX-like) rule
    # Starting from the log_file, set current details for the current rule.

    # Rules should only be primary
    if ( $$Pcmd_type ne 'primary' ) {
	warn "\n$My_name: ==========$My_name: Probable BUG======= \n   ",
             "   rdb_set_from_logB called to set files ",
             "for non-primary rule '$rule'\n\n";
        return;
    }


#??    # We'll prune this by all files determined to be needed for source files.
#??    my %unneeded_source = %$PHsource;

    # Parse log file to find relevant filenames
    # Result in the following variables:
    local %dependents = ();    # Maps files to status
    local @bbl_files = ();
    local %idx_files = ();     # Maps idx_file to (ind_file, base)
    local %generated_log = (); # Lists generated files found in log file
    local $primary_out = $$Pdest;  # output file (dvi or pdf)
    local %conversions = ();   # (pdf)latex-performed conversions.
                     # Maps output file created and read by (pdf)latex
                     #    to source file of conversion.
    # The following are also returned, but are global, to be used by caller
    # $reference_changed, $bad_reference $bad_citation

    &parse_logB;

    # Handle result on output file:
    #   1.  Non-existent output file, which is because of no content.
    #         This could either be because the source file has genuinely
    #         no content, or because of a missing input file.  Since a
    #         missing input file might be correctable by a run of some
    #         other program whose running is provoked AFTER a run of
    #         (pdf)latex, we'll set a diagnostic and leave it to the
    #         rdb_makeB to handle after all circular dependencies are
    #         resolved. 
    #   2.  The output file might be of a different kind than expected
    #         (i.e., dvi instead of pdf, or vv).  This could
    #         legitimately occur when the source file (or an invoked
    #         package or class) sets \pdfoutput. 
    $missing_dvi_pdf = ''; 
    if ($primary_out eq '')  {
        warn "$My_name: For rule '$rule', no output was made\n";
        $missing_dvi_pdf = $$Pdest;
    }
    elsif ($primary_out ne $$Pdest) {
        warn "$My_name: ===For rule '$rule', actual output '$primary_out'\n",
             "       ======appears not to match expected output '$$Pdest'\n";
    }

  IDX_FILE:
    foreach my $idx_file ( keys %idx_files ) {
	my ($ind_file, $ind_base) = @{$idx_files{$idx_file}};
        my $from_rule = "makeindex $idx_file";
        if ( ! rdb_rule_exists( $from_rule ) ){
            print "!!!===Creating rule '$from_rule': '$ind_file' from '$idx_file'\n"
                  if ($diagnostics);
            rdb_create_rule( $from_rule, 'external', $makeindex, '', 1, 
                             $idx_file, $ind_file, $ind_base, 1, 0);
            print "  ===Source file '$ind_file' for '$rule'\n"
                  if ($diagnostics > -1);
            rdb_ensure_file( $rule, $ind_file, $from_rule );
	}
        if ( ! -e $ind_file ) { 
            # Failure was non-existence of makable file
            # Leave failure issue to other rules.
	    $failure = 0;
	}
    }

  BBL_FILE:
    foreach my $bbl_file ( uniqs( @bbl_files ) ) {
	my ($bbl_base, $bbl_path, $bbl_ext) = fileparseA( $bbl_file );
        $bbl_base = $bbl_path.$bbl_base;
        my @new_bib_files;
        my @new_aux_files;
        &parse_aux( "$bbl_base.aux", \@new_bib_files, \@new_aux_files );
        my $from_rule = "bibtex $bbl_base";
        print "!!!===Dealing with rule '$from_rule'\n"
            if ($diagnostics);
        if ( ! rdb_rule_exists( $from_rule ) ){
            print "   ===Creating rule '$from_rule'\n"
              if ($diagnostics);
            rdb_create_rule( $from_rule, 'external', $bibtex, '', 1, 
                             "$bbl_base.aux", $bbl_file, $bbl_base, 1, 0);
	}
        local %old_sources = ();
        rdb_one_rule( $from_rule, sub { %old_sources = %$PHsource; } );
        foreach my $source ( @new_bib_files, @new_aux_files ) {
            print "  === Source file '$source' for '$from_rule'\n"
               if ($diagnostics);
            rdb_ensure_file( $from_rule, $source );
            delete $old_sources{$source};
        }
        if ($diagnostics) {
            foreach ( keys %old_sources ) {
                print "Removing no-longer-needed dependent '$_' from rule '$from_rule'\n";
            }
        }        
        rdb_remove_files( $from_rule, keys %old_sources );
        print "  ===Source file '$bbl_file' for '$rule'\n"
            if ($diagnostics);
        rdb_ensure_file( $rule, $bbl_file, $from_rule );
        if ( ! -e $bbl_file ) { 
            # Failure was non-existence of makable file
            # Leave failure issue to other rules.
            $failure = 0;
	}
    }

NEW_SOURCE:
    foreach my $new_source (keys %dependents, keys %conversions) {
        print "  ===Source file for rule '$rule': '$new_source'\n"
	    if ($diagnostics);
	if ( ($dependents{$new_source} == 5) 
             || ($dependents{$new_source} == 6) ) {
            # (a) File was detected in "No file..." line in log file. 
            #     Typically file was searched for early in run of 
            #     latex/pdflatex, was not found, and then was written 
            #     later in run.
            # or (b) File was written during run. 
            # In both cases, if file doesn't already exist in database, we 
            #    don't know its previous status.  Therefore we tell 
            #    rdb_ensure_file that if it needs to add the file to its 
            #    database, then the previous version of the file should be 
            #    treated as non-existent, to ensure another run is forced.
            rdb_ensure_file( $rule, $new_source, undef, 1 );
	}
	elsif ( $dependents{$new_source} == 7 )  {
            # File was result of conversion by (pdf)latex.
            my $cnv_source = $conversions{$new_source};
            rdb_ensure_file( $rule, $new_source );
	    if ($cnv_source) {
                # Conversion from $cnv_source to $new_source
                #   implies that effectively $cnv_source is a source
                #   of the (pdf)latex run.
                rdb_ensure_file( $rule, $cnv_source );
	    }
            # Flag that changes of the generated file during a run 
            #    do not require a rerun:
            rdb_one_file( $new_source, sub{ $$Pcorrect_after_primary = 1; } );
	}
	else {
            # But we don't need special precautions for ordinary user files 
            #    (or for files that are generated outside of latex/pdflatex). 
            rdb_ensure_file( $rule, $new_source );
	}
    }

    my @more_sources = &rdb_set_dependentsA( $rule );
    my $num_new = $#more_sources + 1;
    foreach (@more_sources) { 
	$dependents{$_} = 4;
        if ( ! -e $_ ) { 
            # Failure was non-existence of makable file
            # Leave failure issue to other rules.
            $failure = 0; 
            $$Pchanged = 1; # New files can be made.  Ignore error.
        }
    }
    if ($diagnostics) {
	if ($num_new > 0 ) {
	    print "$num_new new source files for rule '$rule':\n";
	    foreach (@more_sources) { print "   '$_'\n"; }
	}
	else {
	    print "No new source files for rule '$rule':\n";
	}
    }

    my @files_not_needed = ();
    foreach (keys %$PHsource) { 
        if ( ! exists $dependents{$_} ) {
            print "Removing no-longer-needed dependent '$_' from rule '$rule'\n"
              if $diagnostics;
            push @files_not_needed, $_;
	}
    }
    rdb_remove_files( $rule, @files_not_needed );

} # END rdb_set_from_logB

#************************************************************

sub rdb_find_new_filesB {
    # Call: rdb_find_new_filesB
    # Assumes rule context for primary rule.
    # Deal with files which were missing and for which a method
    # of finding them has become available:
    #   (a) A newly available source file for a custom dependency.
    #   (b) When there was no extension, a file with appropriate
    #       extension
    #   (c) When there was no extension, and a newly available source 
    #       file for a custom dependency can make it.

    my %new_includes = ();

MISSING_FILE:
    foreach my $missing ( keys %$PHsource ) {
        next if ( $$PHsource{$missing} != 0 ); 
        my ($base, $path, $ext) = fileparseA( $missing );
        $ext =~ s/^\.//;
        if ( -e "$missing.tex" ) { 
            $new_includes{"$missing.tex"} = 1;
        }
        if ( -e $missing ) { 
	    $new_includes{$missing} = 1;
        }
        if ( $ext ne "" ) {
            foreach my $dep (@cus_dep_list){
               my ($fromext,$toext) = split(' ',$dep);
               if ( ( "$ext" eq "$toext" )
                    && ( -e "$path$base.$fromext" )
	 	  )  {
                  # Source file for the missing file exists
                  # So we have a real include file, and it will be made
                  # next time by rdb_set_dependents
                  $new_includes{$missing} = 1;
               }
	       else {
                   # no point testing the $toext if the file doesn't exist.
	       }
               next MISSING_FILE;
	    }
       }
       else {
           # $_ doesn't exist, $_.tex doesn't exist,
           # and $_ doesn't have an extension
           foreach my $dep (@cus_dep_list){
              my ($fromext,$toext) = split(' ',$dep);
              if ( -e "$path$base.$fromext" ) {
                  # Source file for the missing file exists
                  # So we have a real include file, and it will be made
                  # next time by &rdb__dependents
                  $new_includes{"$path$base.$toext"} = 1;
#                  next MISSING_FILE;
              }
              if ( -e "$path$base.$toext" ) {
                  # We've found the extension for the missing file,
                  # and the file exists
                  $new_includes{"$path$base.$toext"} = 1;
#                  next MISSING_FILE;
              }
	   }
       }
    } # end MISSING_FILES

    # Sometimes bad line-breaks in log file (etc) create the
    # impression of a missing file e.g., ./file, but with an incorrect
    # extension.  The above tests find the file with an extension,
    # e.g., ./file.tex, but it is already in the list.  So now I will
    # remove files in the new_include list that are already in the
    # include list.  Also handle aliasing of file.tex and ./file.tex.
    # For example, I once found:
# (./qcdbook.aux (./to-do.aux) (./ideas.aux) (./intro.aux) (./why.aux) (./basics
#.aux) (./classics.aux)

    my $found = 0;
    foreach my $file (keys %new_includes) {
        my $stripped = $file;
        $stripped =~ s{^\./}{};
        if ( exists $PHsource{$file} ) {
	    delete $new_includes{$file};
	}
        else {
	    $found ++;
	    rdb_ensure_file( $rule, $file );
	}
    }

    if ( $diagnostics && ( $found > 0 ) ) {
	warn "$My_name: Detected previously missing files:\n";
        foreach ( sort keys %new_includes ) {
            warn "   '$_'\n";
	}
    }
    return $found;
} # END rdb_find_new_filesB

#************************************************************

sub rdb_set_dependentsA {
    # Call rdb_set_dependentsA( rules ...)
    # Returns array (sorted), of new source files.
    local @new_sources = ();
    local @deletions = ();

# Shouldn't recurse.  The definite rules to be examined are given.
    rdb_for_some( [@_],  0, \&rdb_one_depA );
# OLD    rdb_recurseA( [@_],  0, \&rdb_one_depA );
    foreach (@deletions) {
	my ($rule, $file) = @$_;
        rdb_remove_files( $rule, $file );
    }
    &rdb_make_links;
    return uniqs( @new_sources );
} #END rdb_set_dependentsA

#************************************************************

sub rdb_one_depA {
    # Helper for finding dependencies.  One case, $rule and $file given
    # Assume file (and rule) context for DESTINATION file.
    local $new_dest = $file;
    my ($base_name, $path, $toext) = fileparseA( $new_dest );
    $base_name = $path.$base_name;
    $toext =~ s/^\.//;
    my $Pinput_extensions = $input_extensions{$rule};
DEP:
    foreach my $dep ( @cus_dep_list ) {
        my ($fromext,$proptoext,$must,$func_name) = split(' ',$dep);
        if ( $toext eq $proptoext ) {
            my $source = "$base_name.$fromext";
	    # Found match of rule
            if ($diagnostics) {
                print "Found cusdep:  $source to make $rule:$new_dest ====\n";
            }
	    if ( -e $source ) {
	        $$Pfrom_rule = "cusdep $fromext $toext $base_name";
#??		print "?? Ensuring rule for '$$Pfrom_rule'\n";
                local @PAnew_cmd = ( 'do_cusdep', $func_name );
                if ( !-e $new_dest ) {
		    push @new_sources, $new_dest;
		}
                if (! rdb_rule_exists( $$Pfrom_rule ) ) {
                    print "=== Creating rule for '$$Pfrom_rule'\n";
                    rdb_create_rule( $$Pfrom_rule, 'cusdep', '', \@PAnew_cmd, 3, 
                                     $source, $new_dest, $base_name, 0 );
		}
                else {
		    rdb_one_rule( 
                       $$Pfrom_rule, 
                       sub{ @$PAint_cmd = @PAnew_cmd; $$Pdest = $new_dest;}
                    );
		}
		return;
	    }
            else {
                # Source file does not exist
                if ( !$force_mode && ( $must != 0 ) ) {
                    # But it is required that the source exist ($must !=0)
                    $failure = 1;
                    $failure_msg = "File '$base_name.$fromext' does not exist ".
                                   "to build '$base_name.$toext'";
                    return;
		}
                elsif ( $$Pfrom_rule =~ /^cusdep $fromext $toext / )  {
                    # Source file does not exist, destination has the rule set.
                    # So turn the from_rule off
		    $$Pfrom_rule = '';
		}
		else {
		}
	    }
	}
        elsif ( ($toext eq '') 
                && (! -e $file ) 
                && (! -e "$base_name.$proptoext" ) 
                && exists $$Pinput_extensions{$proptoext}
              ) {
            # Empty extension and non-existent destination
            #   This normally results from  \includegraphics{A}
            #    without graphics extension for file, when file does
            #    not exist.  So we will try to find something to make it.
            my $source = "$base_name.$fromext";
            if ( -e $source ) {
                $new_dest = "$base_name.$proptoext";
	        my $from_rule = "cusdep $fromext $proptoext $base_name";
                push @new_sources, $new_dest;
		print "Ensuring rule for '$from_rule', to make '$new_dest'\n"
		    if $diagnostics > -1;
                local @PAnew_cmd = ( 'do_cusdep', $func_name );
                if (! rdb_rule_exists( $from_rule ) ) {
                    rdb_create_rule( $from_rule, 'cusdep', '', \@PAnew_cmd, 3, 
                                     $source, $new_dest, $base_name, 0);
		}
                else {
		    rdb_one_rule( 
                       $$Pfrom_rule, 
                       sub{ @$PAint_cmd = @PAnew_cmd; $$Pdest = $new_dest;}
                    );
		}
                rdb_ensure_file( $rule, $new_dest, $from_rule );
                # We've now got a spurious file in our rule.  But don't mess
                # with deleting an item we are in the middle of!
                push @deletions, [$rule, $file];
		return;
	    }
        } # End of Rule found
    } # End DEP
} #END rdb_one_depA

#************************************************************

sub rdb_list {
    # Call: rdb_list()
    # List rules and their source files
    print "===Rules:\n";
    local $count_rules = 0;
    my @accessible_all = rdb_accessible( keys %requested_filerules ); 
    rdb_for_some( 
        \@accessible_all,
	sub{ $count_rules++; 
             print "Rule '$rule' depends on:\n"; 
           },
	sub{ print "    '$file'\n"; }
    );
    if ($count_rules <= 0) {
	print "   ---No rules defined\n";
    }
} #END rdb_list

#************************************************************

sub rdb_show {
    # Call: rdb_show()
    # Displays contents of rule data base.
    # Side effect: Exercises access routines!
    print "===Rules:\n";
    local $count_rules = 0;
    rdb_for_all( 
	sub{ $count_rules++; 
             my @int_cmd = @$PAint_cmd;
	     foreach (@int_cmd) {
		 if ( !defined($_) ) { $_='undef';}
	     }
             print "  [$rule]: '$$Pcmd_type' '$$Pext_cmd' '@int_cmd' $$Ptest_kind ",
                   "'$$Psource' '$$Pdest' '$$Pbase' $$Pout_of_date $$Pout_of_date_user\n"; },
	sub{ print "    '$file': $$Ptime $$Psize $$Pmd5 '$$Pfrom_rule'\n"; }
    );
    if ($count_rules <= 0) {
	print "   ---No rules defined\n";
    }
} #END rdb_show

#************************************************************

sub rdb_accessible {
    # Call: rdb_accessible( rule, ...)
    # Returns array of rules accessible from the given rules
    local @accessible = ();
    rdb_recurseA( [@_], sub{ push @accessible, $rule; } );
    return @accessible;
} #END rdb_accessible

#************************************************************
#************************************************************
#************************************************************

sub rdb_makeB {
    # Call: rdb_makeB( target, ... )
    # Makes the targets and prerequisites.  
    # Leaves one-time rules to last.
    # Does appropriate repeated makes to resolve dependency loops

    # Returns 0 on success, nonzero on failure.

    # General method: Find all accessible rules, then repeatedly make
    # them until all accessible rules are up-to-date and the source
    # files are unchanged between runs.  On termination, all
    # accessible rules have stable source files.
    #
    # One-time rules are view and print rules that should not be
    # repeated in an algorithm that repeats rules until the source
    # files are stable.  It is the calling routine's responsibility to
    # arrange to call them, or to use them here with caution.
    #
    # Note that an update-viewer rule need not be considered
    # one-time.  It can be legitimately applied everytime the viewed
    # file changes.
    #
    # Note also that the criterion of stability is to be applied to
    # source files, not to output files.  Repeated application of a
    # rule to IDENTICALLY CONSTANT source files may produce different
    # output files.  This may be for a trivial reason (e.g., the
    # output file contains a time stamp, as in the header comments for
    # a typical postscript file), or for a non-trivial reason (e.g., a
    # stochastic algorithm, as in abcm2ps).   
    #
    # This caused me some actual trouble.  In general, circular
    # dependencies produce non-termination, and the the following
    # situation is an example of a generic situation where certain
    # rules must be obeyed in order to obtain proper results:
    #    1.  A/the latex source file contains specifications for
    #        certain postprocessing operations.  Standard (pdf)latex 
    #        already has this, for indexing and bibliography.
    #    2.  In the case in point that caused me trouble, the
    #        specification was for musical tunes that were contained
    #        in external source files not directly input to
    #        (pdf)latex.  But in the original version, there was a
    #        style file (abc.sty) that caused latex itself to call
    #        abcm2ps to make .eps files for each tune that were to be
    #        read in on the next run of latex. 
    #    3.  Thus the specification can cause a non-terminating loop
    #        for latexmk, because the output files of abcm2ps changed
    #        even with identical input.  
    #    4.  The solution was to 
    #        a. Use a style file abc_get.sty that simply wrote the
    #           specification on the tunes to the .aux file in a
    #           completely deterministic fashion.
    #        b. Instead of latex, use a script abclatex.pl that runs
    #           latex and then extracts the abc contents for each tune
    #           from the source abc file.  This is also
    #           deterministic. 
    #        c. Use a cusdep rule in latexmk to convert the tune abc
    #           files to eps.  This is non-deterministic, but only
    #           gets called when the (deterministic) source file
    #           changes.
    #        This solves the problem.  Latexmk works.  Also, it is no
    #        longer necessary to enable write18 in latex, and multiple
    #        unnecessary runs of abcm2ps are no longer used. 
    #
    # The order of testing and applying rules is chosen by the
    # following heuristics: 
    #    1.  Both latex and pdflatex may be used, but the resulting
    #        aux files etc may not be completely identical.  Define
    #        latex and pdflatex as primary rules.  Apply the general
    #        method of repeated circulating through all rules until
    #        the source files are stable for each primary rule
    #        separately.  Naturally the rules are all accessible
    #        rules, but excluding primary rules except for the current
    #        primary.
    #    2.  Assume that the primary rules are relatively
    #        time-consuming, so that unnecessary passes through them
    #        to check stability of the source files should be avoided.
    #    3.  Assume that although circular dependencies exist, the
    #        rules can nevertheless be thought of as basically
    #        non-circular, and that many rules are strictly or
    #        normally non-circular.  In particular cusdep rules are
    #        typically non-circular (e.g., fig2eps), as are normal
    #        output processing rules like dvi2ps.  
    #    4.  The order for the non-circular approximation is
    #        determined by applying the assumption that an output file
    #        from one rule that is read in for an earlier stage is
    #        unchanged. 
    #    HOWEVER, at a first attempt, the ordering is not needed.  It
    #    only gives an optimization
    #    5.  (Note that these assumptions could be violated, e.g., if
    #        $dvips is arranged not only to do the basic dvips
    #        command, but also to extract information from the ps file
    #        and feed it back to an input file for (pdf)latex.)
    #    6.  Nevertheless, the overall algorithm should allow
    #        circularities.  Then the general criterion of stability
    #        of source files covers the general case, and also
    #        robustly handles the case that the USER changes source
    #        files during a run.  This is particularly important in
    #        -pvc mode, given that a full make on a large document can
    #        be quite lengthy in time, and moreover that a user
    #        naturally wishes to make corrections in response to
    #        errors, particularly latex errors, and have them apply
    #        right away.
    # This leads to the following approach:
    #    1.  Classify accessible rules as: primary, pre-primary
    #        (typically cusdep, bibtex, makeindex, etc), post-primary
    #        (typically dvips, etc), and one-time
    #    2.  Then stratify the rules into an order of application that
    #        corresponds to the basic feedforward structure, with the
    #        exclusion of one-time rules.
    #    3.  Always require that one-time rules are among the
    #        explicitly requested rules, i.e., the last to be applied,
    #        were we to apply them.  Anything else would not match the
    #        idea of a one-time rule.  
    #    4.  Then work as follows:
    #        a. Loop over primaries
    #        b. For each primary, examine each pre-primary rule and
    #           apply if needed, then the primary rule and then each
    #           post-primary rule.  The ordering of the pre-primary
    #           and post-primary rules was found in step 2.
    #      BUT applying the ordering is not essential
    #        c. Any time that a pre-primary or primary rule is
    #           applied, loop back to the beginning of step b.  This
    #           ensures that bibtex etc are applied before rerunning
    #           (pdf)latex, and also covers changing source files, and
    #           gives priority to quick pre-primary rules for changing
    #           source files against slow reruns of latex.
    #        d. Then apply post-primary rules in order, but not
    #           looping back after each rule.  This non-looping back
    #           is because the rules are normally feed-forward only.
    #      BUT applying the ordering is not essential
    #        e. But after completing post-primary rules do loop back
    #           to b if any rules were applied.  This covers exotic
    #           circular dependence (and as a byproduct, changing
    #           source files).
    #        f. On each case of looping back to b, re-evaluate the
    #           dependence setup to allow for the effect of changing
    #           source files.  
    #    

    local @requested_targets = @_;
    local %current_primaries = ();   # Hash whose keys are primary rules 
                # needed, i.e., known latex-like rules which trigger
                # circular dependencies
    local @pre_primary = ();   # Array of rules
    local @post_primary = ();  # Array of rules
    local @one_time = ();      # Array of rules


    # For diagnostics on changed files, etc:
    local @changed = ();
    local @disappeared = ();
    local @no_dest = ();       # Non-existent destination files
    local @rules_never_run = ();
    local @rules_to_apply = ();

    &rdb_classify_rules( \%possible_primaries, @requested_targets );

    local %pass = ();
    local $failure = 0;        # General accumulated error flag
    local $missing_dvi_pdf = ''; # Did primary run fail to make its output file? 
    local $runs = 0;
    local $too_many_runs = 0;
    local %rules_applied = ();
    my $retry_msg = 0;         # Did I earlier say I was going to attempt 
                               # another pass after a failure?
  PRIMARY:
    foreach my $primary (keys %current_primaries ) {
        foreach my $rule (keys %rule_db) { 
            $pass{$rule} = 0; 
        }
      PASS:
        while (1==1) {
            $runs = 0;
            my $previous_failure = $failure;
            $failure = 0;
            local $newrule_nofile = 0;  # Flags whether rule created for
                           # making currently non-existent file, which
                           # could become a needed source file for a run
                           # and therefore undo an error condition
	    if ($diagnostics) {
                print "MakeB: doing pre_primary and primary...\n";
            }
            rdb_for_some( [@pre_primary, $primary], \&rdb_makeB1 );
            if ( ($runs > 0) && ! $too_many_runs ) {
                $retry_msg = 0;
                if ( $failure && $newrule_nofile ) { 
                    $retry_msg = 1;
                    print "$My_name: Error on run, but found possibility to ",
                          "make new source files\n";
                    next PASS;
		}
                elsif ( $force_mode || (! $failure) ) {
                    next PASS;
		}
            }
            elsif ($runs == 0) {
                # $failure not set on this pass, so use value from previous pass:
                $failure = $previous_failure;
                if ($retry_msg) {
                    print "But in fact no new files made\n";
		}
	    }
            if ( $missing_dvi_pdf ) { 
               # No output from primary, after completing circular dependence
               warn "Failure to make '$missing_dvi_pdf'\n";
               $failure = 1; 
               last PASS;
            }    
            if ($failure && !$force_mode ) { last PASS; }
	    if ($diagnostics) {
  	        print "MakeB: doing post_primary...\n";
	    }
            rdb_for_some( [@post_primary], \&rdb_makeB1 );
            if ($failure) { last PASS; }
            if ($runs > 0) { next PASS; }
            # Get here if nothing was run.
            last PASS;
	}
	continue {
            # Re-evaluate rule classification and accessibility,
            # but do not change primaries.
            # Problem is that %current_primaries gets altered
            my %old_curr_prim = %current_primaries;
            &rdb_classify_rules( \%possible_primaries, @requested_targets );
            %current_primaries = %old_curr_prim;
            &rdb_make_links;
	}
    }
    rdb_for_some( [@one_time], \&rdb_makeB1 );
    rdb_write( $fdb_file );

    if (! $silent) { 
        # Diagnose of the runs
        if ( $#{keys %rules_applied }  > -1 ) {
            print "$My_name: $runs runs.  Rules applied:\n";
            foreach (sort keys %rules_applied) {
		print "    '$_'\n";
	    }
	}
	elsif ($failure && $force_mode) {
            print "$My_name: Errors, in force_mode: so I tried finishing targets\n";
	}
	elsif ($failure) {
            print "$My_name: Errors, so I did not complete making targets\n";
	}
	else {
            local @dests = ();
            rdb_for_some( [@_], sub{ push @dests, $$Pdest if ($$Pdest); } );
            print "$My_name: All targets (@dests) are up-to-date\n";
	}
    }
    return $failure;
} #END rdb_makeB

#-------------------

sub rdb_show_rule_errors {
    local @messages = ();
    rdb_for_all( 
	       sub{
		   if ($$Plast_message ne '') {
		       push @messages, "$rule: $$Plast_message";
		   }
		   elsif ($$Plast_result == 1) {
		       push @messages, "$rule: failed to create output file";
		   }
		   elsif ($$Plast_result == 2) {
		       push @messages, "$rule: gave an error";
		   }
		   elsif ($$Prun_time == 0) {
#    This can have innocuous causes.  So don't report
#		       push @messages, "$rule: never run";
		   }
	       }
	      );
    if ($#messages > -1) { 
	warn "Collected error summary (may duplicate other messages):\n";
	foreach (@messages){
	    warn "  $_\n";
	}
    }
    return $#messages+1;
}

#-------------------

sub rdb_makeB1 {
    # Call: rdb_makeB1
    # Helper routine for rdb_makeB.
    # Carries out make at level of given rule (all data available).
    # Assumes contexts for recursion, make, and rule, and
    # assumes that source files for the rule are to be considered
    # up-to-date. 
    if ($diagnostics) { print "  MakeB1 $rule\n"; }
    if ($failure & ! $force_mode) {return;}
    if ( ! defined $pass{$rule} ) {$pass{$rule} = 0; } 
    &rdb_clear_change_record;

    # Special fix up for bibtex:
    my $bibtex_not_run = -1;   # Flags status as to whether this is a
        # bibtex rule and if it is, whether out-of-date condition is to
        # be ignored.
        #  -1 => not a bibtex rule
        #   0 => no special treatment
        #   1 => don't run bibtex because of non-existent bibfiles
        #           (and setting to do this test)
        #   2 => don't run bibtex because of setting
    my @missing_bib_files = ();
    if ( $rule =~ /^bibtex/ ) {
        $bibtex_not_run = 0;
        if ($bibtex_use == 0) {
           $bibtex_not_run = 2;
	}
        elsif ($bibtex_use == 1) {
	    foreach ( keys %$PHsource ) {
		if ( ( /\.bib$/ ) && (! -e $_) ) {
                    push @missing_bib_files, $_;
                    $bibtex_not_run = 1;
		}
	    }
	}
    }

    if ( ($$Prun_time == 0) && exists($possible_primaries{$rule}) ) {
        push @rules_never_run, $rule;
        $$Pout_of_date = 1;
        $$Plast_result = -1;
    }
    else {
        if ( $$Pdest && (! -e $$Pdest) ) {
	    # With a non-existent destination, if we haven't made any passes
            #   through a rule, rerunning the rule is good, because the file
            #   may fail to exist because of being deleted by the user (for ex.)
            #   rather than because of a failure on a previous run. 
            # (We could do better with a flag in fdb file.)
	    # But after the first pass, the situation is different.  
	    #   For a primary rule (pdf)latex, the lack of a destination file 
	    #      could result from there being zero content due to a missing
	    #      essential input file.  The input file could be generated 
	    #      by a program to be run later (e.g., a cusdep or bibtex), 
	    #      so we should wait until all passes are completed before 
            #      deciding a non-existent destination file is an error.
	    #   For a custom dependency, the rule may be obsolete, and
	    #      if the source file does not exist also, we should simply
	    #      not run the rule, but not set an error condition.
	    #      Any error will arise at the (pdf)latex level due to a 
	    #      missing source file at that level.
	    if ( ( $$Pcmd_type eq 'cusdep') && $$Psource && (! -e $$Psource) ) {
		# No action
	    }
	    elsif ( $$Pcmd_type eq 'delegated' ) {
		# Delegate to destination rule
	    }
	    elsif ( $pass{$rule}==0) {
		push @no_dest, $$Pdest;
		$$Pout_of_date = 1;
	    }
	    if ( $$Pcmd_type eq 'primary' ) {
		$missing_dvi_pdf = $$Pdest;
	    }
	}
    }

    &rdb_flag_changes_here;

    if (!$$Pout_of_date) {
#??	if ( ($$Pcmd_type eq 'primary') && (! $silent) ) {
#            print "Rule '$rule' up to date\n";
#        }
	return;
    }
    if ($diagnostics) { print "     remake\n"; }
    if (!$silent) { 
        print "$My_name: applying rule '$rule'...\n"; 
        &rdb_diagnose_changes( "Rule '$rule': " );
    }

    $rules_applied{$rule} = 1;
    $runs++;

    # We are applying the rule, so its source file state for when it
    # was last made is as of now:
    # ??IS IT CORRECT TO DO NOTHING IN CURRENT VERSION?

    # The actual run
    my $return = 0;   # Return code from called routine
    # Rule may have been created since last run:
    if ( ! defined $pass{$rule} ) {$pass{$rule} = 0; }
    if ( $pass{$rule} ge $max_repeat ) {
        # Avoid infinite loop by having a maximum repeat count
        # Getting here represents some kind of weird error.
        warn "$My_name: Maximum runs of $rule reached ",
             "without getting stable files\n";
        $too_many_runs = 1;
        $failure = 1;
	$failure_msg = "'$rule' needed too many passes";
        return;
    }
    $pass{$rule}++; 
    if ($bibtex_not_run > 0) {
        if ($bibtex_not_run == 1 ) {
            show_array ("$My_name: I WON'T RUN '$rule' because I don't find the following files:",
                        @missing_bib_files);
        }
        elsif ($bibtex_not_run == 2 ) {
            warn "$My_name: I AM CONFIGURED/INVOKED NOT TO RUN '$rule'\n"; 
	}
        $return = &rdb_dummy_run1;
    }
    else {
        warn_running( "Run number $pass{$rule} of rule '$rule'" );
        if ($$Pcmd_type eq 'primary' ) { 
            $return = &rdb_primary_run; 
        }
        else { $return = &rdb_run1; }
    }
    if ($$Pchanged) {
        $newrule_nofile = 1;
        $return = 0;
    }
    elsif ( $$Pdest && ( !-e $$Pdest ) && (! $failure) ){
        # If there is a destination to make, but for some reason
        #    it did not get made, and no other error was reported, 
        #    then a priori there appears to be an error condition:
        #    the run failed.   But there are two important cases in
        #    which this is a wrong diagnosis.
        if ( ( $$Pcmd_type eq 'cusdep') && $$Psource && (! -e $$Psource) ) {
            # However, if the rule is a custom dependency, this is not by
            #  itself an error, if also the source file does not exist.  In 
            #  that case, we may have the situation that (1) the dest file is no
            #  longer needed by the tex file, and (2) therefore the user
            #  has deleted the source and dest files.  After the next
            #  latex run and the consequent analysis of the log file, the
            #  cusdep rule will no longer be needed, and will be removed.

            # So in this case, do NOT report an error          
            $$Pout_of_date = 0;
	}
        elsif ($$Pcmd_type eq 'primary' ) { 
            # For a primary rule, i.e., (pdf)latex, not to produce the 
            #    expected output file may not be an error condition.  
            # Diagnostics were handled in parsing the log file.
            # Special action in main loop inrdb_makeB1
            $missing_dvi_pdf = $$Pdest;
        }
	else {
            $failure = 1;
	}
    }
    if ($return != 0) {
        $failure = 1; 
	$$Plast_result = 2;
        if ( !$$Plast_message ) {
            $$Plast_message = "Run of rule '$rule' gave a non-zero error code";
	}
# !!??        $failure_msg = $$Plast_message;
        
    }
}  #END rdb_makeB1

#************************************************************

sub rdb_submakeB {
    # Call: rdb_submakeB
    # Makes all the source files for a given rule.
    # Assumes contexts for recursion, for make, and rule.
    %visited = %visited_at_rule_start;
    local $failure = 0;  # Error flag
    my @v = keys %visited;
#??    print "---submakeB $rule.  @v \n";
    rdb_do_files( sub{ rdb_recurse_rule( $$Pfrom_rule, 0,0,0, \&rdb_makeB1 ) } );
    return $failure;
}  #END rdb_submakeB

#************************************************************


sub rdb_classify_rules {
    # Usage: rdb_classify_rules( \%allowed_primaries, requested targets )
    # Assume the following variables are available (global or local):
    # Input:
    #    @requested_targets    # Set to target rules
    
    # Output:
    #    %current_primaries    # Keys are actual primaries
    #    @pre_primary          # Array of rules
    #    @post_primary         # Array of rules
    #    @one_time             # Array of rules
    # @pre_primary and @post_primary are in natural order of application.

    local $P_allowed_primaries = shift;
    local @requested_targets = @_;
    local $state = 0;       # Post-primary
    local @classify_stack = ();

    %current_primaries = ();
    @pre_primary = ();
    @post_primary = ();
    @one_time = ();

    rdb_recurseA( \@requested_targets, \&rdb_classify1, 0,0, \&rdb_classify2 );

    # Reverse, as tendency is to find last rules first.
    @pre_primary = reverse @pre_primary;
    @post_primary = reverse @post_primary;

    if ($diagnostics) {
	print "Rule classification: \n";
	if ($#requested_targets < 0) {
	    print "  No requested rules\n";
	}
	else {
	    print "  Requested rules:\n";
	    foreach ( @requested_targets ) { print "    $_\n"; }
	}
	if ($#pre_primary < 0) {
	    print "  No pre-primaries\n";
	}
	else {
	    print "  Pre-primaries:\n";
	    foreach (@pre_primary) { print "    $_\n"; }
	}
	print "  Primaries:\n";
	foreach (keys %current_primaries) { print "    $_\n"; }
	if ($#post_primary < 0) {
	    print "  No post-primaries\n";
	}
	else {
	    print "  Post-primaries:\n";
	    foreach (@post_primary) { print "    $_\n"; }
	}
	if ($#one_time < 0) {
	    print "  No one_time rules\n";
	}
	else {
	    print "  One_time rules:\n";
	    foreach ( @one_time ) { print "    $_\n"; }
	}
    } #end diagnostics

} #END rdb_classify_rules

#-------------------

sub rdb_classify1 {
    # Helper routine for rdb_classify_rules
    # Applied as rule_act1 in recursion over rules
    # Assumes rule context, and local variables from rdb_classify_rules
    #print "??======= '$rule' $depth $state ========== \n";
    push @classify_stack, [$state];
    if ( exists $possible_one_time{$rule} ) {
        # Normally, we will have already extracted the one_time rules,
        # and they will never be accessed here.  But just in case of
        # problems or generalizations, we will cover all possibilities:
        if ($depth > 1) {
           warn "ONE TIME rule not at outer level '$rule'\n";
        }
        push @one_time, $rule;
    }
    elsif ($state == 0) {
       if ( exists ${$P_allowed_primaries}{$rule} ) {
           $state = 1;   # In primary rule
           $current_primaries{ $rule } = 1;
       }
       else {
	   push @post_primary, $rule;
       }
    }
    else {
        $state = 2;     # in post-primary rule
	push @pre_primary, $rule;
    }
} #END rdb_classify1

#-------------------

sub rdb_classify2 {
    # Helper routine for rdb_classify_rules
    # Applied as rule_act2 in recursion over rules
    # Assumes rule context
    ($state) = @{ pop @classify_stack };
} #END rdb_classify2

#************************************************************


sub rdb_run1 {
    # Assumes contexts for: rule.
    # Unconditionally apply the rule
    # Returns return code from applying the rule.
    # Otherwise: 0 on other kind of success, -1 on error.

    # Source file data, by definition, correspond to the file state just before 
    # the latest run, and the run_time to the time just before the run:
    &rdb_update_filesA;
    $$Prun_time = time;
    $$Pchanged = 0;       # No special changes in files
    $$Plast_result = 0;
    $$Plast_message = '';

    # Return values for external command:
    my $return = 0;

    # Find any internal command
    my @int_args = @$PAint_cmd;
    my $int_cmd = shift @int_args;
    my @int_args_for_printing = @int_args;
    foreach (@int_args_for_printing) {
	if ( ! defined $_ ) { $_ = 'undef'; }
    }
    if ($int_cmd) {
	print "For rule '$rule', running '\&$int_cmd( @int_args_for_printing )' ...\n";
        $return = &$int_cmd( @int_args ); 
    }
    elsif ($$Pext_cmd) {
	$return = &rdb_ext_cmd;
    }
    else {
        warn "$My_name: Either a bug OR a configuration error:\n",
             "    Need to implement the command for '$rule'\n";
        &traceback();
        $return = -1;
        $$Plast_result = 2;
        $$Plast_message = "Bug or configuration error; incorrect command type";
    }
    if ( $rule =~ /^bibtex/ ) {
        my $retcode = check_bibtex_log($$Pbase);
        if ($retcode == 3) {
            $$Plast_result = 2;
            $$Plast_message = "Could not open bibtex log file for '$$Pbase'";
            push @warnings, $$Plast_message;
        }
        elsif ($retcode == 2) {
            $$Plast_message = "Bibtex errors: See file '$$Pbase.blg'";
            push @warnings, $$Plast_message;
        }
        elsif ($retcode == 1) {
            push @warnings, "Bibtex warnings for '$$Pbase'";
        }
    }

    $updated = 1;
    if ($$Ptest_kind == 3) { 
        # We are time-criterion first time only.  Now switch to
	# file-change criterion
        $$Ptest_kind = 1; 
    }
    $$Pout_of_date = $$Pout_of_date_user = 0;

    if ( ($$Plast_result == 0) && ($return != 0) ) {
        $$Plast_result = 2;
        if ($$Plast_message eq '') {
	    $$Plast_message = "Command for '$rule' gave return code $return";
	}
    }
    elsif ( $$Pdest && (! -e $$Pdest) ) {
	$$Plast_result = 1;
    }

    return $return;
}  # END rdb_run1

#-----------------

sub rdb_dummy_run1 {
    # Assumes contexts for: rule.
    # Update rule state as if the rule ran successfully,
    #    but don't run the rule.
    # Returns 0 (success code)

    # Source file data, by definition, correspond to the file state just before 
    # the latest run, and the run_time to the time just before the run:
    &rdb_update_filesA;
    $$Prun_time = time;
    $$Pchanged = 0;       # No special changes in files
    $$Plast_result = 0;
    $$Plast_message = '';

    if ($$Ptest_kind == 3) { 
        # We are time-criterion first time only.  Now switch to
	# file-change criterion
        $$Ptest_kind = 1; 
    }
    $$Pout_of_date = $$Pout_of_date_user = 0;

    return 0;
}  # END rdb_dummy_run1

#-----------------

sub rdb_ext_cmd {
    # Call: rdb_ext_cmd
    # Assumes rule context.  Runs external command with substitutions.
    # Uses defaults for the substitutions.  See rdb_ext_cmd1.
    return rdb_ext_cmd1();
} #END rdb_ext_cmd

#-----------------

sub rdb_ext_cmd1 {
    # Call: rdb_ext_cmd1( options, source, dest, base ) or rdb_ext_cmd1() or ...
    # Assumes rule context.  Returns command with substitutions.
    # Null arguments or unprovided arguments => use defaults.
    # for %S=source, %D=dest, %B=base, %R=root=base for latex, %O='', %T=texfile
    my ($options, $source, $dest, $base ) = @_;
    # Apply defaults
    $options ||= '';
    $source  ||= $$Psource;
    $dest    ||= $$Pdest;
    $base    ||= $$Pbase;
    
    my $ext_cmd = $$Pext_cmd;
    
    #Set character to surround filenames:
    my $q = $quote_filenames ? '"' : '';

    %subst = ( 
       '%O' => $options,
       '%R' => $q.$root_filename.$q,
       '%B' => $q.$base.$q,
       '%T' => $q.$texfile_name.$q,
       '%S' => $q.$source.$q,
       '%D' => $q.$dest.$q,
       '%%' => '%'         # To allow literal %B, %R, etc, by %%B.
    );

    my @tokens = split /(%.)/, $ext_cmd;
    foreach (@tokens) {
        if (exists($subst{$_})) { $_ = $subst{$_}; }
    }
    $ext_cmd = join '', @tokens;

    # print "quote is '$q'; ext_cmd = '$ext_cmd'\n";
    my ($pid, $return) = &Run_msg($ext_cmd);
    return $return;
} #END rdb_ext_cmd1

#-----------------

sub rdb_primary_run {
#?? See multipass_run in previous version Aug 2007 for issues
    # Call: rdb_primary_run
    # Assumes contexts for: recursion, make, & rule.
    # Assumes (a) the rule is a primary, 
    #         (b) a run has to be made,
    #         (c) source files have been made.
    # This routine carries out the run of the rule unconditionally,
    # and then parses log file etc.
    my $return = 0;

    my $return_latex = &rdb_run1;
    if (-e $$Pdest) { $missing_dvi_pdf = '';}

    ######### Analyze results of run:
    if ( ! -e "$root_filename.log" ) {
        $failure = 1;
        $$Plast_result = 2;
        $$Plast_message = $failure_msg = "(Pdf)LaTeX failed to generate a log file";
        return -1;
    }
    ####### NOT ANY MORE! Capture any changes in source file status before we
    #         check for errors in the latex run

    # Find current set of source files:
    &rdb_set_from_logB;

    # For each file of the kind made by epstopdf.sty during a run, 
    #   if the file has changed during a run, then the new version of
    #   the file will have been read during the run.  Unlike the usual
    #   case, we will NOT need to redo the primary run because of the
    #   change of this file during the run.  Therefore set the file as
    #   up-to-date:
    rdb_do_files( sub { if ($$Pcorrect_after_primary) {&rdb_update1;} } );

    # There may be new source files, and the run may have caused
    # circular-dependency files to be changed.  And the regular
    # source files may have been updated during a lengthy run of
    # latex.  So redo the makes for sources of the current rule:
    my $submake_return = &rdb_submakeB;
    &rdb_clear_change_record;
    &rdb_flag_changes_here;
    if ($$Pout_of_date && !$silent) { 
        &rdb_diagnose_changes( "Rule '$rule': " );
    }
    $updated = 1;    # Flag that some dependent file has been remade
    # Fix the state of the files as of now: this will solve the
    # problem of latex and pdflatex interfering with each other,
    # at the expense of some non-optimality
    #??  Check this is correct:
    &rdb_update_filesA;
    if ( $diagnostics ) {
	print "$My_name: Rules after run: \n";
	rdb_show();
    }

    $return = $return_latex;
    if ($return_latex && $$Pout_of_date_user) {
       print "Error in (pdf)LaTeX, but change of user file(s), ",
             "so ignore error & provoke rerun\n"
          if (! $silent);
       $return = 0;
    }

    # Summarize issues that may have escaped notice:
    my @warnings = ();
    if ($bad_reference) {
        push @warnings, "Latex could not resolve all references";
    }
    if ($bad_citation) {
        push @warnings, "Latex could not resolve all citations";
    }
    if ($#warnings > 0) {
	show_array( "$My_name: Summary of warnings:", @warnings );
    }
    return $return;
} #END rdb_primary_run

#************************************************************

sub rdb_clear_change_record {
    # Initialize diagnostics for reasons for running rule.
    @changed = ();
    @disappeared = ();
    @no_dest = ();          # We are not now using this
    @rules_never_run = ();
    @rules_to_apply = ();   # This is used in recursive application
                            # of rdb_flag_changes_here, to list
                            # rules that were out-of-date for some reason.
} #END rdb_clear_change_record 

#************************************************************

sub rdb_flag_changes_here {
    # Flag changes in current rule.  
    # Assumes rule context.
    # Optional argument: if true then fdb_get shouldn't do runtime test
    #                    for recalculation of md5
    local $ignore_run_time = $_[0];
    if ( ! defined $ignore_run_time ) { $ignore_run_time = 0; }
    local $dest_mtime = 0;
    $dest_mtime = get_mtime($$Pdest) if ($$Pdest);
    rdb_do_files( \&rdb_file_change1);
    if ($$Pout_of_date) {
	push @rules_to_apply, $rule;
    }
#??	print "======== flag: $rule $$Pout_of_date ==========\n";
} #END rdb_flag_changes_here

#************************************************************

sub rdb_file_change1 {
    # Call: &rdb_file_change1
    # Assumes rule and file context.  Assumes $dest_mtime set.
    # Flag whether $file in $rule has changed or disappeared.
    # Set rule's make flag if there's a change.
    my $run_time_argument = $ignore_run_time ? 0 : $$Prun_time;
    my ($new_time, $new_size, $new_md5) = fdb_get($file, $run_time_argument);
#??    print "FC1 '$rule':$file $$Pout_of_date TK=$$Ptest_kind\n"; 
#??    print "    OLD $$Ptime, $$Psize, $$Pmd5\n",
#??          "    New $new_time, $new_size, $new_md5\n";
    my $ext_no_period = ext_no_period( $file );
    if ( ($new_size < 0) && ($$Psize >= 0) ) {
	# print "Disappeared '$file' in '$rule'\n";
        push @disappeared, $file;
        # No reaction is good.  
        #$$Pout_of_date = 1;
        # ??? 1 Sep. 2008: I do NOT think so, for cusdep no-file-exists issue
        # ??? 30 Sep 2008: I think I have this fixed.  There were other changes
        #  needed.  No-change-flagged is correct.  The array @disappeared flags 
        #  files that have disappeared, if I need to know.  But having a source
        #  file disappear is not a reason for a remake unless I know how to 
        #  make the file.  If the file is a destination of a rule, that rule
        #  will be rerun.  It may be that the user is changing another source 
        #  in such a way that the disappeared file won't be needed.  Before the
        #  change is applied we get a superfluous infinite loop.
        return;
    }
    if ( ($new_size < 0) && ($$Psize < 0) ) {
	return;
    }
    if ( ($new_size != $$Psize) || ($new_md5 ne $$Pmd5) ) {
#??        print "FC1: changed $file: ($new_size != $$Psize) $new_md5 ne $$Pmd5)\n";
	push @changed, $file;
	$$Pout_of_date = 1;
        if ( ! exists $generated_exts_all{$ext_no_period} ) {
            $$Pout_of_date_user = 1;
	}
    }
    elsif ( $new_time != $$Ptime ) {
#warn "--==-- Unchanged $file, changed time, update filetime in $rule\n";
	$$Ptime = $new_time;
    }
    if ( ( ($$Ptest_kind == 2) || ($$Ptest_kind == 3) )
         && (! exists $generated_exts_all{$ext_no_period} )
         && ( $new_time > $dest_mtime )
        ) {
#??        print "FC1: changed $file: ($new_time > $dest_mtime)\n";
	    push @changed, $file;
	    $$Pout_of_date = $$Pout_of_date_user = 1;
    }
} #END rdb_file_change1

#************************************************************

sub rdb_new_changes {
    &rdb_clear_change_record;
    rdb_recurseA( [@_], sub{ &rdb_flag_changes_here(1); } );
    return ($#changed >= 0) || ($#no_dest >= 0) || ($#rules_to_apply >= 0);
} #END rdb_new_changes

#************************************************************

sub rdb_diagnose_changes {
    # Call: rdb_diagnose_changes or rdb_diagnose_changes( heading )
    # List changes on STDERR
    # Precede the message by the optional heading, else by "$My_name: " 
    my $heading = defined($_[0]) ?   $_[0]  :  "$My_name: "; 

    if ($#rules_never_run >= 0) {
	warn "${heading}Rules & subrules not known to be previously run:\n";
        foreach (@rules_never_run) { warn "   $_\n"; }
    }
    if ( ($#changed >= 0) || ($#disappeared >= 0) || ($#no_dest >= 0) ) {
        warn "${heading}File changes, etc:\n";
        if ( $#changed >= 0 ) {
	    warn "   Changed files, or newly in use since previous run(s):\n";
       	    foreach (uniqs(@changed)) { warn "      '$_'\n"; }
        }
        if ( $#disappeared >= 0 ) {
	    warn "   No-longer-existing files:\n";
	    foreach (uniqs(@disappeared)) { warn "      '$_'\n"; }
        }
        if ( $#no_dest >= 0 ) {
	    warn "   Non-existent destination files:\n";
	    foreach (uniqs(@no_dest)) { warn "      '$_'\n"; }
        }
    }
    elsif ($#rules_to_apply >=0) {
         warn "${heading}The following rules & subrules became out-of-date:\n";
         foreach (@rules_to_apply) { warn "      '$_'\n"; }
    }
    else {
	warn "${heading}No file changes\n";
    }
}     #END rdb_diagnose_changes


#************************************************************
#************************************************************
#************************************************************
#************************************************************

#************************************************************
#************************************************************
#************************************************************
#************************************************************

# Routines for convenient looping and recursion through rule database
# ================= NEW VERSION ================

# There are several places where we need to loop through or recurse
# through rules and files.  This tends to involve repeated, tedious
# and error-prone coding of much book-keeping detail.  In particular,
# working on files and rules needs access to the variables involved,
# which either involves direct access to the elements of the database,
# and consequent fragility against changes and upgrades in the
# database structure, or involves lots of routines for reading and
# writing data in the database, then with lots of repetitious
# house-keeping code.
#
# The routines below provide a solution.  Looping and recursion
# through the database are provided by a set of basic routines where
# each necessary kind of looping and iteration is coded once.  The
# actual actions are provided as references to action subroutines.
# (These can be either actual references, as in \&routine, or
# anonymous subroutines, as in sub{...}, or aas a zero value 0 or an
# omitted argument, to indicate that no action is to be performed.)
#
# When the action subroutine(s) are actually called, a context for the
# rule and/or file (as appropriate) is given by setting named
## NEW ??
# variables to REFERENCES to the relevant data values.  These can be
# used to retrieve and set the data values.  As a convention,
# references to scalars are given by variables named start with "$P",
# as in "$Pdest", while references to arrays start with "$PA", as in 
# "$PAint_cmd", and references to hashes with "$PH", as in "$PHsource".
# After the action subroutine has finished, checks for data
# consistency may be made. 
## ??? OLD
# variables to the relevant data values.  After the action subroutine
# has finished, the database is updated with the values of these named
# variables, with any necessary consistency checks.  Thus the action
# subroutines can act on sensibly named variables without needed to
# know the database structure.  
#
# The only routines that actually use the database structure and need
# to be changed if that is changed are:  (a) the routines rdb_one_rule
# and rdb_one_file that implement the calling of the action subroutines,
# (b) routines for creation of single rules and file items, and (c) to
# a lesser extent, the routine for destroying a file item.  
#
# Note that no routine is provided for destroying a rule.  During a
# run, a rule, with its source files, may become inaccessible or
# unused.  This happens dynamically, depending on the dependencies
# caused by changes in the source file or by error conditions that
# cause the computation of dependencies, particular of latex files, to
# become wrong.  In that situation the files certainly come and go in
# the database, but subsidiary rules, with their content information
# on their source files, need to be retained so that their use can be
# reinstated later depending on dynamic changes in other files.
#
# However, there is a potential memory leak unless some pruning is
# done in what is written to the fdb file.  (Probably only accessible
# rules and those for which source files exist.  Other cases have no
# relevant information that needs to be preserved between runs.)

#
#


#************************************************************

# First the top level routines for recursion and iteration

#************************************************************

sub rdb_recurseA {
    # Call: rdb_recurseA( rule | [ rules],
    #                    \&rule_act1, \&file_act1, \&file_act2, 
    #                    \&rule_act2 )
    # The actions are pointers to subroutines, and may be null (0, or
    # undefined) to indicate no action to be applied.
    # Recursively acts on the given rules and all ancestors:
    #   foreach rule found:
    #       apply rule_act1
    #       loop through its files:
    #          apply file_act1
    #          act on its ancestor rule, if any
    #          apply file_act2
    #       apply rule_act2
    # Guards against loops.  
    # Access to the rule and file data by local variables, only
    #   for getting and setting.

    # This routine sets a context for anything recursive, with @heads,
    # %visited  and $depth being set as local variables.

    local @heads = ();
    my $rules = shift;

    # Distinguish between single rule (a string) and a reference to an
    # array of rules:
    if ( ref $rules eq 'ARRAY' ) { @heads = @$rules; }
    else { @heads = ( $rules ); }

    # Keep a list of visited rules, used to block loops in recursion:
    local %visited = (); 
    local $depth = 0;

    foreach $rule ( @heads ) { rdb_recurse_rule( $rule, @_ ); }

} #END rdb_recurseA

#************************************************************

sub rdb_for_all {
    # Call: rdb_for_all( \&rule_act1, \&file_act, \&rule_act2 )
    # Loops through all rules and their source files, using the 
    #   specified set of actions, which are pointers to subroutines.
    # Sorts rules alphabetically.
    # See rdb_for_some for details.
    rdb_for_some( [ sort keys %rule_db ], @_);
} #END rdb_for_all

#************************************************************

sub rdb_for_some {
    # Call: rdb_for_some( rule | [ rules],
    #                    \&rule_act1, \&file_act, \&rule_act2)
    # Actions can be zero, and rules at tail of argument list can be
    # omitted.  E.g. rdb_for_some( rule, 0, \&file_act ).  
    # Anonymous subroutines can be used, e.g., rdb_for_some( rule, sub{...} ).  
    #
    # Loops through rules and their source files, using the 
    # specified set of rules:
    #   foreach rule:
    #       apply rule_act1
    #       loop through its files:
    #          apply file_act
    #       apply rule_act2
    #
    # Rule data and file data are made available in local variables 
    # for access by the subroutines.

    local @heads = ();
    my $rules = shift;
    # Distinguish between single rule (a string) and a reference to an
    # array of rules:
    if ( ref $rules eq 'ARRAY' ) { @heads = @$rules; }
    else { @heads = ( $rules ); }

    foreach $rule ( @heads ) {
        # $rule is implicitly local
	&rdb_one_rule( $rule, @_ );
    }
}  #END rdb_for_some

#************************************************************

sub rdb_for_one_file {
    my $rule = shift;
    # Avoid name collisions with general recursion and iteraction routines:
    local $file1 = shift;
    local $action1 = shift;
    rdb_for_some( $rule, sub{rdb_one_file($file1,$action1)} );
} #END rdb_for_one_file


#************************************************************

#   Routines for inner part of recursion and iterations

#************************************************************

sub rdb_recurse_rule {
    # Call: rdb_recurse_rule($rule, \&rule_act1, \&file_act1, \&file_act2, 
    #                    \&rule_act2 )
    # to do the work for one rule, recurisvely called from_rules for
    # the sources of the rules.
    # Assumes recursion context, i.e. that %visited, @heads, $depth.
    # We are overriding actions:
    my ($rule, $rule_act1, $new_file_act1, $new_file_act2, $rule_act2)
	= @_;
    # and must propagate the file actions:
    local $file_act1 = $new_file_act1;
    local $file_act2 = $new_file_act2;
    # Prevent loops:
    if ( (! $rule) || exists $visited{$rule} ) { return; }
    $visited{$rule} = 1;
    # Recursion depth
    $depth++;
    # We may need to repeat actions on dependent rules, without being
    # blocked by the test on visited files.  So save %visited:
    local %visited_at_rule_start = %visited;
    # At end, the last value set for %visited wins.
    rdb_one_rule( $rule, $rule_act1, \&rdb_recurse_file, $rule_act2 );
    $depth--;
 } #END rdb_recurse_rule 

#************************************************************

sub rdb_recurse_file {
    # Call: rdb_recurse_file to do the work for one file.
    # This has no arguments, since it is used as an action subroutine,
    # passed as a reference in calls in higher-level subroutine.
    # Assumes contexts set for: Recursion, rule, and file
    &$file_act1 if $file_act1;
    rdb_recurse_rule( $$Pfrom_rule, $rule_act1, $file_act1, $file_act2,
		      $rule_act2 )
        if $$Pfrom_rule;
    &$file_act2 if $file_act2;
} #END rdb_recurse_file

#************************************************************

sub rdb_do_files {
    # Assumes rule context, including $PHsource.
    # Applies an action to all the source files of the rule.
    local $file_act = shift;
    my @file_list = sort keys %$PHsource;
    foreach my $file ( @file_list ){
        rdb_one_file( $file, $file_act );
    }
} #END rdb_do_files

#************************************************************

# Routines for action on one rule and one file.  These are the main
# places (in addition to creation and destruction routines for rules
# and files) where the database structure is accessed.

#************************************************************

sub rdb_one_rule {
    # Call: rdb_one_rule( $rule, $rule_act1, $file_act, $rule_act2 )
    # Sets context for rule and carries out the actions.
#===== Accesses rule part of database structure =======

    local ( $rule, $rule_act1, $file_act, $rule_act2 ) = @_;
#??    &R1;
    if ( (! $rule) || ! rdb_rule_exists($rule) ) { return; }

    local ( $PArule_data, $PHsource ) = @{$rule_db{$rule}};
    local ($Pcmd_type, $Pext_cmd, $PAint_cmd, $Ptest_kind, 
           $Psource, $Pdest, $Pbase,
           $Pout_of_date, $Pout_of_date_user, $Prun_time, $Pchanged,
           $Plast_result, $Plast_message )
        = Parray( $PArule_data );
    # Correct array ref:
    $PAint_cmd = $$PArule_data[2];

    &$rule_act1 if $rule_act1;
    &rdb_do_files( $file_act ) if $file_act;
    &$rule_act2 if $rule_act2;

#??    &R2;
} #END rdb_one_rule

#************************************************************

sub rdb_one_file {
    # Call: rdb_one_file($file, $file_act)
    # Sets context for file and carries out the action.
    # Assumes $rule context set.
#===== Accesses file part of database structure =======
    local ($file, $file_act) = @_;
#??    &F1;
    if ( (!$file) ||(!exists ${$PHsource}{$file}) ) { return; }
    local $PAfile_data = ${$PHsource}{$file};
    local ($Ptime, $Psize, $Pmd5, $Pfrom_rule, $Pcorrect_after_primary ) 
          = Parray( $PAfile_data );
    &$file_act if $file_act;
    if ( ! rdb_rule_exists( $$Pfrom_rule ) ) {
        $$Pfrom_rule = '';
    }
#??    &F2;
} #END rdb_one_file

#************************************************************

# Routines for creation of rules and file items, and for removing file
# items. 

#************************************************************

sub rdb_create_rule {
    # rdb_create_rule( rule, command_type, ext_cmd, int_cmd, test_kind,
    #                  source, dest, base, 
    #                  needs_making, run_time, set_file_not_exists )
    # int_cmd is either a string naming a perl subroutine or it is a
    # reference to an array containing the subroutine name and its
    # arguments. 
    # Makes rule.  Error if it already exists.
    # Omitted arguments: replaced by 0 or '' as needed.    
# ==== Sets rule data ====
    my ( $rule, $cmd_type, $int_cmd, $PAext_cmd, $test_kind, 
         $source, $dest, $base, 
         $needs_making, $run_time, $set_file_not_exists ) = @_;
    my $changed = 0;
    # Set defaults, and normalize parameters:
    foreach ( $cmd_type, $int_cmd, $PAext_cmd, $source, $dest, $base, 
              $set_file_not_exists ) {
        if (! defined $_) { $_ = ''; }
    }
    foreach ( $needs_making, $run_time, $test_kind ) {
        if (! defined $_) { $_ = 0; }
    }
    if (!defined $test_kind) {
        # Default to test on file change
        $test_kind = 1; 
    }
    if ( ref( $PAext_cmd ) eq '' ) {
        #  It is a single command.  Convert to array reference:
        $PAext_cmd = [ $PAext_cmd ];
    }
    else {
        # COPY the referenced array:
        $PAext_cmd = [ @$PAext_cmd ];
    }

    $rule_db{$rule} = 
        [  [$cmd_type, $int_cmd, $PAext_cmd, $test_kind, 
            $source, $dest, $base, $needs_making, 0, $run_time,
            $changed, -1, '' ],
           {}
	];
    if ($source) { 
       rdb_ensure_file( $rule, $source, undef, $set_file_not_exists );  
    }
} #END rdb_create_rule

#************************************************************

sub rdb_ensure_file {
    # rdb_ensure_file( rule, file[, fromrule[, set_not_exists]] )
    # Ensures the source file item exists in the given rule.
    # Then if the fromrule is specified, set it for the file item.
    # If the item is created, then:
    #    (a) by default initialize it to current file state.
    #    (b) but if the fourth argument, set_not_exists, is true, 
    #        initialize the item as if the file does not exist.
    #        This case is typically used when the log file for a run
    #        of latex/pdflatex claims that the file was non-existent
    #        at the beginning of a run. 
#============ rule and file data set here ======================================
    my $rule = shift;
    local ( $new_file, $new_from_rule, $set_not_exists ) = @_;
    if ( ! rdb_rule_exists( $rule ) ) {
	die_trace( "$My_name: BUG in rdb_ensure_file: non-existent rule '$rule'" );
    }
    if ( ! defined $new_file ) {
	die_trace( "$My_name: BUG in rdb_ensure_file: undefined file for '$rule'" );
    }
    if ( ! defined $set_not_exists ) { $set_not_exists = 0; }
    rdb_one_rule( $rule, 
                  sub{
                      if (! exists ${$PHsource}{$new_file} ) {
  		          if ( $set_not_exists ) {
                              ${$PHsource}{$new_file} = [0, -1, 0, '', 0];
		          }
		          else {
                              ${$PHsource}{$new_file} 
                              = [fdb_get($new_file, $$Prun_time), '', 0];
               	          }
		      }
		  }
    );
    if (defined $new_from_rule ) {
	rdb_for_one_file( $rule, $new_file, sub{ $$Pfrom_rule = $new_from_rule; });
    }
} #END rdb_ensure_file 

#************************************************************

sub rdb_remove_files {
    # rdb_remove_file( rule, file,... )
    # Removes file(s) for the rule.  
    my $rule = shift;
    if (!$rule) { return; }
    local @files = @_;
    rdb_one_rule( $rule, 
                  sub{ foreach (@files) { delete ${$PHsource}{$_}; }  }
    );
} #END rdb_remove_files

#************************************************************

sub rdb_rule_exists { 
    # Call rdb_rule_exists($rule): Returns whether rule exists.
    my $rule = shift;
    if (! $rule ) { return 0; }
    return exists $rule_db{$rule}; 
} #END rdb_rule_exists

#************************************************************

sub rdb_file_exists { 
    # Call rdb_file_exists($rule, $file): 
    # Returns whether source file item in rule exists.
    local ( $rule, $file ) = @_;
    local $exists = 0;
    rdb_one_rule( $rule, 
                  sub{ $exists =  exists( ${$PHsource}{$file} ) ? 1:0; } 
		);
    return $exists; 
} #END rdb_file_exists

#************************************************************

sub rdb_update_gen_files {
    # Assumes rule context.  Update source files of rule to current state.
    rdb_do_files( 
        sub{
	    if ( exists $generated_exts_all{ ext_no_period($file) } ) {&rdb_update1;} 
        }
    );
} #END rdb_update_gen_files

#************************************************************

sub rdb_update_filesA {
    # Call: fdb_updateA
    # Assumes rule context.  Update source files of rule to current state.
    rdb_do_files( \&rdb_update1 );
}

#************************************************************

sub rdb_update1 {
    # Call: rdb_update1.  
    # Assumes file context.  Updates file data to correspond to
    # current file state on disk
    ($$Ptime, $$Psize, $$Pmd5) = fdb_get($file);
}

#************************************************************

sub rdb_set_file1 {
    # Call: fdb_file1(rule, file, new_time, new_size, new_md5)
    # Sets file time, size and md5.
    my $rule = shift;
    my $file = shift;
    local @new_file_data = @_;
    rdb_for_one_file( $rule, $file, sub{ ($$Ptime,$$Psize,$$Pmd5)=@new_file_data; } );
}

#************************************************************

sub rdb_dummy_file {
    # Returns file data for non-existent file
# ==== Uses rule_db structure ====
    return (0, -1, 0, '');
}

#************************************************************
#************************************************************

# Predefined subroutines for custom dependency

sub cus_dep_delete_dest {
    # This subroutine is used for situations like epstopdf.sty, when
    #   the destination (target) of the custom dependency invoking
    #   this subroutine will be made by the primary run provided the
    #   file (destination of the custom dependency, source of the
    #   primary run) doesn't exist.
    # It is assumed that the resulting file will be read by the
    #   primary run.

    # Remove the destination file, to indicate it needs to be remade:
    unlink $$Pdest;
    # Arrange that the non-existent destination file is not treated as
    #   an error.  The variable changed here is a bit misnamed.
    $$Pchanged = 1;
    # Ensure a primary run is done
    &cus_dep_require_primary_run;
    # Return success:
    return 0;
}

#************************************************************

sub cus_dep_require_primary_run {
    # This subroutine is used for situations like epstopdf.sty, when
    #   the destination (target) of the custom dependency invoking
    #   this subroutine will be made by the primary run provided the
    #   file (destination of the custom dependency, source of the
    #   primary run) doesn't exist.
    # It is assumed that the resulting file will be read by the
    #   primary run.

    local $cus_dep_target = $$Pdest;
    # Loop over all rules and source files:
    rdb_for_all( 0, 
                 sub { if ($file eq $cus_dep_target) {
                            $$Pout_of_date = 1;
                            $$Pcorrect_after_primary = 1;
                       }
                     }
               );
    # Return success:
    return 0;
}


#************************************************************
#************************************************************
#************************************************************
#
#      UTILITIES:
#

#************************************************************
# Miscellaneous

sub show_array {
# For use in diagnostics and debugging. 
#  On stderr, print line with $_[0] = label.  
#  Then print rest of @_, one item per line preceeded by some space
    warn "$_[0]\n";
    shift;
    if ($#_ >= 0) {  foreach (@_){ warn "  '$_'\n";} }
    else { warn "  NONE\n"; }
}

#************************************************************

sub Parray {
    # Call: Parray( \@A )
    # Returns array of references to the elements of @A
    my $PA = shift;
    my @P = (undef) x (1+$#$PA);
    foreach my $i (0..$#$PA) { $P[$i] = \$$PA[$i]; }
    return @P;
}

#************************************************************

sub glob_list {
    # Glob a collection of filenames.  Sort and eliminate duplicates
    # Usage: e.g., @globbed = glob_list(string, ...);
    my @globbed = ();
    foreach (@_) {
        push @globbed, glob;
    }
    return uniqs( @globbed );
}

#==================================================

sub glob_list1 {
    # Glob a collection of filenames.  
    # But no sorting or elimination of duplicates
    # Usage: e.g., @globbed = glob_list1(string, ...);
    # Since perl's glob appears to use space as separator, I'll do a special check
    # for existence of non-globbed file (assumed to be tex like)

    my @globbed = ();
    foreach my $file_spec (@_) {
        # Problem, when the PATTERN contains spaces, the space(s) are
        # treated as pattern separaters (in MSWin at least).
        # MSWin: I can quote the pattern (is that MSWin native, or also 
        #        cygwin?)
        # Linux: Quotes in a pattern are treated as part of the filename!
        #        So quoting a pattern is definitively wrong.
        # The following hack solves this partly, for the cases that there is no wildcarding 
        #    and the specified file exists possibly space-containing, and that there is wildcarding,
        #    but spaces are prohibited.
        if ( -e $file_spec || -e "$file_spec.tex" ) { 
           # Non-globbed file exists, return the file_spec.
           # Return $file_spec only because this is not a file-finding subroutine, but
           #   only a globber
           push @globbed, $file_spec; 
        }
        else { 
            # This glob fails to work as desired, if the pattern contains spaces.
            push @globbed, glob( "$file_spec" );
        }
    }
    return @globbed;
}

#************************************************************
# Miscellaneous

sub prefix {
   #Usage: prefix( string, prefix );
   #Return string with prefix inserted at the front of each line
   my @line = split( /\n/, $_[0] );
   my $prefix = $_[1];
   for (my $i = 0; $i <= $#line; $i++ ) {
       $line[$i] = $prefix.$line[$i]."\n";
   }
   return join( "", @line );
}


#************************************************************
#************************************************************
#      File handling utilities:


#************************************************************

sub get_latest_mtime
# - arguments: each is a filename.
# - returns most recent modify time.
{
  my $return_mtime = 0;
  foreach my $include (@_)
  {
    my $include_mtime = &get_mtime($include);
    # The file $include may not exist.  If so ignore it, otherwise
    # we'll get an undefined variable warning.
    if ( ($include_mtime) && ($include_mtime >  $return_mtime) )
    {
      $return_mtime = $include_mtime;
    }
  }
  return $return_mtime;
}

#************************************************************

sub get_mtime_raw
{ 
  my $mtime = (stat($_[0]))[9];
  return $mtime;
}

#************************************************************

sub get_mtime { 
    return get_mtime0($_[0]);
}

#************************************************************

sub get_mtime0 {
   # Return time of file named in argument
   # If file does not exist, return 0;
   if ( -e $_[0] ) {
       return get_mtime_raw($_[0]);
   }
   else {
       return 0;
   }
}

#************************************************************

sub get_size {
   # Return time of file named in argument
   # If file does not exist, return 0;
   if ( -e $_[0] ) {
       return get_size_raw($_[0]);
   }
   else {
       return 0;
   }
}

#************************************************************

sub get_size_raw
{ 
  my $size = (stat($_[0]))[7];
  return $size;
}

#************************************************************

sub get_time_size {
   # Return time and size of file named in argument
   # If file does not exist, return (0,-1);
   if ( -e $_[0] ) {
       return get_time_size_raw($_[0]);
   }
   else {
       return (0,-1);
   }
}

#************************************************************

sub get_time_size_raw
{ 
  my $mtime = (stat($_[0]))[9];
  my $size = (stat($_[0]))[7];
  return ($mtime, $size);
}

#************************************************************

sub get_checksum_md5 {
    my $source = shift;
    my $input = new FileHandle;
    my $md5 = Digest::MD5->new;
    my $ignore_pattern = '';

#warn "======= GETTING MD5: $source\n";
    if ( $source eq "" ) { 
       # STDIN:
       open( $input, '-' );
    }
    elsif ( -d $source ) {
        # We won't use checksum for directory
        return 0;
    }
    else {
        open( $input, '<', $source )
        or return 0;
        my ($base, $path, $ext) = fileparseA( $source );
        $ext =~ s/^\.//;
        if ( exists $hash_calc_ignore_pattern{$ext} ) {
            $ignore_pattern = $hash_calc_ignore_pattern{$ext};
        }
    }

    if ( $ignore_pattern ) {
        while (<$input>) {
            if ( /$ignore_pattern/ ){
		$_= '';
	    }
            $md5->add($_);
        }
    }
    else {
        $md5->addfile($input);
    }
    close $input;
    return $md5->hexdigest();
}

#************************************************************
#************************************************************

sub find_file1 {
#?? Need to use kpsewhich, if possible

    # Usage: find_file1(name, ref_to_array_search_path)
    # Modified find_file, which doesn't die.
    # Given filename and path, return array of:
    #             full name 
    #             retcode
    # On success: full_name = full name with path, retcode = 0
    # On failure: full_name = given name, retcode = 1

  my $name = $_[0];
  # Make local copy of path, since we may rewrite it!
  my @path = @{$_[1]};
  if ( $name =~ /^\// ) {
     # Absolute path (if under UNIX)
     # This needs fixing, in general
     if (-e $name) { return( $name, 0 );}
     else { return( $name, 1 );}
  }
  foreach my $dir ( @path ) {
      #??print "-------------dir='$dir',  ";
      # Make $dir concatenatable, and empty for current dir:
      if ( $dir eq '.' ) { 
          $dir = ''; 
      }
      elsif ( $dir =~ /[\/\\:]$/ ) { 
          #OK if dir ends in / or \ or :
      }
      elsif ( $dir ne '' ) { 
          #Append directory separator only to non-empty dir
          $dir = "$dir/"; 
      }
      #?? print " newdir='$dir'\n";
      if (-e "$dir$name") {
          return("$dir$name", 0);
      }
  }
  my @kpse_result = kpsewhich( $name );
  if ($#kpse_result > -1) {
      return( $kpse_result[0], 0);
  }
  return("$name" , 1);
} #END find_file1

#************************************************************

sub find_file_list1 {
    # Modified version of find_file_list that doesn't die.
    # Given output and input arrays of filenames, a file suffix, and a path, 
    # fill the output array with full filenames
    # Return a status code:
    # Retcode = 0 on success
    # Retocde = 1 if at least one file was not found
    # Usage: find_file_list1( ref_to_output_file_array, 
    #                         ref_to_input_file_array, 
    #                         suffix,
    #                         ref_to_array_search_path
    #                       )

  my $ref_output = $_[0];
  my $ref_input  = $_[1];
  my $suffix     = $_[2];
  my $ref_search = $_[3];

#??  show_array( "=====find_file_list1.  Suffix: '$suffix'\n Source:",  @$ref_input );
#??  show_array( " Bibinputs:",  @$ref_search );

  my @return_list = ();    # Generate list in local array, since input 
                           # and output arrays may be same
  my $retcode = 0;
  foreach my $file (@$ref_input) {
    my ($tmp_file, $find_retcode) = &find_file1( "$file$suffix", $ref_search );
    if ($tmp_file)  {
    	push @return_list, $tmp_file;
    }
    if ( $find_retcode != 0 ) {
        $retcode = 1;
    }
  }
  @$ref_output = @return_list;
#??  show_array( " Output", @$ref_output );
#??  foreach (@$ref_output) { if ( /\/\// ) {  print " ====== double slash in  '$_'\n"; }  }
  return $retcode;
} #END find_file_list1

#************************************************************

sub kpsewhich {
# Usage: kpsewhich( filespec, ...)
# Returns array of files with paths as found by kpsewhich
#    kpsewhich( 'try.sty', 'jcc.bib' );
# Can also do, e.g.,
#    kpsewhich( '-format=bib', 'trial.bib', 'file with spaces');
    my $cmd = $kpsewhich;
    my @args = @_;
    foreach (@args) {
        if ( ! /^-/ ) {
            $_ = "\"$_\"";
	}
    }
    foreach ($cmd) {
        s/%[RBTDO]//g;
    }
    $cmd =~ s/%S/@args/g;
    my @found = ();
    local $fh;
    open $fh, "$cmd|"
        or die "Cannot open pipe for \"$cmd\"\n";
    while ( <$fh> ) {
	s/^\s*//;
        s/\s*$//;
        push @found, $_;
    }
    close $fh;
#    show_array( "Kpsewhich: '$cmd', '$file_list' ==>", @found );
    return @found;
}

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

sub add_cus_dep {
    # Usage: add_cus_dep( from_ext, to_ext, flag, sub_name )
    # Add cus_dep after removing old versions
    my ($from_ext, $to_ext, $must, $sub_name) = @_;
    remove_cus_dep( $from_ext, $to_ext );
    push @cus_dep_list, "$from_ext $to_ext $must $sub_name";
}

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

sub remove_cus_dep {
    # Usage: remove_cus_dep( from_ext, to_ext )
    my ($from_ext, $to_ext) = @_;
    my $i = 0;
    while ($i <= $#cus_dep_list) {
	if ( $cus_dep_list[$i] =~ /^$from_ext $to_ext / ) {
	    splice @cus_dep_list, $i, 1;
	}
	else {
	    $i++;
	}
    }
}

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

sub show_cus_dep {
    show_array( "Custom dependency list:", @cus_dep_list );
}

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

sub add_input_ext {
    # Usage: add_input_ext( rule, ext, ... )
    # Add extension(s) (specified without a leading period) to the 
    # list of input extensions for the given rule.  The rule should be
    # 'latex' or 'pdflatex'.  These extensions are used when an input
    # file without an extension is found by (pdf)latex, as in
    # \input{file} or \includegraphics{figure}.  When latexmk searches
    # custom dependencies to make the missing file, it will assume that
    # the file has one of the specified extensions.
    my $rule = shift;
    if ( ! exists $input_extensions{$rule} ) {
       $input_extensions{$rule} = {};
    }
    my $Prule = $input_extensions{$rule};
    foreach (@_) { $$Prule{$_} = 1; }
}

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

sub remove_input_ext {
    # Usage: remove_input_ext( rule, ext, ... )
    # Remove extension(s) (specified without a leading period) to the 
    # list of input extensions for the given rule.  The rule should be
    # 'latex' or 'pdflatex'.  See sub add_input_ext for the use.
    my $rule = shift;
    if ( ! exists $input_extensions{$rule} ) { return; }
    my $Prule = $input_extensions{$rule};
    foreach (@_) { delete $$Prule{$_}; }
}

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

sub show_input_ext {
    # Usage: show_input_ext( rule )
    my $rule = shift;
    show_array ("Input extensions for rule '$rule': ", 
                keys %{$input_extensions{$rule}} );
}

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

sub find_dirs1 {
   # Same as find_dirs, but argument is single string with directories
   # separated by $search_path_separator
   find_dirs( &split_search_path( $search_path_separator, ".", $_[0] ) );
}


#************************************************************

sub find_dirs {
# @_ is list of directories
# return: same list of directories, except that for each directory 
#         name ending in //, a list of all subdirectories (recursive)
#         is added to the list.
#   Non-existent directories and non-directories are removed from the list
#   Trailing "/"s and "\"s are removed
    local @result = ();
    my $find_action 
        = sub 
          { ## Subroutine for use in File::find
            ## Check to see if we have a directory
	       if (-d) { push @result, $File::Find::name; }
	  };
    foreach my $directory (@_) {
        my $recurse = ( $directory =~ m[//$] );
        # Remove all trailing /s, since directory name with trailing /
        #   is not always allowed:
        $directory =~ s[/+$][];
        # Similarly for MSWin reverse slash
        $directory =~ s[\\+$][];
	if ( ! -e $directory ){
            next;
	}
	elsif ( $recurse ){
            # Recursively search directory
            find( $find_action, $directory );
	}
        else {
            push @result, $directory;
	}
    }
    return @result;
}

#************************************************************

sub uniq 
# Read arguments, delete neighboring items that are identical,
# return array of results
{
    my @sort = ();
    my ($current, $prev);
    my $first = 1;
    while (@_)
    {
	$current = shift;
        if ($first || ($current ne $prev) )
	{
            push @sort, $current; 
            $prev = $current;
            $first = 0;
        }
    }
    return @sort;
}

#==================================================

sub uniq1 {
   # Usage: uniq1( strings )
   # Returns array of strings with duplicates later in list than
   # first occurence deleted.  Otherwise preserves order.

    my @strings = ();
    my %string_hash = ();

    foreach my $string (@_) {
        if (!exists( $string_hash{$string} )) { 
            $string_hash{$string} = 1;
            push @strings, $string; 
        }
    }
    return @strings;
}

#************************************************************

sub uniqs {
    # Usage: uniq2( strings )
    # Returns array of strings sorted and with duplicates deleted
    return uniq( sort @_ );
}

#************************************************************

sub ext {
    # Return extension of filename.  Extension includes the period
    my $file_name = $_[0];
    my ($base_name, $path, $ext) = fileparseA( $file_name );
    return $ext;
 }

#************************************************************

sub ext_no_period {
    # Return extension of filename.  Extension excludes the period
    my $file_name = $_[0];
    my ($base_name, $path, $ext) = fileparseA( $file_name );
    $ext =~ s/^\.//;
    return $ext;
 }

#************************************************************

sub fileparseA {
    # Like fileparse but replace $path for current dir ('./' or '.\') by ''
    # Also default second argument to get normal extension.
    my $given = $_[0];
    my $pattern = '\.[^\.]*';
    if  ($#_ > 0 ) { $pattern = $_[1]; }
    my ($base_name, $path, $ext) = fileparse( $given, $pattern );
    if ( ($path eq './') || ($path eq '.\\') ) { 
        $path = ''; 
    }
    return ($base_name, $path, $ext);
 }

#************************************************************

sub fileparseB {
    # Like fileparse but with default second argument for normal extension
    my $given = $_[0];
    my $pattern = '\.[^\.]*';
    if  ($#_ > 0 ) { $pattern = $_[1]; }
    my ($base_name, $path, $ext) = fileparse( $given, $pattern );
    return ($base_name, $path, $ext);
 }

#************************************************************

sub split_search_path 
{
# Usage: &split_search_path( separator, default, string )
# Splits string by separator and returns array of the elements
# Allow empty last component.
# Replace empty terms by the default.
    my $separator = $_[0]; 
    my $default = $_[1]; 
    my $search_path = $_[2]; 
    my @list = split( /$separator/, $search_path);
    if ( $search_path =~ /$separator$/ ) {
        # If search path ends in a blank item, the split subroutine
	#    won't have picked it up.
        # So add it to the list by hand:
        push @list, "";
    }
    # Replace each blank argument (default) by current directory:
    for ($i = 0; $i <= $#list ; $i++ ) {
        if ($list[$i] eq "") {$list[$i] = $default;}
    }
    return @list;
}

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


sub tempfile1 {
    # Makes a temporary file of a unique name.  I could use file::temp,
    # but it is not present in all versions of perl
    # Filename is of form $tmpdir/$_[0]nnn$suffix, where nnn is an integer
    my $tmp_file_count = 0;
    my $prefix = $_[0];
    my $suffix = $_[1];
    while (1==1) {
        # Find a new temporary file, and make it.
        $tmp_file_count++;
        my $tmp_file = "${tmpdir}/${prefix}${tmp_file_count}${suffix}";
        if ( ! -e $tmp_file ) {
            open( TMP, ">$tmp_file" ) 
               or next;
            close(TMP);
            return $tmp_file;
	 }
     }
     die "$My_name.tempfile1: BUG TO ARRIVE HERE\n";
}

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

#************************************************************
#************************************************************
#      Process/subprocess routines

sub Run_msg {
    # Same as Run, but give message about my running
    warn_running( "Running '$_[0]'" );
    Run($_[0]);
}

sub Run {
# Usage: Run ("program arguments ");
#    or  Run ("start program arguments");
#    or  Run ("NONE program arguments");
# First form is just a call to system, and the routine returns after the 
#    program has finished executing.  
# Second form (with 'start') runs the program detached, as appropriate for
#    the operating system: It runs "program arguments &" on UNIX, and 
#    "start program arguments" on WIN95 and WINNT.  If multiple start
#    words are at the beginning of the command, the extra ones are removed.
# Third form (with 'NONE') does not run anything, but prints an error
#    message.  This is provided to allow program names defined in the
#    configuration to flag themselves as unimplemented.
# Return value is a list (pid, exitcode):
#   If process is spawned sucessfully, and I know the PID,
#       return (pid, 0),
#   else if process is spawned sucessfully, but I do not know the PID,
#       return (0, 0),
#   else if process is run, 
#       return (0, exitcode of process)
#   else (I fail to run the requested process)
#       return (0, suitable return code)
#   where return code is 1 if cmdline is null or begins with "NONE" (for
#                      an unimplemented command)
#                     or the return value of the system subroutine.


# Split command line into one word per element, separating words by 
#    one (OR MORE) spaces:
# The purpose of this is to identify latexmk-defined pseudocommands
#  'start' and 'NONE'.
# After dealing with them, the command line is reassembled
    my $cmd_line = $_[0];
    if ( $cmd_line eq '' ) {
	traceback( "$My_name: Bug OR configuration error\n".
                   "   In run of'$rule', attempt to run a null program" );
        return (0, 1);
    }
    if ( $cmd_line =~ /^start +/ ) {
        #warn "Before: '$cmd_line'\n";
        # Run detached.  How to do this depends on the OS
        # But first remove extra starts (which may have been inserted
        # to force a command to be run detached, when the command
	# already contained a "start").
        while ( $cmd_line =~ s/^start +// ) {}
        #warn "After: '$cmd_line'\n";
        return &Run_Detached( $cmd_line );
    }
    elsif ( $cmd_line =~ /^NONE/ ) {
        warn "$My_name: ",
             "Program not implemented for this version.  Command line:\n";
	warn "   '$cmd_line'\n";
        return (0, 1);
    }
    else { 
       # The command is given to system as a single argument, to force shell
       # metacharacters to be interpreted:
       return( 0, system( $cmd_line ) );
   }
}

#************************************************************

sub Run_Detached {
# Usage: Run_Detached ("program arguments ");
# Runs program detached.  Returns 0 on success, 1 on failure.
# Under UNIX use a trick to avoid the program being killed when the 
#    parent process, i.e., me, gets a ctrl/C, which is undesirable for pvc 
#    mode.  (The simplest method, system ("program arguments &"), makes the 
#    child process respond to the ctrl/C.)
# Return value is a list (pid, exitcode):
#   If process is spawned sucessfully, and I know the PID,
#       return (pid, 0),
#   else if process is spawned sucessfully, but I do not know the PID,
#       return (0, 0),
#   else if I fail to spawn a process
#       return (0, 1)

    my $cmd_line = $_[0];

##    warn "Running '$cmd_line' detached...\n";
    if ( $cmd_line =~ /^NONE / ) {
        warn "$My_name: ",
             "Program not implemented for this version.  Command line:\n";
	warn "   '$cmd_line'\n";
        return (0, 1);
    }

    if ( "$^O" eq "MSWin32" ){
        # Win95, WinNT, etc: Use MS's start command:
        # Need extra double quotes to deal with quoted filenames: 
        #    MSWin start takes first quoted argument to be a Window title. 
        return( 0, system( "start \"\" $cmd_line" ) );
    } else {
        # Assume anything else is UNIX or clone
        # For this purpose cygwin behaves like UNIX.
        ## warn "Run_Detached.UNIX: A\n";
        my $pid = fork();
        ## warn "Run_Detached.UNIX: B pid=$pid\n";
        if ( ! defined $pid ) {
            ## warn "Run_Detached.UNIX: C\n";
	    warn "$My_name: Could not fork to run the following command:\n";
            warn "   '$cmd_line'\n";
            return (0, 1);
	}
        elsif( $pid == 0 ){
           ## warn "Run_Detached.UNIX: D\n";
           # Forked child process arrives here
           # Insulate child process from interruption by ctrl/C to kill parent:
           #     setpgrp(0,0);
           # Perhaps this works if setpgrp doesn't exist 
           #    (and therefore gives fatal error):
           eval{ setpgrp(0,0);};
           exec( $cmd_line );
           # Exec never returns; it replaces current process by new process
           die "$My_name forked process: could not run the command\n",
               "  '$cmd_line'\n";
        }
        ##warn "Run_Detached.UNIX: E\n";
        # Original process arrives here
        return ($pid, 0);
    }
    # NEVER GET HERE.
    ##warn "Run_Detached.UNIX: F\n";
}

#************************************************************

sub find_process_id {
# find_process_id(string) finds id of process containing string and
# being run by the present user.  Typically the string will be the
# name of the process or part of its command line.
# On success, this subroutine returns the process ID.
# On failure, it returns 0.
# This subroutine only works on UNIX systems at the moment.

    if ( $pid_position < 0 ) {
        # I cannot do a ps on this system
        return (0);
    }

    my $looking_for = $_[0];
    my @ps_output = `$pscmd`;

# There may be multiple processes.  Find only latest, 
#   almost surely the one with the highest process number
# This will deal with cases like xdvi where a script is used to 
#   run the viewer and both the script and the actual viewer binary
#   have running processes.
    my @found = ();

    shift(@ps_output);  # Discard the header line from ps
    foreach (@ps_output)   {
	next unless ( /$looking_for/ ) ;
        my @ps_line = split (' ');
# OLD       return($ps_line[$pid_position]);
        push @found, $ps_line[$pid_position];
    }

    if ($#found < 0) {
       # No luck in finding the specified process.
       return(0);
    }
    @found = reverse sort @found;
    if ($diagnostics) {
       print "Found the following processes concerning '$looking_for'\n",
             "   @found\n",
             "   I will use $found[0]\n";
    }
    return $found[0];
}

#************************************************************
#************************************************************
#************************************************************

#   Directory stack routines

sub pushd {
    push @dir_stack, cwd();
    if ( $#_ > -1) { chdir $_[0]; }
}

#************************************************************

sub popd {
    if ($#dir_stack > -1 ) { chdir pop @dir_stack; }
}

#************************************************************

sub ifcd_popd {
    if ( $do_cd ) {
        warn "$My_name: Undoing directory change\n";
        &popd;
    }
}

#************************************************************

sub finish_dir_stack {
    while ($#dir_stack > -1 ) { &popd; }
}

#************************************************************
#************************************************************
# Break handling routines (for wait-loop in preview continuous)

sub end_wait {
    #  Handler for break: Set global variable $have_break to 1.
    # Some systems (e.g., MSWin reset) appear to reset the handler.
    # So I'll re-enable it
    &catch_break;
    $have_break = 1;
}

#========================

sub catch_break {
# Capture ctrl/C and ctrl/break.
# $SIG{INT} corresponds to ctrl/C on LINUX/?UNIX and MSWin
# $SIG{BREAK} corresponds to ctrl/break on MSWin, doesn't exist on LINUX
    $SIG{INT} = \&end_wait;
    if ( exists $SIG{BREAK} ) {
        $SIG{BREAK} = \&end_wait;
    }
}

#========================

sub default_break {
# Arrange for ctrl/C and ctrl/break to give default behavior
    $SIG{INT} = 'DEFAULT';
    if ( exists $SIG{BREAK} ) {
        $SIG{BREAK} = 'DEFAULT';
    }
}

#************************************************************
#************************************************************