#!/bin/sh
# /usr/local/bin/lpf_vsl	by Julian Stacey
# Installed from ~jhs/public_html/src/bsd/jhs/bin/local/lpf_vsl/lpf_vsl
# See also:			/usr/share/doc/handbook/printing.{html,latin1}
# Alternative I haven't looked at yet JJLATER:	/usr/ports/print/apsfilter &

# Bug:
#	An ascii file with 2 blank lines, then all indented 3 spaces, gets:
#		"unrecognised format" mail
#	Also a calendar file beinning with /* British calendar .. */ does too.
#
# Bug: date|lpr in /var/log/lpd-errs produces 2 x "[: Tue: unexpected operator"
#	I presume it''s the input format detector going wrong.
#	I never wrote it to work on one line files.
#	I don''t care.
# This filter also blows on some long lines eg .tiff
# & some blows on files with just 1 char (#) in 1st line eg /etc/ttys.

log=/var/log/lpf_vsl
dbg=echo

$dbg " " >> $log
$dbg "`date`: lpf_vsl start" >> $log

tmp=/var/spool/output/lpd.lpf_vsl/$$.tmp	# $$ is PID

# Destroy security for now. JJLATER fix to 775 or 770.
xs chmod 777 /var/spool/output/lpd.lpf_vsl

$dbg "tmp = $tmp" >> $log
#	Keep $tmp same as in /etc/printcap
#	chmod 755 `dirname $tmp` ; chown daemon:daemon `dirname $tmp`

# man printcap:
#	The if filter is invoked with arguments:
#		if [-c] -wwidth -llength -iindent -n login -h host acct-file
# I just ignore arguments delivered here for now.
# JJLATER If the "-h host" component has a bjc in it, generate bjc not pcl.

gs="/usr/local/bin/gs -sPAPERSIZE=a4 -dFIXEDMEDIA -dNOPAUSE -q -sDEVICE=ljet3 -sOutputFile=- -dBATCH -dSAFER -"

# PDF Processing: gs now accepts PDF as input
pdfin="$gs"
$dbg "pdfin = $pdfin" >> $log
# Previously, I used to try to convert PDF to PS with this:
#	pdf2ps="/usr/local/bin/pdf2ps -sPAPERSIZE=a4"
#	Avoid pdf2ps error "exec: gs: not found" in /var/log/lpd-errs
#	PATH=:/bin:/usr/bin:/usr/local/bin ; export PATH
#	Although man pdf2ps says
#		pdf2ps [ options ] input.pdf [output.ps]
#	The second parameter is not actually optional, defaulting to pipe if
#	absent, it's just:	`basename "$1" \.pdf`.ps
#	From /usr/local/bin/pdf2ps: -c save pop -f "$1"

# Suppress Inter Field Seperator so it will not swallow the leading
# white space on the first line of plain text input, between left
# margin & first printable character.
#	IFS=

# Read first characters of first line, to determine file type.
# This syntax was not good enough:
#	read first_line
# because a PCL file contains nulls \0 (a .tiff too).
# None of the shell variables & other unix commands (EG echo printf etc)
# tools I've tried cope with it, so to avoid writing a C program, I swallow
# the first line whole & unchanged with head.

# I lost some PCL data with this:
#	head -1 > $tmp
#	first_line=`cat $tmp`
# So now I swallow whole file (tough if its real big, this proc. wont
# work as a pipe.
cat > $tmp 

yyy=`file -b --dereference $tmp`
$dbg "`date`: Format (from file command): $yyy" >> $log
ok="file command says OK it is"

if [ "$yyy" = "JPEG image data" ]; then	# {
	$dbg "Format (from file command): JPG" >> $log
	# .jpg data as examined from cameras:
	#  Frank:
	#    377 330 377 341   u 344   E   x   i   f  \0  \0   I   I   *  \0
	#    377 330 377 341   w 327   E   x   i   f  \0  \0   I   I   *  \0
	#    377 330 377 341   p 371   E   x   i   f  \0  \0   I   I   *  \0
	#  Julian:
	#    377 330 377 341 023   p   E   x   i   f  \0  \0   I   I   *  \0
	#    377 330 377 341 023   .   E   x   i   f  \0  \0   I   I   *  \0
	#    377 330 377 341 023   D   E   x   i   f  \0  \0   I   I   *  \0
	# Notice that "I   I   *  \0" as at start of tiff.
	/usr/local/bin/jpegtopnm < $tmp | /usr/local/bin/pnmtops \
		 | $gs && rm $tmp && exit 0
	# Alternate filter not tried:
	#	/usr/local/bin/djpeg -pnm < $tmp | \
	#		/usr/local/bin/pnmtotiff > $tmp.tmp
	#	/usr/local/bin/tiff2ps -a $tmp.tmp | \
	#		$gs && rm $tmp $tmp.tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }

elif [ "$yyy" = "TIFF image data, little-endian" ]; then	# {
	# JJLATER add code to allow for big endian.
	$dbg "Format (from file command): TIFF" >> $log # From eg http://www.berklix.com/scanjet/
	/usr/local/bin/tiff2ps -a $tmp | $gs && rm $tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }

else 	#{
	$dbg "Format testing (with file command) for ASCII" >> $log
	#  Hope its not a binary that will consume masses of paper.
	#  Plain text: print a form at the end to eject the last page.
	# EscE: Initialise printer to default state with EscE,
	#	in case last use has left it in a mangled state
	#	EG host computer might have died in mid print.
	# EscE: Initialise again for good measure:
	#	maybe the printer might not recognise first Reset if it
	#	gets swallowed into an embedded sequence, or perhaps the
	#	printer is on a flakey dynamic multiplexer, & the first
	#	couple of characters get lost, while the multiplexer is
	#	switching the printer to a different host computer.
	#	Unlike with a hayes escape, a pcl scape does not seem to
	#	require a quiet period afterward, according to the big
	#	HP PCL manual.
	# Esc&k2G: Tell printer to treat LF as CR+LF.
	# EscE	Reset to default for next user (that on a dynamic
	#	multiplexer may not even be Unix+lpd )
	#	A printer reset also has the side effect of causing
	#	the printer to flush the last partial page (if any), whereas
	#	If we sent a FF instead, we could end up printing an
	#	extra blank page, if either { (A) the input stream of data had
	#	terminated with a page aligning FF, or (B) the last line feed
	#	happened to align the printer to end of page. }

	# File reports various sorts of Ascii, eg:
	#	cd /etc ; file -b * | grep -i ascii | sort | uniq
	#		ASCII C++ program text
	#		ASCII English text
	#		ASCII make commands text
	#		ASCII text
	#	See also: man file & /usr/share/misc/magic
	if [ "$yyy" = "ASCII text" ]; then	# {
		$dbg "$ok $yyy" >> $log ; printf "\033E\033&k2G" && cat $tmp && \
			printf "\033E" && rm $tmp && exit 0
		# }
	elif [ "$yyy" = "ASCII English text" ]; then	#{
		$dbg "$ok $yyy" >> $log ; printf "\033E\033&k2G" && cat $tmp && \
			printf "\033E" && rm $tmp && exit 0
		# }
	elif [ "$yyy" = "ASCII make commands text" ]; then	#{
		$dbg "$ok $yyy" >> $log ; printf "\033E\033&k2G" && cat $tmp && \
			printf "\033E" && rm $tmp && exit 0
		# }
	elif [ "$yyy" = "ASCII C++ program text" ]; then	#{
		$dbg "$ok $yyy" >> $log ; printf "\033E\033&k2G" && cat $tmp && \
			printf "\033E" && rm $tmp && exit 0
		# }

	# In a mail Removing Received: lines changes file report
	# From	"ASCII mail text"
	# To:	"ASCII HTML document text" OR "ASCII text" if a non MIME
	# so JJLATER trap "ASCII mail text".

	# I could also add Many other strings such as
	#	"POSIX shell script text executable"
	#	"troff or preprocessor input text"
	# But troff should be formatted first anyway.
	# I could extend to use a swith as below,
	# but I suspect below broke cos of spaces in strings
	# switch ( "$yyy") #{
	#         case "ASCII text":
	#         case "ASCII English text":
	#         case "ASCII make commands text":
	#         case "ASCII C++ program text":
	# 		$dbg "Format OK to print" >> $log
	#                 breaksw
	#         case default:
	# 		$dbg "Format not ASCII so skipped" >> $log
	#                 breaksw
	#         endsw   #}

	else	# {
		$dbg "No more file command Ascii tests, end of test with file command." >> $log
	fi	# }
fi	# }

# Some .tif, the the 1st line from head -1 can be 360K long. 
# So something below here breaks, but hopefully .tiff will be caugh above.

$dbg "Start of own ($0) analysis of first line." >> $log

first_line=`head -1 $tmp`

# Restore Inter Field Seperator to space tab nl
# IFS="\040\011\012"

first_two_chars=`expr "$first_line" : '\(..\)'`
first_four_chars=`expr "$first_line" : '\(....\)'`
first_six_chars=`expr "$first_line" : '\(......\)'`

# If the input text is
#	/* British calendar <en_UK.ISO8859-1/calendar.events>
# The output is:
#	+ expr '/* British calendar <en_UK.ISO8859-1/calendar.events>' : '\(...\)'
#	+ [ /Distfile /FIRE /ad4s1 /ad4s2 /ad4s3 /ad4s4 /ad4s4dcrypt \
#		/ad6s1 /ad6s2 /ad6s3 /ad6s4 /av /bin /boot /cdrom /cdrw \
#		/chroot /compat /dev /devfs /devusb /dvd /dvdw /entropy \
#		/etc /home /host /lib /libexec /media /mnt /opt /pri \
#		/proc /pub /rescue /root

if [ "$first_four_chars" = "%!PS" ]; then	# {
	$dbg "Format (from own ($0) analysis): POSTSCRIPT from GROFF" >> $log
	# PostScript: use Ghostscript to convert and print it
	# groff generated .ps files start: "%!PS-Adobe-3.0"
	# I do not need	to emit $first_line for gs, as I tested ghostview
	# without first line, & it displayed OK, & ghostview uses gs internaly.
	# Example:
	#	%!PS-Adobe-3.0
	#	%%Creator: Applixware
	$gs $tmp && rm $tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }


elif [ "$first_four_chars" = "%!ps" ]; then	# {
	$dbg "Format (from own ($0) analysis): POSTSCRIPT not from GROFF" >> $log
	$gs $tmp && rm $tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }


elif [ `expr "$first_line" : '\(...\)'` = '.\"' ]; then	# {
	$dbg "Format (from own ($0) analysis): GROFF" >> $log
	# Note the backslash while needed in the .rof file, used to be
	# discarded by
	#	read first_line
	# before I received it here, however now using:
	#	cat > $tmp ; first_line=`head -1 $tmp`
	# it is there.
	cat $tmp | groff -U -b -s -t -Tps -dformat=ps -dumlauts=p | \
		 $gs && rm $tmp && exit 0
	# I could instead generate pcl direct from groff.
	$dbg "Format has failed" >> $log
	exit 1
	# }


elif [ `expr "$first_line" : '\(........\)'` = '%PDF-1.2' ]; then	# {
	$dbg "Format (from own ($0) analysis): PDF-1.2" >> $log	# from ps2pdf, groff, external
	$pdfin $tmp && rm $tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }


elif [ `expr "$first_line" : '\(........\)'` = '%PDF-1.3' ]; then	# {
	$dbg "Format (from own ($0) analysis): PDF-1.3" >> $log	# From external
	# From Phil Walters I received %PDF-1.3
%âãÏÓ
	$pdfin $tmp && rm $tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }


elif [ `expr "$first_line" : '\(........\)'` = '%PDF-1.4' ]; then	# {
	$dbg "Format (from own ($0) analysis): PDF-1.4" >> $log	# From external
	# From Companies House change of address form
	#	%PDF-1.4?%??????
	#	2544232302EECD00
	#	5046D1E4D523F3DA
	# 0D=Carriage Return
	# From GEA Prog 2008_04.pdf
	#	%PDF-1.4?
	#	254423230
	#	5046D1E4A
	# 0A=New Line
	$pdfin $tmp && rm $tmp && exit 0
	#	GNU Ghostscript 7.07: Unrecoverable error, exit code 1
	#	Segmentation fault
	# However xpdf can display it, & can export it to a .ps file
	# which this can print. So I discovered this works OK generating .ps
	#	 pdftops $tmp - > t.ps
	# So merge the 2:
	# /usr/local/bin/pdftops $tmp - | $pdfin && rm $tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }


elif [ `expr "$first_line" : '\(........\)'` = '%PDF-1.5' ]; then	# {
	$dbg "Format (from own ($0) analysis): PDF-1.5" >> $log
	# Never seen a 1.5, but presumably they exist.
	/usr/local/bin/pdftops $tmp - | $pdfin && rm $tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }


elif [ `expr "$first_line" : '\(........\)'` = '%PDF-1.6' ]; then	# {
	$dbg "Format (from own ($0) analysis): PDF-1.6" >> $log
	# from mk@, scanner@msd.
	/usr/local/bin/pdftops $tmp - | $pdfin && rm $tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }


elif [ `expr "$first_line" : '\(....\)'` = '%PDF' ]; then	# {
	$dbg "Format (from own ($0) analysis): PDF Unknown Version" >> $log # Not from ps2pdf, upper case
	$pdfin $tmp && rm $tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }


elif [ `expr "$first_line" : '\(....\)'` = '%pdf' ]; then	# {
	$dbg "Format (from own ($0) analysis): pdf unknown version" >> $log # not from ps2pdf, lower case
	$pdfin $tmp && rm $tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }


elif [ $first_two_chars = '\033E' ]; then	# {
	( echo "PCL detected by /usr/local/bin/lpf_vsl" | mail -s "PCL from lpf_vsl" jhs )
	$dbg "Format (from own ($0) analysis): PCL" >> $log # Escape E
	#  HP/PCL: The big HP PCL 5 book recomends all PCL files should
	#  start with this, & ghostscript generates it, so it's a safe bet.
	# Example:	"^[E^[&l26A" from gs
	# I hope data in first line is not longer than $first_line can read !
	# Some PCL files go berserk, & the whole damn file has no \n,
	# or just one at end.
	cat $tmp && rm $tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }

elif [ `expr "$first_line" : '\(...\)'` = 'II*' ]; then	# {
	$dbg "Format (from own ($0) analysis): TIFF" >> $log # From eg http://www.berklix.com/scanjet/
	/usr/local/bin/tiff2ps -a $tmp | $gs && rm $tmp && exit 0
		# The match pattern should have a null (shows in vi as ^@),
		# ie 'II*\0'
		# but I haven't got the elif to match on that.
	$dbg "Format has failed" >> $log
	exit 1
	# }


elif [ $first_six_chars = 'GIF87a' ]; then	# {
	# from XV as frame grabber on FreeBSD-7.1-RELEASE
	$dbg "Format (from own ($0) analysis): GIF"  >> $log
	/usr/local/bin/gif2ps < $tmp | $gs && rm $tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }

elif [ $first_six_chars = 'GIF89a' ]; then	# {
	$dbg "Format (from own ($0) analysis): GIF"  >> $log
	/usr/local/bin/gif2ps < $tmp | $gs && rm $tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }

elif [ "`echo \"$first_line\" | file -b -`" = "JPEG image data" ]; then	# {
	# .jpg data as examined from cameras:
	#  Frank:
	#    377 330 377 341   u 344   E   x   i   f  \0  \0   I   I   *  \0
	#    377 330 377 341   w 327   E   x   i   f  \0  \0   I   I   *  \0
	#    377 330 377 341   p 371   E   x   i   f  \0  \0   I   I   *  \0
	#  Julian:
	#    377 330 377 341 023   p   E   x   i   f  \0  \0   I   I   *  \0
	#    377 330 377 341 023   .   E   x   i   f  \0  \0   I   I   *  \0
	#    377 330 377 341 023   D   E   x   i   f  \0  \0   I   I   *  \0
	# Notice that "I   I   *  \0" as at start of tiff.
	$dbg "Format (from own ($0) analysis): JPG"  >> $log
	/usr/local/bin/jpegtopnm < $tmp | /usr/local/bin/pnmtops \
		 | $gs && rm $tmp && exit 0
	# Alternate filter not tried:
	#	/usr/local/bin/djpeg -pnm < $tmp | \
	#		/usr/local/bin/pnmtotiff > $tmp.tmp
	#	/usr/local/bin/tiff2ps -a $tmp.tmp | \
	#		$gs && rm $tmp $tmp.tmp && exit 0
	$dbg "Format has failed" >> $log
	exit 1
	# }

fi	# }

# About Trailing Form Feed:	JJLATER
# - The trailing Reset (\033E) Form Feed = New Page =Control L = (\014)
#  should be selective, sampling data at end of file, but it's not (yet),
# - Printers can be set up to also require FF or not,
#   generatin their own instead. If both do FF you get a wasted page.
# - If neither do it, the job doesn't print till next job, or until you manually
#   send an empty page.
# - My HP is set up to generate its own trailing NP=FF
# - My Brother is set up to want an FF from computer.
# - Applixware.pcl does not have a terminating FF.
# - Groff.pcl produced by groff -U -b -s -t -Tlj4 -dformat=lj4
#	thing.rof > thing.pcl ) Have a terminating ^L^[E
# - .pcl made by groff -U -b -s -t -Tlj4 -dformat=lj4  thing.rof > thing.pcl
#    Ends with ^L^[E (& has bits of words in, not pure bitmap), for CV ~55K.
# - .pcl made by gs -sDEVICE=ljet3 -sOutputFile=thing.pcl -- thing.ps
#    also ends in ^L^[E (but seems to have no strings, just bit map, & is
#    much larger, for my CV ~650K)

# JJLATER add a trap to rm $tmp

# Should never get this far.
echo "`date`: /usr/local/bin/lpf_vsl failure unrecognised format $tmp" | mail -s "lpf_vsl failure" root
echo "`date`: /usr/local/bin/lpf_vsl failure unrecognised format $tmp" >> $log
exit 2	# only on error.

# /usr/share/doc/handbook/handbook93.html:
# The text filter, confusingly called the input filter in LPD documentation,
# handles regular text printing. Think of it as the default filter. LPD expects
# every printer to be able to print plain text by default, and it is the text
# filter's job to make sure backspaces, tabs, or other special characters do not
# confuse the printer. If you are in an environment where you have to
# account for printer usage, the text filter must also account for pages
# printed, usually by counting the number of lines printed and comparing that
# to the number of lines per page the printer supports. The text filter is
# started with the following argument list:
#	[-c] -wwidth -llength -iindent -n login -h host acct-file