Newsgroups: comp.os.linux.announce,comp.os.linux.admin,comp.answers,news.answers From: gtaylor@cs.tufts.edu (Grant Taylor) Subject: Linux Printing HOWTO Keywords: Linux, printing, lpr Summary: HOWTO on printing under Linux Followup-To: poster Approved: linux-announce@tc.cornell.edu (Matt Welsh) Archive-name: linux/howto/printing Last-Modified: 1 Dec 93 The Linux Printing HOWTO by Grant Taylor and Brian McCauley Contents: 1 Intro 2 Printing and Faxing Software 3 LPR 4 Previewing 5 Ascii Translation 0.1 Intro This is the Printing HOWTO for Linux. This is a member of the second generation of Linux FAQS. The original Linux FAQ gradually became a monolithic beast, and has now been re-written in a new syle, refering to HOWTOs for details. This HOWTO details how to properly set up most types of printers, how to configure the software to make them print well, what types of fax software are available, and how to preview many types of printable output. Other HOWTOs detail networking, kernel hacking, etc. All HOWTOs are found in several places, the official location being sunsite.unc.edu:/pub/Linux/docs/HOWTO, and another being rtfm.mit.edu. Please send comments, bug reports, etc, to printing@god.ext.tufts.edu. This address goes to both Brian and Grant. If you know anything or can point us towards any useful programs not in this document that should be, please let us know! Grant has a mail server set up on his machine. It stores documents and other things related to printing and previewing with Linux, including the latest and greatest version of this document. Mail to listserv@god.ext.tufts.edu with a body of ``info'' will have you sent a list of available files. A body of ``get file1 [ file2 file3... ] '' will send you specific files. 0.1.1 Summary of changes * Now in La [ me ] TeX! * New mailserver info. * Assorted little changes 0.2 Printing Software These sections describe printing software known to be available for Linux. Note that most Un*x printing software can be easily compiled under Linux. 0.2.1 text Under Linux, or under most any Un*x operating system, the easiest thing to print with most printers is plain ASCII. Any way you can arrange to send the text to the printer is perfectly valid. If you have a serial printer, then try devices /dev/ttys?, /dev/ttyS?, /dev/cua?, etc, and if you have a regular parallel hookup, use /dev/lp?. Typing ``cat file >/dev/????'' should do it. You may need root privileges, or you may wish to chmod your printer device if this is all you ever print. (Note the security hole there if you've more than one user) Some printers (ie, HP DeskJet) want dos-style end of lines: newline+carriage return. These printers will show the 'staircase effect'; further information on how to deal with it is in ``How do I prevent the staircase effect,'' 0.2.2 pr Most plain ascii files in the un*x world have a tendency to be just that - plain, unformatted ascii with no page breaks or anything else to make a printed copy look nice and not have lines printed over perferations. The answer to this problem is to run your text through a formatter such as pr. Pr is a standard un*x utility designed to format plain text for printing on a line printer. The usual appearance of the resulting formatted text has a header and/or footer, page numbers, the date, possibly margins, double spacing, etc. As is common with un*x utils, pr has a bazillion options. They are detailed in the man page. 0.2.3 Postscript Printing almost anything other than plain text under Un*x usually involves the ability to print Postscript. If you have a Postscript printer, you're all set. But for most, this is not so easy. The established way to interpret Postscript under Linux is to use ghostscript, which, like nearly everything else, comes from the GNU project. Ghostscript is a Postscript interpreter which accepts Postscript input and generates output appropriate for X displays, most printers, some specialized display hardware, and fax software. The following devices are available as of ghostscript 2.6.1pl4: linux PC vga using linux vgalib # x11 X Windows version 11, release >=4 [Unix and VMS only] # Printers: # * appledmp Apple Dot Matrix Printer (Imagewriter) # bj10e Canon BubbleJet BJ10e # * bj200 Canon BubbleJet BJ200 # * cdeskjet H-P DeskJet 500C with 1 bit/pixel color # * cdjcolor H-P DeskJet 500C with 24 bit/pixel color and # high-quality color (Floyd-Steinberg) dithering # * cdjmono H-P DeskJet 500C printing black only # * cdj500 H-P DeskJet 500C (same as cdjcolor) # * cdj550 H-P DeskJet 550C # * declj250 alternate DEC LJ250 driver # + deskjet H-P DeskJet and DeskJet Plus # * dfaxhigh DigiBoard, Inc.'s DigiFAX software format # * dfaxlow DigiFAX low (normal) resolution # djet500 H-P DeskJet 500 # * djet500c H-P DeskJet 500C # epson Epson-compatible dot matrix printers (9- or 24-pin) # + eps9high Epson-compatible 9-pin, interleaved lines # (triple resolution) # * epsonc Epson LQ-2550 and Fujitsu 3400/2400/1200 color printers # * escp2 Epson ESC/P 2 language printers, including Stylus 800 # + ibmpro IBM 9-pin Proprinter # * jetp3852 IBM Jetprinter ink-jet color printer (Model #3852) # + laserjet H-P LaserJet # * la50 DEC LA50 printer # * la75 DEC LA75 printer # * lbp8 Canon LBP-8II laser printer # * ln03 DEC LN03 printer # * lj250 DEC LJ250 Companion color printer # + ljet2p H-P LaserJet IId/IIp/III* with TIFF compression # + ljet3 H-P LaserJet III* with Delta Row compression # + ljet4 H-P LaserJet 4 (defaults to 600 dpi) # + ljetplus H-P LaserJet Plus # * m8510 C.Itoh M8510 printer # * necp6 NEC P6/P6+/P60 printers at 360 x 360 DPI resolution # * nwp533 Sony Microsystems NWP533 laser printer [Sony only] # * oki182 Okidata MicroLine 182 # paintjet H-P PaintJet color printer # * pj alternate PaintJet XL driver # * pjxl H-P PaintJet XL color printer # * pjxl300 H-P PaintJet XL300 color printer # * r4081 Ricoh 4081 laser printer # * sparc SPARCprinter # * t4693d2 Tektronix 4693d color printer, 2 bits per R/G/B component # * t4693d4 Tektronix 4693d color printer, 4 bits per R/G/B component # * t4693d8 Tektronix 4693d color printer, 8 bits per R/G/B component # * tek4696 Tektronix 4695/4696 inkjet plotter #%* trufax TruFax facsimile driver [Unix only] # File formats and others: # bit A plain "bit bucket" device # bmpmono Monochrome MS Windows .BMP file format # bmp16 4-bit (EGA/VGA) .BMP file format # bmp256 8-bit (256-color) .BMP file format # bmp16m 24-bit .BMP file format # gifmono Monochrome GIF file format # gif8 8-bit color GIF file format # pcxmono Monochrome PCX file format # pcxgray 8-bit gray scale PCX file format # pcx16 Older color PCX file format (EGA/VGA, 16-color) # pcx256 Newer color PCX file format (256-color) # pbm Portable Bitmap (plain format) # pbmraw Portable Bitmap (raw format) # pgm Portable Graymap (plain format) # pgmraw Portable Graymap (raw format) # ppm Portable Pixmap (plain format) # ppmraw Portable Pixmap (raw format) # * tiffg3 TIFF/F (G3 fax) Lines beginning with a ` # ' are drivers included with the gs261 source distribution. Drivers marked with % require commercial software to work Installations of ghostscript are available from several places: A full, and presumably correct, installation is available in the SLS; note that this binary requires libX.so.???, as it includes the X11 display driver. The `official' installation is to obtain the sources and build it yourself: prep.ai.mit.edu:/pub/gnu/ghostscript-xxxx.tar.gz and ghostscript-fonts-xxxx.tar.gz This is probably best, as you can get the latest version (currently 2.6.1pl4 (the four patches are separate, they are available from my mail-server as gs261-patches)). A patch which uses the Linux svgalib, and hence does not require X for previewing is available from ws105.zfn.uni-bremen.de:/pub/gs261-linuxdriver.sh. A minimal binary installation of ghostscript and several other packages needed for printing the Linux documentation is available as sunsite.unc.edu:/pub/linux/apps/tex/texmin/texmin-0.1.tar.z. Note that this does not contain any postscript fonts. (Nor do you need them to print dvi (aka [ La ] TeX).) The main Ghostscript documentation is contained in the file use.doc, either in the source directory, or .../lib/ghostscript/doc/use.doc if you haven't the sources. To print ps, first determine your driver name with: gs -help which lists installed drivers. If the device you need is not listed, you must compile gs yourself from the source distribution (do not panic. do follow the instructions in make.doc. you will need 5 or 6 megs of space to build it in.) Then type gs -dNOPAUSE -sDEVICE=?????? -sOutputFile=/dev/???? file.ps and your output should (hopefully) appear at your printer. Those of you with non-US paper sizes may wish to build gs yourself with the right default, or you may use the -sPAPERSIZE=a4 option. Ghostscript may be used to print at most of the resolutions your printer supports; -r300, -r150, -r360x180 are examples of the option used to control this. Dot matrix printers in particular need to choose an appropriate resolution, since they do not typically run at the most common 300. Note that versions 2.6 and greater of ghostscript have more drivers. 0.2.4 fonts (This section contains font information not specific to any ghostscript driver (nor even specific to ghostscript). Font information pertaining to the X11 gs drivers (and thus to ghostview) is included in the ghostview section under previewing) All versions of ghostscript come with assorted public-domain fonts, most of which were generated from bitmaps, and are therefore of relatively poor quality. However, ghostscript can use any PostScript Type 1 or 3 fonts you may have handy. For example, the Adobe Type Manager (for any platform other than mac (a conversion util for mac ps fonts was on simtel-20 in the windows3 directory)) comes with fonts you can use. Place the fonts (typically *.pc?) in .../lib/ghostscript/fonts/ and add to your fontmap lines such as: /Courier (com_____.pfb) ; A full fontmap file for the ususal set of fonts included with the Adobe Type Manager is available as fontmap.atm from the printing mailserver. TimesNewRomanPS and ArialMT fonts are interchangable with Times Roman and Helvetica, so you can alias them this way if this is what you have. Adobe Type 1 fonts may be found on the net in several places: ftp.cica.indiana.edu:/pub/pc/win3/fonts and archive.umich.edu:/msdos/mswindows/fonts I have not looked in these places, but this information is lifted straight from the comp.fonts FAQ (which you should read if fonts are a thing for you. I'll be glad to mail you a copy). Conversion between various font types is tricky. Ghostscript comes with the tools needed to take a bitmap (hopefully large) and make a scalable ps font. Groff comes with the tools to allow use of tfm/mf (TeX) and pfb (Type 1) fonts in *roff documents. X11R5 includes several font utilities and Type 1 rendering code contributed by IBM. I have used none of the above, but they all come with manpages, so read them. Other font conversion utilities are listed in the comp.fonts FAQ. Also look into the package fontutils on prep.ai.mit.edu:/pub/gnu 0.2.5 faxing If you have a fax modem, you can arrage to fax things out (and in), including PostScript, dvi, ascii, etc... Arranging for e-mail to fax itself somewhere is also straightforward. Fax modems correspond to one of two types: Classes 1 and 2. Class one modems are fairly common and have less of what goes on in a fax supported in firmware (thus the software has to do more and do it carefully. supporting such timing-critical things under a preemptive multitasking environment such as Linux is tricky at best). The class 1 standard is EIA 578. Class 2 modems tend to be more expensive and comply with the standard EIA 592. Mention of your fax modem's level of support should be in its manuals. Do not confuse class and group. You would be hard pressed to find a non-group III fax modem these days (but you probably could). Fax software which runs under linux must convert input in whatever format into a Group III compatible image format for transmission. As usual, Ghostscript does the job. The device tiffg3 generates standard g3/tiff encoded fax messages. You must compile them in if they are not there already; this is yet another reason to get the source. Some fax sofware on commercial platforms can use the display PostScript renderer to render the fax image; this is not yet an option under Linux. The GNU program netfax supports Class 2 fax modems only. It builds nearly out of the box under linux. I have patches for version 3.2.1; you can get them from the printing mail server as ``netfaxpatch''. It runs a server and accepts jobs into a queue from the local machine and the net. netfax is in use at MIT. It accepts postscript, dvi, and ascii, and can be configured to work as an email gate. Documentation in the package is limited to compile notes and man pages for the individual programs; actual installation tips are few and far between. prep.ai.mit.edu:/pub/gnu/fax-3.2.1.tar.gz FlexFax, available from sgi.com(192.48.153.1):/sgi/fax/?????.src.tar.Z, is written in C++, and thus requires g++ to build. FlexFax supports class 1 and 2 modems, uses ghostview (or a display postscript server) for rendering, and generally is more complete, or at least general than the somewhat MIT-specific netfax. It also can run in place of getty and either receive a fax or spawn a login as appropriate for the type of call. It includes precise instructions on configuring smail3 (and other mail agents) to send mail for so-and-so@12345678.fax to the fax at 12345678. This is the package I recommend if you have real faxing needs, as it is properly documented and full-featured. Version 2.x of FlexFax is in real release now and fully supports Linux - just type ./configure and make. mgetty+sendfax is a Linux and SCO-specific getty for faxmodems and a simple sendfax program. This package is on sunsite. This is a good package for those who need to send a fax every now and then. modgetty (which uses tcl, an interpreted language) can receive faxes. Try harbor.ecu.purdue.edu:/pub/tcl 0.2.6 *roff, man pages Man pages can sometimes be printed straight from the cat pages which come with the SLS (in lieu of the normal nroff source pages) as though they were a normal text file (which they mostly are). However, many printers do not like the VT escape codes for highlighting and what not that find themselves imbedded in these. A filter for this purpose comes with the net-2 lpd package. It is called lpf. If you have the nroff source to the page (the finding of which I highly recommend) you can say man -t foobar |lpr and your man program will (hopefully) format the man page using groff into PostScript, which will then be sent to your lpd and on to the printer. This form of man page output looks MUCH better than the plain ASCII version. Unfortunately, this depends very much on which man program and what supporting software you have installed. If your man doesn't do this, you might try the perl version of man, available near sunsite.unc.edu:/pub/Linux/system/Manual-Pagers/ It is written entirely in perl, and is thus easily customizable (perl being an interpreted language reminiscent of C and sh). You can also find the nroff source file in the man directories (most versions of man have an option to just spit out the filename) and do it yourself, into any format supported by groff: groff -mandoc -T ascii,dvi,ps,X100,X75,latin8 foobar.1 [ |lpr ] 0.2.7 mpage The package mpage converts plain text into postscript and/or prints more than one page onto each peice of paper given postscript or text. It is available as wuarchive.wustl.edu:/pub/mirrors/unix-c/postscript/mpage.tar-z or thereabouts. Note that wuarchive uses the -z suffix to mean .Z, ie, compress, not GNUzip or freeze. ``man -t foobar | mpage'' will send a 2-up (depending on the environment variable MPAGE) version of the man page to lpr and its postscript interpreter. This saves paper and speeds up printing. 0.2.8 a2ps A2ps will take ASCII and turn it into a nice postscript document with headers and footers and page numbers, printed two pages on one (or otherwise, if you say so). A2ps does a very nice job at this. It is available at the same place mpage is. 0.2.9 enscript Enscript is a program which does basically the same thing as a2ps. I do not know where to get it. It comes with most commercial Un * xes. A clone version of enscript is called nenscript, available as /pub/Linux/system/Printing/nenscript-1.13++.bin.tar.z or /pub/Linux/system/Printing/nenscript-1.13++.tar.z on sunsite. 0.2.10 gslp Gslp is one of the uilities which comes with ghostscript 2.6.x and purports to do the same ascii --> ps conversion as enscript and a2ps. I have not used it, but the docs say that gs -q -sDEVICE=????? -dNOPAUSE -- gslp.ps text.file [ options ] should do the trick. (gslp.ps is the actual program, which is written in postscript. Here it is run with the argument `text.file.' Postscript is in many respects a programming language more than a mere printer command language.) Further documentation is in the file gslp.ps. There is a script file which does the above for you. 0.2.11 psutils Those of you who deal with large amounts of postscript may wish for more utility programs than come with the SLS. There are probably millions of little programs which do things to your Postscript. A representative package of them may be found in achilles.doc.ic.ac.uk:/tex/inter/psutils/ or guardian.cs.psu.edu:/pub/src/psutil.tar.Z These handle page selection, double-sided printing, booklet creation, etc. Most large ftp sites (eg, wuarchive.wustl.edu, ftp.uu.uunet) will have many such packages to choose from. 0.2.12 tex/dvi [ La ] TeX is the text formatting package used by many in the academic world and elsewhere. TeX works much like any other compiler -- source code is run through the program ``tex'' to become a .dvi file (analogous to an .o object file in C) which can be further manipulated to produce printed output. This manipulation of a dvi (DeVice Independant) file usually takes a little bit of doing. It is well worth it; TeX's output is of professional quality. For those in the real world who cannot afford a dvi understanding printer, it is usually best to convert the dvi into postscript that you can pipe into ghostscript or lpr. The SLS comes with a functioning installation of both TeX and dvips. Typing ``dvips -f1 file.dvi | lpr'' will do it. Dvips responds to either command line arguments or a file /usr/TeX/lib/tex/ps/config.ps (in the SLS TeX distribution, at least) in which you can arrange to have dvips automatically send its output to lpr. Thus ``dvips file.dvi'' will do everything that needs to be done. Note that some .dvi's may include PostScript graphics in the dvips stage rather than the tex stage of the game; if they are not found, you will get a hole instead of a picture. This follows naturally from the object file analogy above. Usually, pre-made documentation in this form has a makefile or script to do everything for you. The LILO documentation is an example of this. Dvips has several interesting options; for exemple, ``dvips -r1 file.dvi'' will print it out backwards. We deskjet users love this one. If all you are given is a file with a .tex ending, try either ``tex file.tex'' or ``latex file.tex''. One of these is bound to work. Then you have a dvi. (You may have to run it twice for indexing) 0.2.13 eps Eps is a program which converts dvi files directly into the standard Epson printer language; thus it is a dvi driver for epsons. I beleive it supports MakeTeXPK, the automatic font rendering system used by many dvi drivers, and is available as sunsite.unc.edu:/pub/Linux/apps/tex/eps-060.tar.gz or in ftp.ctd.comsat.com:/pub/. Note that it is still coming out with new versions fairly often, so there may be a newer version than 060. 0.2.14 dvilj For LaserJet owners, there is a separate program which I beleive is in the SLS that will take dvi and convert it directly into your printer's language (PCL). It is called dvilj2p or dvilj, and if not on tsx or sunsite, is certainly available on ftp.uu.net. (Description by Nils Rennebarth) Its a nice driver, but a little out of fashion in the sense that configuration (especially of font-paths) font-paths is not very flexible and that it doesn't support virtual fonts (at least the version 0.51 not). The biggest advantage over the dvips/ghostscript combo is that it uses downloadable fonts which: a) reduces data transmission to the printer drastically, which makes the printer usable even on a serial line. b) reduces printer-memory-requirements drastically. A standard Laserjet with 512k memory is able to print almost every TeX-document. It has support for double side printing and most options you expect a driver to have. It compiles cleanly and worked flawlessly on our diverse hardware here. 0.2.15 texinfo This is the native documentation format of the GNU project. Emacs can be coerced into producing an info file out of this, and TeX can produce nice printed documantation from the same file. It is a bit of a stretch for both systems, but it works. It is really just TeX source which expects the macro file texinfo.tex to be installed on your system. Just do ``tex '' twice (for index generation purposes), and you end up with a plain dvi file, to print or preview at your leisure. In emacs, you can also do M-x texinfo-format-buffer to convert the texinfo file into an info file viewable with emacs M-x info or an info viewer of your choice. There are also separate programs which read and format info from a texinfo file. These are available in prep.ai.mit.edu:/pub/gnu/ 0.2.16 hardware and drivers There are two ways the kernel driver may be used to run the parallel ports. One, the original, is the polling driver. The other is the interrupt driver. In principle, the interrupt driver only deals with the port when it gets an interrupt and should therefore be more efficient. In practice, people have found that it depends on the machine. It probably doesn't make too much difference in most situations. For the polling driver, you may adjust its polling frequency with the program lptune without kernel twiddling. The actual driver is in .../linux/kernel/chr_drv/lp.c. To choose the interrupt driver rather than the polled, use the program lptune to set it. (Lptune is available on sunsite, or from the printing mail server.) Just put the appropriate line in /etc/rc.local Seven is the usual "LPT1:" IRQ, 5 is "LPT2:" for an AT class machine. Note that if your printer is on lp0, the interrupt driver will probably not work. The first parameter should already contain the correct i/o base address. Many bios boot screens have this information if you need it. 0.2.17 Printer device names On an XT bus system LPT1: becomes /dev/lp0 (major=6, minor=0), on an AT LPT1: becomes /dev/lp1 (major=6, minor=1). To be more precise: Name Major Minor I/O address lp0 6 0 0x3bc lp1 6 1 0x378 lp2 6 2 0x278 For a serial printer use the ttyS? (or ttys?, if you follow that scheme) device, not the cua? device. The ttyS? devices have major 4; the cua? devices have major 5. 0.3 LPR This section used to the the lpd-HOWTO then it became part of the linux-HOWTO and now it is part of printing.how-to. Who knows where it will go next? 0.3.1 What lpr and lpd do Most un*x systems use lpd (or the System V variant lp), the line printer daemon, and friends to spool print jobs and run them through all needed filters. While line printers are certainly on their way out, spooling print jobs and running them through filters is definitely a convenient thing. Thus lpr. (subliminal note: PLEASE find and read the lpr, lpd, printcap, lpc, lpq, and lprm man pages. They are in the source dist of the net-2, if you haven't got them.) Lpd and family are available in several places. The new Linux net-2 package contains a working lpd. I now use this one with my stock 0.99pl10 kernel after following the directions in Matt Welsh's net-2 HOWTO to get my NET-2 going (you only need loopback). A binary package which can be coaxed into working with the old net things (pre 0.99pl10 kernel) is in the SLS. Note a subtle difference between the two versions: the old one placed lock files and such in /var/spool, while the net-2 version requires the directory /var/spool/lpd to exist for it's lock file. You will need to edit /etc/printcap to configure lpd to use your printer. Setting up lpd to accept postscript and print it is not very difficult; simply make a shell script fiter (qv). Setting up lpr correctly is definitely worth the trouble if you are going to do any printing at all. Of course, if all you ever print is the occasional man page or instruction book, then you can probably get by without it (but even then it's nice). Ghostscript 2.6.x also comes with a complicated script and utilities to generate printcap entries and manage multiple types of queues for a printer. I have not tested these, but I assume they can be made to work. If you are in a large-scale environment with varied printing needs, these may be worth a look. 0.3.2 Where do the files go and what should their permissions be? There is quite a bit of variation between the various releases but I'll try where possible to offer solutions that are universally applicable. Put /etc/lpd in your /etc/rc or /etc/rc.local (usually in /etc/rc.local after you start syslogd (if you use syslogd)). Set the group fields of the file permissons/ownership as follows: -rwxr-s--- 1 root lp 37892 Nov 19 23:32 /etc/lpd -rwx--s--x 1 root lp 21508 Nov 19 23:32 /usr/bin/lpc -rws--s--x 1 root lp 17412 Nov 19 23:32 /usr/bin/lpq -rwx--s--x 1 root lp 17412 Nov 19 23:32 /usr/bin/lpr -rwx--s--x 1 root lp 17412 Nov 19 23:32 /usr/bin/lprm -rwxr-xr-x 1 root lp 2816 May 10 13:37 /usr/bin/lptest ...and for each of the spool directories listed in the sd fields of /etc/printcap... /var/spool/lpd: total 5 drwxrwxr-x 2 root lp 1024 May 18 23:00 . drwxr-xr-x 11 root root 1024 Feb 19 20:56 .. -rw-rw-r-- 1 root lp 4 May 18 23:00 .seq -rw-rw-r-- 1 root lp 18 May 18 23:00 lock -rw-rw-r-- 1 root lp 25 May 18 23:00 status Note these 3 files are created by lpr and lpd so if you've never run these they will be absent and you should touch them into being. In older versions the group was ``daemon'' not ``lp''. There's also a named socket but lpd creates/deletes it as needed. Some of these permissions may be overkill so don't be surprised if your system works with different permissions. The fact that lpq needs to be suid(root) is inelegant and may change in the future. People tell me that lpr must be suid(root) but I've not seen evidence that this is really the case as long as the file permissions on the spool queues are right. Still as far as I know lpr is designed to be secure when installed setuid(root). Alternative approach: just make lpc, lpr and lprm setuid(root) then you can almost forget the file permissions on the spool queues! You're free to choose different directories for the executables on your system (notably lpc is usually in /etc even though it has commands that are useful to non-root). The master lpd lock file is fixed at compile time to be /var/spool/lpd/lpd.lock so you must have a /var/spool/lpd directory even if you choose not to keep your spool queue there. In the older binaries the lock file was /usr/spool/lpd.lock so this was not an issue. My advise is keep your primary spool queue in /usr/spool/lpd and make /var a symlink to usr or keep it in /var/spool/lpd and make /usr/spool a symlink to ../var/spool. This gives the greatest compatibility with the pathnames that are compiled into the various distributed binaries. The main configuration file is /etc/printcap although the old binaries look for it in /usr/etc/. The old binaries' README says hosts.equiv and hosts.lpd files are looked for in /etc/inet but again the more accepted convention is to locate them in /etc. Since, IMHO, working out in which particular etc directory each file should reside is tedious, I would advise having only one: mv -i /usr/etc/inet/* /etc/inet/* /usr/etc/* /etc/ rmdir /usr/etc/inet /etc/inet /usr/etc ln -s ../etc /usr/etc ln -s . /etc/inet [ But make sure the files in /usr/etc and /etc/inet are not already symlinks into /etc or you'll be in deep water. ] This makes your /etc huge but I think it's safe. All the latest stuff has only /etc paths compiled into it so eventually we can get rid of these links. 0.3.3 What are all these diretories for? People like to put all things that aren't needed during bootstrap in /usr so that / is as small as possible. They therefore move unessential big files (usually the binaries) from /etc/ to /usr/etc/. Unless you know what you are doing (or following the example of someone that does) I'd advise keeping it all in /etc with /usr/etc and /etc/inet being symlinks back to /etc. (In fact these links are probably soon to be redundant as most recent binaires now look in /etc/.) The Linux/PRO approach is to have a real /usr/etc directory containing all the non-essential stuff and make symlinks to each file therein from its namesake in /etc. To spread out the configuration files for the sake of tidiness there is a directory tree called /conf and many files in /etc symlinks into this tree. Also a few people like all their internet stuff in /etc/inet but as recent binaries excpect to find this stuff in /etc they must create links for each file from /etc into /etc/inet. To avoid the problem of your /usr filesystem filling up as a result of stuff in spool queues some poeple perfer to keep their spool queues and such like in a separate filesystem called /var. If you don't have a /var file system then you should make /var a symlink to usr. If you do have a /var filesystem to may want to make /usr/spool a symlink to ../var/spool so that any programs expecting to find /usr/spool will not be dissapointed. 0.3.4 lpd not working If ``ps ax'' does not reveal a lpd then you daemon has died (or was never started) - usually a sign that it couldn't create its lockfile (or you tried to start it before all your filesystems were mounted). If lpr works only for root then you've probably got a permission problem. If you get ``jobs queued, but cannot start daemon'' or ``lpc: connect: No such file or directory'' you are having trouble with the socket connection to lpd. "start" in the context of this error really means "wake". I don't really understand this but the problem was common in early days with the old binaries but developments in the TCP/IP code in the kernel fixed it for most people. Basically is stems from a network porblem. If you're not on a network then try congifuring youself as a single node network and the problem should vanish. 0.3.5 Where do I get a printcap for a xxxxx? This question is essentially meaningless - see next question. 0.3.6 The semantics of /etc/printcap Given the similarity in appearence and name between /etc/termcap and /etc/printcap one could be forgiven for assuming that they contain analogous infomation. This is not the case. Whereas termcap contains informations about terminal *types* - (mostly escape seqences) printcap contains information about *specific* printers (like the directory that holds the spool queue and the device name of the printer). The information about a printer model's escape sequences and so on are held in the various filters which are programs called by lpd to drive the printer. /etc/printcap simply gives the locations of these filters. For details RTFM(printcap). [ Alternatively the net-HOWTO has a summary of some of the more important fields. ] One last point you should always specify ``suppress header'' ``:sh:'' unless you have *text* (not postscript) printer and want banners. On a text printer they are usually a waste of time and paper. On a postscript printer they usually stop your printer working. (See the banners question for further expanation). 0.3.7 The syntax of a /etc/printcap Ideally RTFM(termcap) (yes, I said *termcap*) but since most people don't have TFM(termcap) here are the essentials. Lines starting # are comments (as you might have guessed). For each printer usable from the lpr command on your system there is one logical line in the file. For the sake of readability each logical line may be spread over several physical lines by making the last character on all but the last physical line a backslash. Each logical line has the following format: name1|name2|name3:string_capability=string:\ :numeric_capability#number:boolean_capability: The leading spaces and colon on the second line are for readability only. A printer can have as many names as you like but conventionally the final name is used as a longhand description of the printer. (Still people are free to say `lpr -P "Fred's grotty teletype"' if that's the description you've given.) The list of capabilities can be as long as needed and the order is not significant. Each ``capability'' is denoted by a two character code. (The name ``capability'' comes form the file format's termcap heritage - parameter or attribute would be a more sensible terms.) [ Note from Ross Biro: capabilities with 3 character names don't work properly which is why the serial port stuff in the old binaries failed. ] Capabilities having string value and have a = delimiter between the capability name and the value while those having a numeric value use a # (actually they can use either a # or an =). Boolean ``capablilties'' are true if they appear in the list and false if they do not. Special characters in a string value can be expressed using ` \\'-escape sequences as in C; in addition, `\E' stands for ESC. ` ' is also a kind of escape character; ` ' followed by CHAR stands for the control-equivalent of CHAR. Thus, ` a' stands for the character control-a, just like `\001'. `\' and ` ' themselves can be represented as `\\' and `\' respectively. `\:' for `:' seems to work but the source code contains a waring that it can confuse the parser and `\072' is a better idea. Example: lp|bam|BAM's Epson FX-80:lp=/dev/lp1:sd=/usr/spool/lp1:sh:mx#0:\ :df=/usr/local/lib/magic-filter/lp.df:\ :if=/usr/local/lib/magic-filter/lp.if: Name is lp (this is the printer that lpr uses by default). It's also known as bam or "BAM's Epson FX-80". The printer is on /dev/lp1 (aka AT-bus LPT1:). I don't want a burst page. I don't want a file length limit. Files queued by `lpr -d' are passed through /usr/local/lib/magic-filter/lp.df and those queued by `lpr' through /usr/local/lib/magic-filter/lp.lf. See also the next section. 0.3.8 An /etc/printcap gotcha Two /etc/printcap files can look identical and yet one works and the other doesn't. See if ``lpc stat'' reports a printer called ` :'. The last character on a continued line must be a\. If there are whitespace characters after the\then it doesn't register the next line as a continuation. 0.3.9 The minimum /etc/printcap? This is a silly question but it is frequently asked. The answer is ``lp:sh'' (that's 6 bytes including the required linefeed character on the end). To use this /etc/printcap you must make /dev/lp a symlink to your printer and create your spool queue directory as /usr/spool/lpd. (You might think that if you wanted banner pages you could loose the ``:sh'' but the termcap syntax requires at least one capability per entry). 0.3.10 How do I prevent the `staircase effect'? Un*x terminates each line of a file with a linefeed but not a carriage return so taken literally a Un*x text file printed on an ASCII device will start each line below the end of the previous line. Some printers can be set to treat "linefeed" as "carriage return, linefeed", others can't. If yours can then do simply do that. If the printer cannot be fixed create a shell script filter that reads: #!/bin/sh if [ $1 = -c ]; then cat else sed -e s/$/^M/ fi # the ``echo -ne'' assumes that /bin/sh is realy bash echo -ne "f where M is a carriage return character not a followed by a M. To type M in emacs use the sequence C-q C-m and in vi use C-v C-m. Conventionally this script is called /usr/lib/lpf. The test of $ 1 allows the isertion of carriage returns to be switched off by the `-l' switch on lpr. Alternatively your printer may have an escape sequence that will set the way it handles linefeed characters. A simple filter that uses an `echo -ne' command to send this sequence may be appropriate. #!/bin/sh # Filter for HP printers to treat LF as CRLF # the ``echo -ne'' assumes that /bin/sh is realy bash echo -ne "033\&k2G cat echo -ne "f 0.3.11 Resetting the printer after each printout Either make your filters do it or define the `tr' ``capability'' in /etc/printcap to be your printer's font reset command. For details of the format of this string see the question on the format of printcap. This may not work if a printout crashes in the middle of an escape sequence - putting a lot of @ on the front may help but this probably won't be enough it you were printing raster graphics when the filter died. 0.3.12 Preventing formfeed after each printout If you don't have an `if' specified in /etc/printcap then lpd will automatically put a formfeed at the end of each file. If you're using a filter then it's up to the filter to decide if it wants to put a formfeed. To disable formfeed completely if you don't have an `if' put :ff=: in your /etc/printcap. But please note this suppresses the formfeed that would usually be printed if a filter dies. If you want formfeeds after text printouts but not on printouts printed with `lpr -l' then create the following `if' filter: #!/bin/sh # the ``echo -ne'' assumes that /bin/sh is realy bash cat if [ "$1" != -c ]; then echo -ne "f fi If you want a formfeed after `lpr -l' to be optional you can misuse the `-i' switch to suppress the formfeed with the following trick (after all `lpr -i -l' would usually not be implemented). #!/bin/sh cat # use lpr -i -l to print raw without trailing formfeed if [ "$1" != -c -o "$4" = -i0 ]; then # the ``echo -ne'' assumes that /bin/sh is realy bash echo -ne "f fi 0.3.13 Printing to a serial port The first section applies to both net-2 and pre-net-2 implementations of lpd; then there is a section applicable only to the older implementation; the rest applies to net-2 only. First, there is a kernel patch you need to make. This applies to pre-net-2 lpd as well (actually, the use of this should probably just be commented out of the lpd source code). The change is for at least patchlevel 10, but probably all others as well. in /usr/src/linux/kernel/chr_drv/tty_ioctl.c, around line 390, you will see: case TIOCEXCL: return -EINVAL; /* not implemented */ Change this to return 0, recompile the kernel, and you'll be all set. The following applies only to the "old" lpd, before net-2. You cannot set any serial port settings (including baud rate) in your printcap. Leave them all out and use stty to set up the port (if necessary) in your rc.local, and you should be all set (to find the correct stty settings, start with the default and twiddle until you can correctly cat a file out the port). The rest of this applies only to net-2 lpd although it may be helpful to read. I put more detail into this because it's what I use. 0.3.14 Serial settings in printcap There are two sets of flags which you will need to set, plus the baud rate (Note: the "fc" flag setting seems to override the br # capability, so be sure to set that correctly as well as the br # !). Each of the flags can have bits set and cleared. Clearing is done first, so specify the clear flags (fc # and xc # ) before the set flags (fs and xs). Setting the br # capability is self-explanatory. Example: br # 9600 It is very easy to translate from 'stty' settings to printcap flag settings. If you need to, see the man page for stty now. Use stty to set up the printer port so that you can cat a file to it and have it print correctly. Here's what "stty -a" looks like for my printer port: dina:/usr/users/andy/work/lpd/lpd# stty -a < /dev/ttyS2 speed 9600 baud; rows 0; columns 0; line = 0; intr = ^C; quit = ^\; erase = ^?; kill = ^U; eof = ^D; eol = ; eol2 = ; start = ^Q; stop = ^S; susp = ^Z; rprnt = ^R; werase = ^W; lnext = ^V; min = 1; time = 0; -parenb -parodd cs8 hupcl -cstopb cread -clocal -crtscts -ignbrk -brkint -ignpar -parmrk -inpck -istrip -inlcr -igncr -icrnl ixon -ixoff -iuclc -ixany -imaxbel -opost -olcuc -ocrnl -onlcr -onocr -onlret -ofill -ofdel nl0 cr0 tab0 bs0 vt0 ff0 -isig -icanon -iexten -echo -echoe -echok -echonl -noflsh -xcase -tostop -echoprt -echoctl -echoke The only changes between this and the way the port is initialized at bootup are -clocal, -crtscts, and ixon. Your port may well be different depending on how your printer does flow control. Once you have your stty settings right, so that cat file > /dev/ttyS2 (in my case) sends the file to the printer, look at the file /usr/src/linux/include/linux/termios.h. This contains a lot of # defines and a few structs (You may wish to cat this file to the printer (you do have that working, right?) and use it as scratch paper -- I did!). Go to the section that starts out /* c_cflag bit meaning */ #define CBAUD 0000017 This section lists the meaning of the ``fc # '' and ``fs # '' bits. You will notice that the names there (after the baud rates) match up with one of the lines of stty output. Didn't I say this was going to be easy? Note which of those settings are preceded with a - in your stty output. Sum up all those numbers (they are octal). This represents the bits you want to *clear*, so the result is your fc # capability. Of course, remember that you will be setting bits directly after you clear, so you can just use fc # 0177777 (I do). Now do the same for those settings (listed in this section) which do not have a - before them in your stty output. In my example the important ones are CS8 (0000060), HUPCL (0002000), and CREAD (0000200). Also note the flags for your baud rate -- mine is 0000015. Add those all up, and in my example you get 0002275. This goes in your fs # capability (fs # 02275 works fine in my example). Do the same with set & clear for the next section of the include file, "c_lflag bits". In my case I didn't have to set anything, so I just use ``xc # 0157777'' and ``xs # 0''. Once your printcap is set up, try it out. If things don't work, see the next section. 0.3.15 cat works to the serial port, but not lpd (1) Generally getting lpd up and running is explained elsewhere, but if you are having trouble with serial port settings you can prevent lpd from trying to configure your port by treating you printer as one that does not present a normal device interface (qv). 1. Set your printer (in your printcap) to /dev/null1. Create /dev/null1 using mknod /dev/null1 c 1 3 (because you don't want /dev/null to be opened exclusively). remove the baud rate and flags settings from your printcap. 2. Create a script such as this: #!/bin/sh echo if: $* >> /var/spool/lpd/results # /dev/lp is linked to /dev/ttyS2 which has the printer exec your-old-input-filter $* > /dev/lp ...or if you didn't have an old if installed... #!/bin/sh echo if: $* >> /var/spool/lpd/results cat > /dev/lp # the ``echo -ne'' assumes that /bin/sh is realy bash echo -en "f > /dev/lp Make sure it's world-executable and world-readable. I call my script /var/spool/lpd/if, to make things easy. Try out your script (/var/spool/lpd/if < somefile) and see if it prints. 3. Set the if= capability in your printcap to call this script, e.g. if=/var/spool/lpd/if 4. Use stty to correctly set your port settings. Try to print now. You should be able to tell if things are being spooled, and things *should* be printed, if your manual testing of the if script works. But this is a kludge, so the idea is not to use the if script. 0.3.16 cat works to the serial port, but not lpd (2) Assuming the above method using the "if" filter works and that you belive that you have specified what you think are the correct flags and baud rate in printcap; check ``stty -a < /dev/ttyS2'' (or whatever your printer port is). If the settings are not correct, check your flags against your printout from termios.h. If the settings are *way* not correct, you may need to fix lpd itself. The patch follows, and you can probably see why it's needed :-) It gets applied in the lpd/ subdirectory of the lpd source. (the patch is coming in just a sec) When I was setting mine up, I followed a sequence like this: lprm (whatever -- make sure queue is empty and lpd is running) stty (correct settings) < /dev/ttyS2 lpr (something small) stty -a < /dev/ttyS2 (often had to ctrl-c out of this one) (twiddle with flags) lprm (whatever -- make sure queue is empty again...) Here's the patch (I think its reversed so apply it with -R): -------------------------------Cut Here------------------------------------- *** printjob.c Thu Jul 8 20:56:59 1993 --- printjob.c~ Sat Feb 27 09:07:01 1993 *************** *** 1271,1277 **** } #ifdef LINUX ttybuf.c_cflag &= ~FC; /* not quite right! */ ! ttybuf.c_cflag |= FS; /* not quite right! */ #else ttybuf.sg_flags &= ~FC; ttybuf.sg_flags |= FS; --- 1271,1277 ---- } #ifdef LINUX ttybuf.c_cflag &= ~FC; /* not quite right! */ ! ttybuf.c_cflag |= ~FS; /* not quite right! */ #else ttybuf.sg_flags &= ~FC; ttybuf.sg_flags |= FS; -------------------------------Cut Here------------------------------------- 0.3.17 Printers that aren't simple character devices [ Firstly I'll explain the subject. ] The most common example is a printer that is connected via a network in some strange way. For example consider a printer conneccted to a host running some esoteric OS with which you can only communicate via E-mail. To use this printer without going through lpr you would say something like: | uuencode tmp | mail printer-request@eosteric in place of: > /dev/lp To use such a printer through lpr the ``lp'' capability of the print queue should be directed to a /dev/null type device (e.g. ``mknod /dev/null1 c 1 3'') but not /dev/null itself as lpd opens the device exclusively. Each filter must must explicitly uuencode and mail its output. In more complex cases if you already have an ``if'' or ``of'' filter for a strangely connected printer then other filters can pass their output to/through this filter to avoid duplication of effort. In this case the ``if'' filter should usually be called with the ``-c'' switch to minimise the further manipulations if performs. 0.3.18 Getting burst/banner pages For a simple text printer (in particular not postscript) and a simple text banner simply take :sh: out of the printcap record. If you want to prevent the banner comming out in whatever font was last used on the printer then define the `tr' ``capability'' to be your printer's font reset command. If you want a fancy customised banner (or have a postscript printer) leave :sh: in the printcap and make each of your filters print the banner. All the information to put on the banner is included in the filter's positional parameters. RTFM(printcap) for details. [ If you're using 's magic-filter package then call the code to print the banners from the config script. ] 0.3.19 Printing text on a PS printer You need a filter based on a program that converts ascii to postscript. The most well known of these is ``enscript'' but it's also the hardest to find. Others include ``a2ps'', ``nenscript'', and ``mpage'' (for details of these see sections 0.2 - 0.2 ). 0.3.20 Trunation of graphics files This is usually because you've got a limit set on the maximum size file that can sit in the spool queue. Put `mx # 0' in your printcap. 0.3.21 Why doesn't `lpr -i' work? To get lpr -i to work you need a filter istalled as `if' that implements it. The -i switch is simply passed on by lpd to the filter. There is such a filter usually called /usr/lib/lpf that can be obtained from bsd source archives in the directory bsd-source/lpr/filters. It's also part of Fred's port but he put it in /usr/bin/. If you whant to use this program but still want your filter to do some printer specific initialisation then write script thus: #!/bin/sh # My initialisation stuff goes here exec /usr/lib/lpf $* 0.3.22 Why doesn't `lpr -p' work? Because it's broken. lpd always thinks that the printer is 0 characters wide regardless of what printcap says. Previous versions of the lpd-HOWTO contained a patch but it has now been applied to the latest distribution of the net-2 stuff. One other thing: lpd calls pr by full pathname so if you keep pr somewhere different from /usr/bin/pr you will need a symlink. (Where lpd expects to find pr may vary from version to version). 0.3.23 lprm says `too many users' You have an anchient libc.so with a broken scandir(). 0.3.24 lpc complains that it hasn't stopped a daemon One lpd process runs all the time and it spawns children to handle each printer as needed. Sometimes these children leave lock files lying about. This is harless. 0.3.25 lpr over a network To print on the printer listed as foor "foo" in the printcap on machine "bar.baz.net" from the machine "mine.baz.net" you put an entry like this in your /etc/printcap (on mine.baz.net): foo:lp=:rm=bar.baz.net:rp=foo:sd=/usr/lpd/spool/foo: and, of course, create the spool directory /usr/lpd/spool/foo. There's no point specifying filters and the like in mine.baz.net:/etc/printcap as it's the ones in bar.baz.net:/etc/printcap that will get used. On the machine bar.baz.foo, you need to put "mine.baz.net" on a line by itself in either /etc/hosts.equiv or /etc/hosts.lpd; note that putting it in /etc/hosts.equiv will allow for unauthenticated logins as well as printing. /etc/hosts.lpd is printing only. 0.3.26 Writing lpd filters In normal Un*x terminology, filters are just programs (so they must have execute permission) that read a stream from their standard input and write to their standard output. lpd filters are filters in the sense that thay read STDIN and write to STDOUT, but are unusual in that they may assume that their standard input is a file and perform lseek() operations on it. All lpd filters have a common command line syntax (or more often simply ignore command line parameters). For details of the command line parameters RTFM(printcap). It's easier to debug filters if you test them in an immediate shell before you install them. (If your filter makes use of its command line arguments you'll have to specify them too). my-new-filter /dev/lp1 If you want to write a shell script filter it must have a # !/bin/sh header. Here is the generic form of a filter to accept postscript. #!/bin/sh /path.../gs -q -dSAFER -dNOPAUSE -r??? -sDevice=?????? -sOutputFile=- - and place its full pathname as one of the filter (usually `if' but not 'of'!) parameter in the printcap for your printer. I suggest putting such scripts in a subdirectory off /usr/lib (sat /usr/lib/lpd/). (-dSAFER attempts to protect against ps interpreter security holes, -q and -dNOPAUSE make it churn right along quietly, and Device is the appropriate driver for your printer). Here is an Epson FX-80 dvi filter using ghostscript: #!/bin/sh /usr/TeX/bin/dvips -f | \ /usr/bin/gs -q -dSAFER -sDEVICE=eps9high -r120x216 \ -dNOPAUSE -sOutputFile=- - More tools useful for makeing filters are described elsewhere in this document. 0.3.27 A filter tests OK but fails when lpd uses it You may need a # !/bin/sh header. You may also need to set PATH within the script since the daemon's PATH may not have everything you need. 0.3.28 `of' filters Never use these. (Well strictly speaking there are circumstances but you're unlikey to meet them). 0.3.29 Getting filters for given printers From: B.A.McCauley@bham.ac.uk (Brian McCauley) >From BSD source archives (I think). [ Editors note: I need a better answer here ] . If you already have a program to print say DVI on your printer by some mechanism then making it into a filter is usually a matter of writting trivial shell script - see this HOWTO. If the program you are using insists on reading a names file as input see the next question. Text mode filters are trivial too (see this HOWTO) unless you want lpr to have a choice of fonts in which case they are slightly harder than trivial. You will probably want to insert and `echo -ne' command at the beginning and end of your filter to set up the font etc to your liking. 0.3.30 Filters from programs that won't read STDIN For example dvilj2p insists on a named file as its input (and what's more expects one with a .dvi suffix) so do this: #!/bin/sh ln -s /proc/self/fd/0 /tmp/$$.dvi dvilj2p /tmp/$$ rm /tmp/$$.dvi Note: If it wasn't for the fact that dvilj2p adds a .dvi suffix you wouldn't need the temporary symlink and could just specify /proc/self/fd/0 directly. People who use this trick often usually keep permanant symlinks to /proc/self/fd/* in /dev. If you don't have procfs you'll need to create a temporary file. 0.3.31 Having many filters Historically the lpr command was created to support a finite set of possible file types. You can, in fact, use any of the filters for any reason. If you're never going to use Benson Varian raster files you could use the -v switch for GIF files. You could even use -d for low res and -v for high res. Remember that if you create a filter for a file format that takes a long time to process then your printer may sit idle between print jobs even when there are things in the queue. If you are on a network remember that the filter setups go on the print server. One way to avoid running out of filter options is to define several logical printers in /etc/printcap that all point to the same physical one and put each filter in the ``if'' field of a different printcap entry. This has the advantage that you can set the PRINTER enviroment variable to choose your filter rather than having to specify it on the command line each time. One small problem with this is that you have no control over the order in which files from separate queues are printed. Another way to avoid running out of possible types is to use magic filters. 0.3.32 What are magic filters? Magic filters deduce their input files' types from `magic numbers' (distictive byte patterns at particular offsets). Magic filters are usually perl scripts, shell scripts or C programs that simply identify the file type then call the appropriate non-magic filter. Blatent plug :-) I have a generic magic filter bash script that selects the right filter to use based on the output of the `file' command. With my magic ``dvi'' filter (and 3 associated non-magic filters) I can do things like: lpr -d file1.dvi file2.div.Z file3.ps file4.texinfo.gz (BTW confguring lpr to handle texinfo files is getting a bit silly - see previous question). This is now at: tsx-11.mit.edu:pub/linux/sources/usr.bin/magic-filter-0.3.tar.gz (Although the release number will possibly change in future). Magic filters should never specified as `of' as the output filter only gets called once if a number of files are printed without a gap. IMHO magic filters as `if' can also cause problems as they may prevent you, say, listing a PostScript or nroff file. (Other people disagree as, if you are making your printer available on a network, it can be a pain having people accidently listing ps files.) The following is an example of a magic shell script which should take either postscript or text and deal with it: 0.3.33 Magic filter examples The following is an example of a magic shell script which should take either postscript or text and deal with it: #!/bin/sh # This is based on a script I received from Scott Doty and which was # written by Keith Walker. Keith's script made use of the fact that # lpd passes options to if: # # -w -l -i -n -h # # to print text out well at any size. This one does not. These options # are also handy if you want to do your own snazzy header page, much # like NeWSPrint from Sun does (although running PostScript through # the display server to get it interpreted is a bit much :) # # This is untested, but bits of it came from working programs, so it # should work. # # gs will reset the printer anyway, so the this text setup doesn't matter # setup should include the escape code for \n conversion, if applicable # printf "" read first_line first_two_chars=`expr $first_line : '(..)'` if [ "$first_two_chars" = "%!" ]; then # it's postscript /usr/bin/gs -dSAFER -dNOPAUSE -q -sDEVICE=??????? -sOutputFile=- - else # it's plain text echo -n $first_line cat printf "\014" fi Note that for the paranoid, shell scripts run as someone other that the user are a security hole, but they are illustrative. An example written in C, which may be easily adapted to most installations is available from the printing mail server as lpr _if.c 0.4 Previewing These sections describe various ways to preview things under Linux - that is, how to view them in a way approximating their final form without printing them out. 0.4.1 ghostview Ghostview, a companion program for gs, previews postscript on an X display. It also lets you select individual or ranges of pages from a postscript document to print using lpr. The new version, 1.5, has fixed a few glitches which never bothered me but may make a difference to you. It also calls gs with the -dSAFER option and has a few more resource and command-line options relative to 1.4.1. The real installation is from prep.ai.mit.edu:/pub/gnu/ghostview-XXX.tar.gz and builds out of the box. Ghostview requires gs to work. The new version of gs, 2.6.x, will use X display fonts in an effort to improve legibility at the low resolutions of a video monitor (a previous failing of this pair relative to commercial display-postscript based systems such as Sun's pageview). This works very well for me at least, at the expense of exact character positioning (x fonts have different widths). In fact, I thought that Ghostview looks better than Sun's pageview the other day when I looked at the same page in oth programs side-by-side. Ghostview/gs also has much more intelligent color handling than pageview. You might wish to let gs render some Type 1 fonts you install instead of using platform fonts (or the awful fonts gs comes with. To do this while in ghostview (or in any situation involving the X11 driver), place ``ghostscript.useExternalFonts: false'' in your .Xdefaults file, and the platform fonts will not be used. This is part of a message posted to gnu.ghostscript.bug by Tim Theisen : (note that the usual Linux X-server, XFree, is simply an enhanced version of MIT's effort at a i386 X-server (X386), and does contain the X11R5 Type 1 rasterizer which I beleive was contributed by IBM.) Ghostscript now uses the X Toolkit to pick up X Resources. Now ghostscript uses the standard X rules that allow more specific resources to override less specific ones giving users the full power of X resources to control the X11 driver. It also allows system administrators to establish an application defaults file with resources specific to their ghostscript installation. The customization choices mentioned in make.doc have been moved into X resources and are now configured at run time rather than compile time. Sorry, this section of make.doc did not get revised for the 2.6.1 release. If useBackingPixmap is set, ghostscript will attempt to allocate a backing pixmap. If one cannot be allocated, ghostscript will issue a warning and ask for backing store instead. (Since there were insufficient resources for a backing pixmap, the X server may not provide backing store either.) Color Handling was totally revamped for gs 2.6. Ghostscript first checks for a suitable standard colormap. If you have static colormap in your X server, it would be best to store a standard colormap property on the root window describing the color layout. Ghostscript will then be able to take full advantage of the device. If you have a standard colormap installed, ghostscript will start slightly faster since it does not have to allocate colors for a cube or ramp. If no standard colormap is available, ghostscript will allocate an RGB cube or gray ramp. Ghostscript tries for a 5x5x5 cube on a color device, and a 128 gray ramp on grayscale devices. It will never ask for more than 1/2 of the colors for a RGB cube or gray ramp. It also takes into account the number of significant bits per pixel. (i.e. It won't ask for 128 gray levels if you only have 16 available.) Ghostscript will attempt to allocate colors that are off the color cube/ramp as the picture is being rendered. Ghostscript will keep track of 256 dynamic colors. After all these are allocated, ghostscript asks the X server directly. The foreground and background color can be set explicitly. This is important for the visually impaired and when using the ghostview widget. Color Resources: * palette(Palette): Default value: "Color". Other allowable settings: "Grayscale", "Monochrome". The palette resource is used to restrict the palette used for display. One can set palette to "Grayscale" or "Monochrome" to see how a file would be rendered in grayscale or monochrome on a color display. I use it to avoid dithering of gray- scale figures on a color display with 4-bit DACs. * maxGrayRamp(MaxGrayRamp): Default value: 128. Maximum number of gray levels that ghostscript will attempt to allocate. (It won't try for more than this on an 8-bit pseudo color display even if you set it higher.) Set this lower if you want a smaller ramp and would prefer ghostscript to use dynamic colors. * maxRGBCube(MaxRGBCube): Default value: 5. Maximum number of colors levels that ghostscript will attempt to allocate. (It won't try for more than this on an 8-bit pseudo color display even if you set it higher.) Set this lower if you want a smaller ramp and would prefer ghostscript to use dynamic colors. I believe these values to be a good compromise between dynamic allocation and fall back onto a fairly good color cube for dithering. You can use the foreground and background colors to accomplish "reverse video". However, if you have a grayscale device, it may be better to reverse the gray ramp using the following PostScript code fragment: [ 1 exch sub /exec load currenttransfer /exec load ] cvx settransfer The X11 driver now supports native X11 fonts. If you have installed the HP XLFD font extensions into your font or X server. Ghostscript will also be able to use platform fonts at rotations of 90 degrees, with mirroring, and anamorphic scaling. The X11 driver does most if its work silently. You can get it to report when it is using an X11 font by setting the logExternalFonts boolean in your X resources. The X11 driver is setup to use the standard fonts distributed with X11R5. We purchased the Adobe Type Manager and the Adobe Plus Pack. These font packages give all the fonts normally found in the Apple LaserWriter Plus. The X11 driver is setup to handle these fonts as well. (They are a superset of the bitmap fonts distributed with X11.) You may set the regularFonts, symbolFonts, or dinbatFonts resources if you have different fonts available. Each font name must have 7 dashes or it will be ignored. Minimize the use of wildcards to promote faster matching. (I once encountered an X server that took many seconds to do a font lookup when wildcards were carelessly used.) There is a different list of fonts for each common encoding. Regular fonts may be accessed in standard or ISO Latin 1 encoding. The bdf files that are distributed with X11 are in the ISO Latin 1 encoding. This leaves out the ligatures. Luckily, the ligatures are present in the bdf files, but are not given an encoding, essentially commenting them out. You can use the fixfont program from the xproof distribution (ftp.cs.wisc.edu:/pub/X/xproof.tar.Z, or ftp.x.org:/contrib/xproof.tar.Z) to reencode the bdf files and build X11 fonts that contain the ligatures (i.e standard encoding). If you have the Type1 fonts mentioned above, and you installed the Type1 rasterizer into you font or X server, you can use the appended fonts.scale to name your fonts so that ghostscript can find them. Font resources: * useExternalFonts(UseExternalFonts): Default value: true. This resource controls whether X11 fonts will be used. * useScalableFonts(UseScalableFonts): Default value: true. This resource controls whether scalable fonts will be used. If you have an outline scaler in your X server, you should have this on. If you have an X terminal, you may get slightly better performance with this on. If you have to use the X11 bitmap scaler, turn this off. Fonts scaled by the bitmap scaler look worse than the default ghostscript fonts. * logExternalFonts(LogExternalFonts): Default value: false. Controls whether to report when X11 fonts are being used. The following fonts.scale makes all of the fonts of the Adobe Type Manager and Adobe Plus pack available in standard and ISO Latin 1 encoding. (We were able to purchase the above two packages at an educational discount price of $ 150.) 70 agw_____.pfb -Adobe-ITC Avant Garde Gothic-Book-r-normal--0-0-0-0-p-0-iso8859-1 agwo____.pfb -Adobe-ITC Avant Garde Gothic-Book-o-normal--0-0-0-0-p-0-iso8859-1 agd_____.pfb -Adobe-ITC Avant Garde Gothic-Demi-r-normal--0-0-0-0-p-0-iso8859-1 agdo____.pfb -Adobe-ITC Avant Garde Gothic-Demi-o-normal--0-0-0-0-p-0-iso8859-1 bkl_____.pfb -Adobe-ITC Bookman-Light-r-normal--0-0-0-0-p-0-iso8859-1 bkli____.pfb -Adobe-ITC Bookman-Light-i-normal--0-0-0-0-p-0-iso8859-1 bkd_____.pfb -Adobe-ITC Bookman-Demi-r-normal--0-0-0-0-p-0-iso8859-1 bkdi____.pfb -Adobe-ITC Bookman-Demi-i-normal--0-0-0-0-p-0-iso8859-1 com_____.pfb -Adobe-Courier-Medium-r-normal--0-0-0-0-m-0-iso8859-1 coo_____.pfb -Adobe-Courier-Medium-o-normal--0-0-0-0-m-0-iso8859-1 cob_____.pfb -Adobe-Courier-Bold-r-normal--0-0-0-0-m-0-iso8859-1 cobo____.pfb -Adobe-Courier-Bold-o-normal--0-0-0-0-m-0-iso8859-1 hv______.pfb -Adobe-Helvetica-Medium-r-normal--0-0-0-0-p-0-iso8859-1 hvo_____.pfb -Adobe-Helvetica-Medium-o-normal--0-0-0-0-p-0-iso8859-1 hvb_____.pfb -Adobe-Helvetica-Bold-r-normal--0-0-0-0-p-0-iso8859-1 hvbo____.pfb -Adobe-Helvetica-Bold-o-normal--0-0-0-0-p-0-iso8859-1 hvn_____.pfb -Adobe-Helvetica-Medium-r-Narrow--0-0-0-0-p-0-iso8859-1 hvno____.pfb -Adobe-Helvetica-Medium-o-Narrow--0-0-0-0-p-0-iso8859-1 hvnb____.pfb -Adobe-Helvetica-Bold-r-Narrow--0-0-0-0-p-0-iso8859-1 hvnbo___.pfb -Adobe-Helvetica-Bold-o-Narrow--0-0-0-0-p-0-iso8859-1 ncr_____.pfb -Adobe-New Century Schoolbook-Medium-r-normal--0-0-0-0-p-0-iso8859-1 nci_____.pfb -Adobe-New Century Schoolbook-Medium-i-normal--0-0-0-0-p-0-iso8859-1 ncb_____.pfb -Adobe-New Century Schoolbook-Bold-r-normal--0-0-0-0-p-0-iso8859-1 ncbi____.pfb -Adobe-New Century Schoolbook-Bold-i-normal--0-0-0-0-p-0-iso8859-1 por_____.pfb -Adobe-Palatino-Medium-r-normal--0-0-0-0-p-0-iso8859-1 poi_____.pfb -Adobe-Palatino-Medium-i-normal--0-0-0-0-p-0-iso8859-1 pob_____.pfb -Adobe-Palatino-Bold-r-normal--0-0-0-0-p-0-iso8859-1 pobi____.pfb -Adobe-Palatino-Bold-i-normal--0-0-0-0-p-0-iso8859-1 sy______.pfb -Adobe-Symbol-Medium-r-normal--0-0-0-0-p-0-iso8859-1 tir_____.pfb -Adobe-Times-Medium-r-normal--0-0-0-0-p-0-iso8859-1 tii_____.pfb -Adobe-Times-Medium-i-normal--0-0-0-0-p-0-iso8859-1 tib_____.pfb -Adobe-Times-Bold-r-normal--0-0-0-0-p-0-iso8859-1 tibi____.pfb -Adobe-Times-Bold-i-normal--0-0-0-0-p-0-iso8859-1 zcmi____.pfb -Adobe-ITC Zapf Chancery-Medium-i-normal--0-0-0-0-p-0-iso8859-1 zd______.pfb -Adobe-ITC Zapf Dingbats-Medium-r-normal--0-0-0-0-p-0-iso8859-1 agw_____.pfb -Adobe-ITC Avant Garde Gothic-Book-r-normal--0-0-0-0-p-0-adobe-fontspecific agwo____.pfb -Adobe-ITC Avant Garde Gothic-Book-o-normal--0-0-0-0-p-0-adobe-fontspecific agd_____.pfb -Adobe-ITC Avant Garde Gothic-Demi-r-normal--0-0-0-0-p-0-adobe-fontspecific agdo____.pfb -Adobe-ITC Avant Garde Gothic-Demi-o-normal--0-0-0-0-p-0-adobe-fontspecific bkl_____.pfb -Adobe-ITC Bookman-Light-r-normal--0-0-0-0-p-0-adobe-fontspecific bkli____.pfb -Adobe-ITC Bookman-Light-i-normal--0-0-0-0-p-0-adobe-fontspecific bkd_____.pfb -Adobe-ITC Bookman-Demi-r-normal--0-0-0-0-p-0-adobe-fontspecific bkdi____.pfb -Adobe-ITC Bookman-Demi-i-normal--0-0-0-0-p-0-adobe-fontspecific com_____.pfb -Adobe-Courier-Medium-r-normal--0-0-0-0-m-0-adobe-fontspecific coo_____.pfb -Adobe-Courier-Medium-o-normal--0-0-0-0-m-0-adobe-fontspecific cob_____.pfb -Adobe-Courier-Bold-r-normal--0-0-0-0-m-0-adobe-fontspecific cobo____.pfb -Adobe-Courier-Bold-o-normal--0-0-0-0-m-0-adobe-fontspecific hv______.pfb -Adobe-Helvetica-Medium-r-normal--0-0-0-0-p-0-adobe-fontspecific hvo_____.pfb -Adobe-Helvetica-Medium-o-normal--0-0-0-0-p-0-adobe-fontspecific hvb_____.pfb -Adobe-Helvetica-Bold-r-normal--0-0-0-0-p-0-adobe-fontspecific hvbo____.pfb -Adobe-Helvetica-Bold-o-normal--0-0-0-0-p-0-adobe-fontspecific hvn_____.pfb -Adobe-Helvetica-Medium-r-Narrow--0-0-0-0-p-0-adobe-fontspecific hvno____.pfb -Adobe-Helvetica-Medium-o-Narrow--0-0-0-0-p-0-adobe-fontspecific hvnb____.pfb -Adobe-Helvetica-Bold-r-Narrow--0-0-0-0-p-0-adobe-fontspecific hvnbo___.pfb -Adobe-Helvetica-Bold-o-Narrow--0-0-0-0-p-0-adobe-fontspecific ncr_____.pfb -Adobe-New Century Schoolbook-Medium-r-normal--0-0-0-0-p-0-adobe-fontspecific nci_____.pfb -Adobe-New Century Schoolbook-Medium-i-normal--0-0-0-0-p-0-adobe-fontspecific ncb_____.pfb -Adobe-New Century Schoolbook-Bold-r-normal--0-0-0-0-p-0-adobe-fontspecific ncbi____.pfb -Adobe-New Century Schoolbook-Bold-i-normal--0-0-0-0-p-0-adobe-fontspecific por_____.pfb -Adobe-Palatino-Medium-r-normal--0-0-0-0-p-0-adobe-fontspecific poi_____.pfb -Adobe-Palatino-Medium-i-normal--0-0-0-0-p-0-adobe-fontspecific pob_____.pfb -Adobe-Palatino-Bold-r-normal--0-0-0-0-p-0-adobe-fontspecific pobi____.pfb -Adobe-Palatino-Bold-i-normal--0-0-0-0-p-0-adobe-fontspecific sy______.pfb -Adobe-Symbol-Medium-r-normal--0-0-0-0-p-0-adobe-fontspecific tir_____.pfb -Adobe-Times-Medium-r-normal--0-0-0-0-p-0-adobe-fontspecific tii_____.pfb -Adobe-Times-Medium-i-normal--0-0-0-0-p-0-adobe-fontspecific tib_____.pfb -Adobe-Times-Bold-r-normal--0-0-0-0-p-0-adobe-fontspecific tibi____.pfb -Adobe-Times-Bold-i-normal--0-0-0-0-p-0-adobe-fontspecific zcmi____.pfb -Adobe-ITC Zapf Chancery-Medium-i-normal--0-0-0-0-p-0-adobe-fontspecific zd______.pfb -Adobe-ITC Zapf Dingbats-Medium-r-normal--0-0-0-0-p-0-adobe-fontspecific 0.4.2 gspreview This is another front-end for ghostscript. I have gotten and built it, and actually preferred the user interface, but it had a slight glitch which was probably me doing a careless install rather than a program bug. It didn't seem as feature-full as ghostview, though. (Not that there are all THAT many features in ghostview, but it does its job well). ftp.x.org:/contrib/GSPreview... 0.4.3 xdvi A beautifully legible previewing program for dvi with a handy zoom+pan feature. Will not interpret ps specials, which are understood only by dvips (back to the compiler, object file, and now linker analogy :) To view a file, do ``xdvi file.dvi''. Mine (from ftp.x.org:/contrib/... ) has the selfile feature, where xdvi puts up a file browser... cute but useless. This comes with either TeX or X in the SLS -- I don't remember. Either way, you've probably got one. 0.4.4 xtex Xtex is similar in purpose to xdvi. I have tried to build it under Linux and failed. It is available as ftp.x.org:/contrib/xtex-2.18.5.tar.Z 0.4.5 gxditview Ditview produces a preview version of troff source using X fonts ``groff -TX100 -mandoc man_page.1'' will run gxditview to show you a typeset version of the man page. -TX75 is the same thing, but tiny. My SLS didn't have a working one at all. A good one comes with the source to groff, which you might want to get anyway for the additional drivers (the SLS, until recently, was missing several, including postscript). prep.ai.mit.edu:/pub/gnu/groff-xxxx.tar.z 0.4.6 non-x-previewing Gs comes with pc video hardware drivers, but under un*x these are not a good thing. However, there are gs binaries around which will use the Linux VGA library (vgalib). The ghostscript device for this is called linux, thus ``gs -sDEVICE=linux file.ps'' will show you an image of the Postscript. The environment variable GSVGAMODE is important for this. Set it to the nuber of the video mode you want, taken from the vga.h which comes with vgalib. At one point the SLS binaries had this driver, then they didn't, and now they do again (I saw an announcement to this effect Aug 14). (``gs -help'' will list the available drivers in your binary.) If you need this driver, a patch to put in Linux svgalib is available as ws105.zfn.uni-bremen.de:/pub/gs261-linuxdriver.sh. The plain vgalib driver is available on Sunsite. Texmgr is a program which will preview dvi under MGR. I don't beleive that it currently works under Linux MGR, but if it does, MGR uses sufficiently less memory and disk that this might be an attractive option for some. dvgt is a program which will preview dvi with Linux svgalib, or on one of several types of graphics terminals including vt, tek, or a PC with MS-Kermit. It is available on sunsite. 0.5 Ascii translation These sections describe various programs which can generate plain ascii from some file formats. 0.5.1 from TeX Lametex will generate ascii from tex source. It is available as sunsite.unc.edu:/pub/Linux/apps/tex/lametex.tar.z 0.5.2 from dvi Dvi2tty is the name of a program which will process dvi into text. Aparently, it will also make an effort at reproducing graphics as well. 0.5.3 from ps Ghostscript 2.6.1 comes with a script file which will use gs to extract just the text from a ps file, called ps2ascii. See the gs information above for where it can be found. Further documentation is in the ghostscript 2.6.1 distribution files `gs_2asc.ps' and `use.doc' 0.5.4 from troff groff -Tascii or -Tlatin1..... 0.5.5 from ascii/latin1 The GNU program recode handles conversion between various forms of straight text encoding, ie from Latin-1 to ASCII. This would be available on prep.ai.mit.edu.