Foomatic 4.0.17 =============== foomatic-filters ---------------- Filter scripts used by the printer spoolers to convert the incoming PostScript data into the printer's native format using a printer/driver specific, but spooler-independent PPD file. http://www.openprinting.org/ This README contains mainly info for developers. See the file USAGE if you want to know how to use Foomatic. Copying ------- This package is free software and can be redistributed and/or modified under the terms of the GNU General Public License; either version 2 of the License, or (at your option) any later version. Authors ------- Till Kamppeter Lars Uebernickel Grant Taylor Helge Blischke Bugs ---- If you spot a data error or any other bug, please report it on the OpenPrinting bug tracking system: http://bugs.linux-foundation.org/ Choose "OpenPrinting" as the product and "foomatic-filters" as the component. Intro ----- For getting Foomatic PPD files for this version, go to http://www.openprinting.org/ See the README file of "foomatic-db-engine" for a (more or less) complete overview of Foomatic. User discussion happens on: http://forums.openprinting.org/list.php?34 Developer mailing list: https://lists.linux-foundation.org/mailman/listinfo/printing-foomatic Supported printing systems/spoolers ----------------------------------- CUPS - Common Unix Printing System (http://www.cups.org/) SOLARIS - Solaris LP (lpsched) LPD - Line Printer Daemon (Does this have a home page anywhere?) LPRng - LPR - New Generation (https://sourceforge.net/projects/lprng/) GNUlpr - An enhanced LPD (http://sf.net/projects/lpr, development stopped) PPR - Page PRinter spooler (http://ppr.sourceforge.net/, development stopped) PDQ - Print, Don't Queue (http://pdq.sf.net/, development stopped) CPS - Coherent Printing System (http://www.tww.cx/cps.php, development stopped) --- - Direct, spooler-less printing (http://www.openprinting.org/) As most of the supported printing systems are not actively developed any more, support for them will be removed in a later version of foomatic-rip. This can even happen before 5.x. For now their support is deprecated and there were only few tests done during the development of foomatic-rip 4.x. So be prepared that there are bugs (patches welcome). Testing were mostly done with the de-facto standard CUPS and spooler-less. What is in this package? ------------------------ This package contains only two programs, its man pages, and a test suite. The programs arw foomatic-rip and beh (Backend Error Handler). foomatic-rip is the main PostScript/PDF-to-printer's-native-language filter and beh is a CUPS backend wrapper for fine-tuning how a CUPS queue should behave on failure of its backend. foomatic-rip works with all spoolers and always uses PPD files for printer/driver capabilities info. Manufacturer-supplied PPDs (in most cases of of PostScript printers) can be used, too. foomatic-rip is now written in C, in contrary to Perl in Foomatic 3.x. This allows foomatic-rip to link directly with libraries. foomatic-rip currently uses only the standard libraries. The functionality of the former foomatic-gswrapper is now integrated into foomatic-rip, so the foomatic-gswrapper is not needed any more. beh is still written in Perl and does not need anything else than a Perl interpreter. It is only for use with CUPS. The test/ subdirectory contains a test suite, originally developed for the distribution compliance tests of the LSB. It also serves as a regression test for further development of foomatic-rip. To execute the test, go into that directory and run "./testfoomaticrip" as root (as normal user the test series #13, CUPS queues, will not work) and you test the foomatic-rip installed on your system. You can modify the testing configuration by the variables in the beginning of the script. The test suite will not get installed with the "make install" command. The main directory contains the source code of foomatic-rip, beh, and the infrastructure for building and installing the package. It also contains the documentation. Dependencies ------------ To build this package you need a C compiler, and its standard libraries. To run beh a Perl interpreter (5.6.0 and newer) is needed. To connect to remote printers in non-CUPS printing environments, you need additional connectivity software (as "rlpr", "nc", "smbspool', ...). How does it work? ----------------- foomatic-rip is a filter which takes PostScript or PDF (and for non-CUPS operation also certain other formats which will get converted to PostScript) from standard input and translates it to the printer's native language. The resulting data is usually directed to standard output, but depending on the spooler abnd its configuration it can also be directed to elsewhere. The information how to do this translation it gets from a PPD file, from command line options, environment variables, and spooler configuration files. foomatic-rip does the following steps to do its work: Spooler auto-detection At first, foomatic-rip reads its command line and a certain assortment of environment variables. With this information it determines from which spooler it was called, since every spooler calls its filter(s) with different command lines and different information supplied via environment variables. Gathering all information to execute the print job Next step after figuring out what the spooler is, is collecting the information about the print job which was not found in the first step. Now the knowledge of which spooler is used is taken into account for interpreting the information. Reading the PPD file In one of the previous steps we have found the name of the PPD file assigned to the print queue currently in use. Now the PPD file is read to get all information needed to build the renderer's (Usually, the renderer is Ghostscript, when no renderer is needed, as for a PostScript printer, "cat" is used) command line, the available options, their default values, and how to apply them. After having parsed the PPD file we have a renderer command line and a list of options with the range of possible settings and a default setting. For LPRng, LPD, GNUlpr, and spooler-less printing we get also the so-called postpipe here, defining a shell command line into which foomatic-rip should direct its output. If no postpipe is found, the output data goes to standard output. The postpipe allows to print to destinations which are not directly supported by the spooler. Applying user-supplied settings All option settings which the user has supplied on the command line are checked whether they are valid (option exists, choice in range) and then applied to the list of default settings, replacing the defaults by the values given by the user. The options not mentioned on the command line keep their default values from the PPD file. Check for the "docs" option foomatic-rip accepts a special option which is not defined in the PPD file, the "docs" option. When the user supplies it, he wants to print a listing of all options available for the printer/driver combo in use. So the incoming data on standard input is discarded and a sub-process for generating the option listing in plain text form is launched. Standard input of the main process is connected to the output of the sub-process. Now the main process behaves as the option listing would be the job which the user has sent. Print files With some spoolers the jobs to be printed are supplied in files. In this case we close standard input and open the file on the standard input handler. This way the following steps read from the file instead of from standard input. The rest of the foomatic-rip process is repeated for every input file, to print them one after the other. Raw queue When we have a raw queue, all the rest of the incoming data is directly passed to standard output or to the postpipe now. The following steps will be omitted then. Print the job Jobs are usually expected to be sent in PostScript or PDF format (handling of other formats in non-CUPS environment will be shown later). PostScript is the standard format for print jobs under Unix and Linux for many years. Currently, we are on the way to move to PDF as standard format (http://www.linuxfoundation.org/en/OpenPrinting/PDF_as_Standard_Print_Job_Format). Therefore we accept both formats. In most cases Ghostscript is the renderer which understands both formats equally. So we do not convert PDF to PostScript if either the renderer command line starts with "gs ..." as then Ghostscript is the renderer and no pre-processing of incoming PostScript happens or we have a dedicated commad line prototype for PDF defined in the PPD file ("*FoomaticRIPCommandLinePDF:" keyword). In addition there must be no option at all which is implemented by inserting active PostScript code (not only comments starting with "%") into a PostScript input data stream. If these criteria are not fulfilled, the PDF job will be converted to PostScript. After all the preparation, and if the job is PostScript the data stream is examined for traces of option settings supposed to be applied to the renderer's command line or to the JCL (Job Command Language, for example PJL) header which is sent to the printer before the renderer's output is sent. PPD-aware applications and spoolers stuff option settings directly into the file, they do not necessarily send PPD options by the command line. There is also stuffed in PostScript code to apply option settings given by the command line of the printing command ("lpr", "lp", ...) and to set the defaults given in the PPD file. Examination strategy: We read lines from standard input until the first %%Page: comment appears and save them as @psheader. This is the page-independent header part of the PostScript file. The PostScript interpreter (renderer) must execute this part once before rendering any assortment of pages. Then pages can be printed in any arbitrary selection or order. All option settings we find here will be collected in the default option set for the RIP (Raster Image Processor, renderer) command line. Now the pages will be read and sent to the renderer, one after the other. Every page is read into memory until the %%EndPageSetup comment appears (or a certain amount of lines was read in the case that there is no %%EndPageSetup). So we can get option settings only valid for this page. If we have such settings we set them in the modified command set for this page. If the renderer is not running yet (first page) we start it with the command line built from the current modified command set and send the first page to it, in the end we leave the renderer running and keep input and output pipes open, so that it can accept further pages. If the renderer is still running from the previous page and the current modified command set is the same as the one for the previous page, we send the page. If the command set is different, we close the renderer, re-start it with the command line built from the new modified command set, send the header again, and then the page. After the last page the trailer (%%Trailer) is sent. The output pipe of this program stays open all the time so that the spooler does not assume that the job has finished when the renderer is re-started. Non DSC-conforming documents will be read until a certain line number is reached. Options for the renderer's command line or the JCL header appearing later will be ignored. This means that option settings in the page headers will not be taken into account. If options are implemented by PostScript code supposed to be stuffed into the job's PostScript data we stuff the code for all these options into our job data, So all default settings made in the PPD file (the user can have edited the PPD file to change them) are taken care of and command line options get also applied. To give priority to settings made by applications we insert the options's code in the beginnings of their respective sections, so that sommething, which is already inserted, gets executed after our code. Missing sections are automatically created. In non-DSC-conforming files we insert the option code in the beginning of the file. This is the same policy as used by the "pstops" filter of CUPS. If CUPS is the spooler, the option settings were already inserted by the "pstops" filter (both PPD defaults and user-supplied options), so we don't insert them again. The only thing we do is correcting settings of numerical options when they were set to a value not available as choice in the PPD file, As "pstops" does not support "real" numerical options, it sees these settings as an invalid choice and stays with the default setting. In this case we correct the setting in the first occurence of the option's code, as this one is the one added by CUPS, later occurences come from applications and should not be touched. If the input data is PDF then we do not search for options in the data stream nor do we try to insert settings, as this is not forseen by PDF. As told above, data in PDF format is only passed on as PDF if there are no options in the PPD file which are implemented by PostScript code to be embedded in the data stream. All options are command line or JCL options. If there are page-dependent option changes we always restart the renderer with the changed JCL and command line options. We tell the renderer by command line options which pages he has to print. If we are in a non-CUPS environment and the input is neither PostScript nor PDF (if there is no "%!" after $maxlinestopsstart lines) a file conversion filter (input format -> PostScript) will automatically be applied to the incoming data, so that we will process the resulting PostScript here. This way we have always PostScript data here and so we can apply the printer/driver features described in the PPD file. For the file conversion filter two subprocesses are started, the task of the first one is to pass the already buffered lines into the filter and then to continue reading standard input (without parsing the data) to pass the rest of the job to the filter. The second subprocess is the filter itself, getting its standard input from the first subprocess and the giving its standard output to the main process. This way the main process has again PostScript as its standard input. Supported file conversion filters are "a2ps", "enscript", "mpage", and spooler-specific filters. All filters convert plain text to PostScript, "a2ps" also other formats. The conversion filter is always used when one prints the documentation pages, as they are created as plain text, when CUPS is the spooler "pstops" is executed after the filter so that the default option settings from the PPD file and CUPS-specific options as N-up get applied. On regular printouts one gets always PostScript or PDF when CUPS is the spooler and PostScript in the case of PPR, so the filter is only used for regular printouts under LPD, LPRng, GNUlpr, PDQ, or without spooler. The main process keeps always parsing the PostScript input or feeding through the PDF input, it launches the renderer in one subprocess and launches an additional subprocess for bracketing the renderer's output with the JCL commands and putting the resulting data to standard output or to the postpipe. Overview of the subprocesses ---------------------------- To do the filtering without loading the whole file into memory we work on a data stream, we read the data line by line, analyse it to decide what filters to use and start the filters if we have found out which we need. We buffer the data only as long as we didn't determine which filters to use for this piece of data and with which options. foomatic-rip splits into up to 6 parallel processes to do the whole filtering (listed in the order of the data flow): KID0: Generate documentation pages (only jobs with "docs" option) KID2: Put together already read data and current input stream for feeding into the file conversion filter (only non-PostScript and "docs" jobs) KID1: Run the file conversion filter to convert non-PostScript input into PostScript (only non-PostScript and "docs" jobs) MAIN: Prepare the job auto-detecting the spooler, reading the PPD, extracting the options from the command line, and parsing the job data itself. It analyses the job data to check whether it is PostScript and starts KID1/KID2 if not, it also stuffs PostScript code from option settings into the PostScript data stream. It starts the renderer (KID3/KID4) as soon as it knows its command line and restarts it when page-specific option settings need another command line or different JCL commands. KID3: The rendering process. In most cases Ghostscript, "cat" for native PostScript printers with their manufacturer's PPD files. KID4: Put together the JCL commands and the renderer's output and send all that either to STDOUT or pipe it into the command line defined with $postpipe.