MAN.CGI(8) System Manager's Manual MAN.CGI(8)

man.cgiCGI program to search and display manual pages

The man.cgi CGI program searches for manual pages on a WWW server and displays them to HTTP clients, providing functionality equivalent to the man(1) and apropos(1) utilities. It can use multiple manual trees in parallel.

At the top of each generated HTML page, man.cgi displays a search form containing these elements:

  1. An input box for search queries, expecting either a name of a manual page or an expression using the syntax described in the apropos(1) manual; filling this in is required for each search.

    The expression is broken into words at whitespace. Whitespace characters and backslashes can be escaped by prepending a backslash. The effect of prepending a backslash to another character is undefined; in the current implementation, it has no effect.

  2. A man(1) submit button. The string in the input box is interpreted as the name of a manual page.
  3. An apropos(1) submit button. The string in the input box is interpreted as a search expression.
  4. A dropdown menu to optionally select a manual section. If one is provided, it has the same effect as the man(1) and apropos(1) -s option. Otherwise, pages from all sections are shown.
  5. A dropdown menu to optionally select an architecture. If one is provided, it has the same effect as the man(1) and apropos(1) -S option. By default, pages for all architectures are shown.
  6. A dropdown menu to select a manual tree. If the configuration file /var/www/man/manpath.conf contains only one manpath, the dropdown menu is not shown. By default, the first manpath given in the file is used.

The man.cgi program generates five kinds of output pages:

The index page.
This is returned when calling man.cgi without PATH_INFO and without a QUERY_STRING. It serves as a starting point for using the program and shows the search form only.
A list page.
Lists are returned when searches match more than one manual page. The first column shows the names and section numbers of manuals as clickable links. The second column shows the one-line descriptions of the manuals. For man(1) style searches, the content of the first manual page follows the list.
A manual page.
This output format is used when a search matches exactly one manual page, or when a link on a list page or an Xr link on another manual page is followed.
A no-result page.
This is shown when a search request returns no results - either because it violates the query syntax, or because the search does not match any manual pages.
An error page.
This cannot happen by merely clicking the “Search” button, but only by manually entering an invalid URI. It does not show the search form, but only an error message and a link back to the index page.

For each manual tree, create one first-level subdirectory below /var/www/man. The name of one of these directories is called a “manpath” in the context of man.cgi. Create a single ASCII text file /var/www/man/manpath.conf containing the names of these directories, one per line. The directory given first is used as the default manpath.

Inside each of these directories, use the same directory and file structure as found below /usr/share/man, that is, second-level subdirectories /var/www/man/*/man1, /var/www/man/*/man2 etc. containing source mdoc(7) and man(7) manuals with file name extensions matching the section numbers, second-level subdirectories /var/www/man/*/cat1, /var/www/man/*/cat2 etc. containing preformatted manuals with the file name extension ‘0’, and optional third-level subdirectories for architectures. Use makewhatis(8) to create a mandoc.db(5) database inside each manpath.

Configure your web server to execute CGI programs located in /cgi-bin. When using OpenBSD httpd(8), the slowcgi(8) proxy daemon is needed to translate FastCGI requests to plain old CGI.

To compile man.cgi, first copy cgi.h.example to cgi.h and edit it according to your needs. It contains the following compile-time definitions:

Only useful for running on www.openbsd.org to deal with old URIs containing "manpath=OpenBSD " where the blank character has to be translated to a hyphen. When compiling for other sites, this definition can be deleted.
An optional file system path to the directory containing the file mandoc.css, to be specified relative to the server's document root, and to be specified without a trailing slash. When empty, the CSS file is assumed to be in the document root. Otherwise, a leading slash is needed. This is used in generated HTML code.
An ASCII string to be used for the HTML <TITLE> element.
A file system path to the man.cgi data directory relative to the web server chroot(2) directory, to be specified with a leading slash and without a trailing slash. It needs to have at least one component; the root directory cannot be used for this purpose. The files manpath.conf, header.html, and footer.html are looked up in this directory. It is also prepended to the manpath when opening mandoc.db(5) and manual page files.
The initial component of URIs, to be specified without leading and trailing slashes. It can be empty.

After editing cgi.h, run

make man.cgi

and copy the resulting binary to the proper location, for example using the command:

make installcgi

In addition to that, make sure the default manpath contains the files man1/apropos.1 and man8/man.cgi.8, or the documentation links at the bottom of the index page will not work.

man.cgi uniform resource identifiers are not needed for interactive use, but can be useful for deep linking. They consist of:

  1. The http:// or https:// protocol specifier.
  2. The host name.
  3. The SCRIPT_NAME, preceded by a slash unless empty.
  4. To show a single page, a slash, the manpath, another slash, and the name of the requested file, for example /OpenBSD-current/man1/mandoc.1. This can be abbreviated according to the following syntax: [/manpath][/mansec][/arch]/name[.sec]
  5. For searches, a query string starting with a question mark and consisting of key=value pairs, separated by ampersands, for example ?manpath=OpenBSD-current&query=mandoc. Supported keys are manpath, query, sec, arch, corresponding to apropos(1) -M, expression, -s, -S, respectively, and apropos, which is a boolean parameter to select or deselect the apropos(1) query mode. For backward compatibility with the traditional man.cgi, sektion is supported as an alias for sec.

For security reasons, in particular to prevent cross site scripting attacks, some strings used by man.cgi can only contain the following characters:

  • lower case and upper case ASCII letters
  • the ten decimal digits
  • the dash (‘-’)
  • the dot (‘.’)
  • the slash (‘/’)
  • the underscore (‘_’)

In particular, this applies to all manpaths and architecture names.

The web server may pass the following CGI variables to man.cgi:

The initial part of the URI passed from the client to the server, starting after the server's host name and ending before PATH_INFO. This is ignored by man.cgi. When constructing URIs for links and redirections, the SCRIPT_NAME preprocessor constant is used instead.
The final part of the URI path passed from the client to the server, starting after the SCRIPT_NAME and ending before the QUERY_STRING. It is used by the show page to acquire the manpath and filename it needs.
The HTTP query string passed from the client to the server. It is the final part of the URI, after the question mark. It is used by the search page to acquire the named parameters it needs.

/var/www
Default web server chroot(2) directory. All the following paths are specified relative to this directory.
/cgi-bin/man.cgi
The usual file system path to the man.cgi program inside the web server chroot(2) directory. A different name can be chosen, but in any case, it needs to be configured in httpd.conf(5).
/htdocs
The file system path to the server document root directory relative to the server chroot(2) directory. This is part of the web server configuration and not specific to man.cgi.
/htdocs/mandoc.css
A style sheet for mandoc(1) HTML styling, referenced from each generated HTML page.
/man
Default man.cgi data directory containing all the manual trees. Can be overridden by MAN_DIR.
/man/manpath.conf
The list of available manpaths, one per line. If any of the lines in this file contains a slash (‘/’) or any character not contained in the Restricted character set, man.cgi reports an internal server error and exits without doing anything.
/man/header.html
An optional file containing static HTML code to be inserted right after opening the <BODY> element.
/man/footer.html
An optional file containing static HTML code to be inserted right before closing the <BODY> element.
/man/OpenBSD-current/man1/mandoc.1
An example mdoc(7) source file located below the “OpenBSD-current” manpath.

The man.cgi CGI program is call-compatible with queries from the traditional man.cgi script by Wolfram Schneider. However, the output looks quite different.

apropos(1), mandoc.db(5), makewhatis(8), slowcgi(8)

A version of man.cgi based on mandoc(1) first appeared in mdocml-1.12.1 (March 2012). The current mandoc.db(5) database format first appeared in OpenBSD 6.1.

The man.cgi program was written by Kristaps Dzonsons <kristaps@bsd.lv> and is maintained by Ingo Schwarze <schwarze@openbsd.org>, who also designed and implemented the database format.

May 20, 2018 OpenBSD 6.7