debiman

Goals

debiman makes (Debian) manpages accessible in a web browser. Its goals are, in order:

completeness: all manpages in Debian should be available.
visually appealing and convenient: reading manpages should be fun, convenience features (e.g. permalinks, URL redirects, easy navigation) should be available
speed: manpages should be quick to load, new manpages should be quickly ingested, the program should run quickly for pleasant development

Currently, there is one known bug with regards to completeness (#12).

With regards to speed, debiman can process all manpages of Debian unstable in less than 10 minutes on a modern machine. Incremental updates complete in less than 15 seconds. For more details, see PERFORMANCE.md.

Prerequisites

mandoc
a local or remote Debian mirror or an apt-cacher-ng running on localhost:3142
a number of Go packages (which go get will automatically get for you, see below)
- pault.ag/go/debian
- pault.ag/go/archive
- github.com/golang/protobuf/proto
- golang.org/x/crypto/openpgp
- golang.org/x/net/html
- golang.org/x/sync/errgroup
- golang.org/x/text

Architecture overview

debiman works in 4 stages:

All Debian packages of all architectures of the specified suites are discovered. The following optimizations are used to reduce the number of packages, and hence the input size/required bandwidth:
1. packages which do not own any files in /usr/share/man (as per the Contents- archive files) are skipped.
2. each package is downloaded only for 1 of its architectures, as manpages are architecture-independent.
Man pages and auxiliary files (e.g. content fragment files which are included by a number of manpages) are extracted from the identified Debian packages.
All man pages are rendered into an HTML representation using mandoc(1).
An index file for debiman-auxserver (which serves redirects) is written.

Each stage runs concurrently (e.g. Contents and Packages files are inspected concurrently), but only one stage runs at a time, e.g. extraction needs to complete before rendering can start.

Development quick start

Set up Go

Install the latest supported version of Go from https://go.dev/dl. If you prefer to install Go from Debian, ensure you get the same version — if you use Debian stable, you likely need to install from backports.

Install debiman

To download, compile and install debiman to ~/go/bin, run:

go install github.com/Debian/debiman/cmd/...@main

Run debiman

To synchronize Debian testing to ~/man and render a handful of packages, run:

~/go/bin/debiman -serving_dir=~/man -only_render_pkgs=qelectrotech,i3-wm,cron

Test the output

To serve manpages from ~/man on localhost:8089, run:

~/go/bin/debiman-minisrv -serving_dir=~/man

Note that for a production setup, you should not use debiman-minisrv. Instead, refer to the web server example configuration files in example/.

Recompile debiman

To update your debiman installation after making changes to the HTML templates or code in your debiman git working directory, run:

go generate github.com/Debian/debiman/...
go install github.com/Debian/debiman/cmd/...

Synchronizing

For https://manpages.debian.org, we run:

flock /srv/manpages.debian.org/debiman/exclusive.lock \
nice -n 5 \
ionice -n 7 \
debiman \
  -sync_codenames=oldstable,oldstable-backports,stable,stable-backports \
  -sync_suites=testing,unstable,experimental \
  -serving_dir=/srv/manpages.debian.org/www \
  -local_mirror=/srv/mirrors/debian

…resulting in the directories wheezy/, wheezy-backports/, jessie/, jessie-backports/, testing/, unstable/ and experimental/ (respectively).

Note that you will NOT need to change this command line when a new version of Debian is released.

When interrupted, you can just run debiman again with the same options. It will resume where it left off.

If for some reason you notice corruption or other mistakes in some manpages, just delete the directory in which they are placed, then re-run debiman to download and re-process these pages from scratch.

It is safe to run debiman while you are serving from -serving_dir. debiman will swap files atomically using rename(2).

Customization

You can copy the assets/ directory, modify its contents and start debiman with -inject_assets pointed to your directory. Any files whose name does not end in .tmpl are treated as static files and will be placed in -serving_dir (compressed and uncompressed).

There are a few requirements for the templates, so that debiman can re-use rendered manpages (for symlinked manpages):

In assets/manpage.tmpl and assets/manpageerror.tmpl, the string <a class="toclink" is used to find table of content links.
</div>\n</div>\n<div id="footer"> is used to delimit the mandoc output from the rest of the page.

interesting test cases

crontab(5) is present in multiple Debian versions, multiple languages, multiple sections and multiple conflicting packages. Hence, it showcases all debiman features.

w3m(1) has a Japanese translation which is only present in UTF-8 starting with Debian jessie. It also has a German translation starting with Debian stretch.

qelectrotech(1) has a French translation in 3 different encodings (none specified, ISO8859-1, UTF-8).

mysqld(8) is present in two conflicting packages: mariadb-server-core-10.0 and mysql-server-core-5.6.

URLs

The URL schema which debiman uses is (<suite>/)(<binarypkg/>)<name>(.<section>(.<lang>)). Any part aside from name can be omitted; here are a few examples:

Without suite and binary package:

With binary package:

With suite:

With suite and binary package:

Cheap and reliable Node.js hosting starts at $3/month, and $1/month static HTML hosting

Debian / debiman

Programming Languages

Labels

Projects that are alternatives of or similar to debiman