Should we include rst2man?

[Martin Geisler]

Hi guys,

Our new reST documentation system requires an unreleased version of
Docutils to generate the manpages.

I have talked off-list with Mads Kiilerich about this, and he made me
aware that it is not just a question of asking people to install
Docutils (which is trivial, it's a pure Python package) but also a
question of how Mercurial packages can declare their build-dependencies.

The Debian maintainer has it easy: there is already a suitable rst2man
package in Debian. But not so for Fedora and I guess other distributions
also lack a rst2man package.

Being unable to build a package using only other packages from the same
distribution is a bad situation. Some users might also not be
comfortable with installing a snapshot version of Docutils, especially
if they already have (and use) a stable Docutils on their system.

So I suggest we make life easier for both package maintainers and users
in general and include the rst2man tool in Mercurial itself, similarly
to how we include byterange.py from urlgrabber.

The script is ~1100 lines and could live as /doc/rst2man. The only
requirement for a Mercurial install will then be a "normal" Docutils
installation.

What do you think of this idea?

We can remove the rst2man tool again when Docutils 0.6 has been released
and packaged and when important bugs are fixed in the upstream rst2man.

-- 
Martin Geisler

VIFF (Virtual Ideal Functionality Framework) brings easy and efficient
SMPC (Secure Multiparty Computation) to Python. See: http://viff.dk/.