steve losh
 

Posted on June 1, 2009.

After my last post on What I Hate About Mercurial, seydar commented that I should contribute a patch to add untrack/forget functionality. I decided to take the advice, stop being lazy and write one.

It’s a very, very simple change and I’m not sure if it will even make it into the core repository, but even so it gave me a chance to figure out how to set up a decent development environment for hacking on Mercurial. I figured that other people might find that information useful so I decided to write this.

About This Guide

This is meant to be a guide to get yourself set up to work on Mercurial’s code base and contribute your work back to the community. It’s not a guide for actually working on the code.

I’m going to go ahead and assume a few (small) things:

  1. You’re running OS X or some form of Linux. If you’re using Windows, I’m sorry, but I just don’t use it enough to really help.
  2. You’ve got Python installed.
  3. You’ve got a copy of Mercurial installed in some way (MacPorts, apt-get, source, etc).
  4. You’re familiar with (and have clients for) IRC and email.

If those things are true, let’s get started. The entire process should take between 15 and 45 minutes, depending on how comfortable you are with bash, Python, and Mercurial itself.

Join the Community

The first step to contributing is to join the community. Things will probably go more smoothly if other developers have seen your name before and those other developers are a great resource if you get stuck.

Join the #mercurial channel on Freenode. A lot of other Mercurial developers hang out in there and it’s a good place to go if you want some really quick feedback. Take a few minutes and register your nick on Freenode, just because it’s a good thing to do.

You should also subscribe to the Mercurial development mailing list. That’s where you’re going to be sending your work, so it’s good to start getting an idea of how people use the list as soon as possible.

Grab the Source

Now you’ll want to grab the Mercurial source code so you can start changing it. The repository you almost certainly want to clone is the crew repository.

Several people (listed here) have write access to this and it’s where most changes go before they’re pulled into the main Mercurial repositories. Your want your changes to apply cleanly to the tip of crew if you want them to have the best chance of being accepted.

Go ahead and clone it somewhere on your machine (replace hg-crew if you prefer a different name for the folder):

hg clone http://hg.intevation.org/mercurial/crew hg-crew

It’s going to take a little while. Mercurial might be fast but its repository has over 8,600 changesets. While you’re waiting you can move on to the next step.

Set Up virtualenv

This might not seem like it’s necessary, but trust me, it’s a good thing to do. It will make things easier down the road, I promise.

No, really, take ten minutes and do this. You’ll thank me.

virtualenv is a tool that lets you create separate Python development environments that are isolated from each other. The packages and binaries for one won’t contaminate the others. It will make testing our work on Mercurial painless.

Open a new terminal window while the crew repository is busy cloning. Install virtualenv in the usual Python fashion:

sudo easy_install virtualenv

Now install the virtualenvwrapper tool:

sudo easy_install virtualenvwrapper

This is a nifty little shell script that makes working with virtual environments much more pleasant. Before you can use it, you’ll need to source it into your ~/.bashrc or ~/.bash_profile file, as well as choose a place to stick the environments themselves:

export WORKON_HOME=$HOME/lib/virtualenvs
source /(path to python)/bin/virtualenvwrapper_bashrc

As you can see, I prefer to put the virtualenvs folder inside my ~/lib folder. You can change that to any place you like.

You’ll also need to replace the (path to python) with the actual path to wherever Python is installed on your system. If you’re not sure, here’s a simple, slow, and horribly inefficient but effective command to figure it out:

find / -name 'virtualenvwrapper_bashrc' 2> /dev/null

Once you’ve got the necessary lines in your ~/.bashrc you can close that terminal window.

Set Up A Public Repository

This isn’t strictly required, but it’s polite and only takes a minute or two. If you’ve already got a method of sharing your repositories (using hgwebdir with your website, for example) go ahead and use that.

If not, head over to BitBucket, sign up for an account if you don’t already have one (it’s free), and create a new public repository there. I’ve named mine hg-crew-sjl to be explicit about what it contains, but that’s not terribly important.

Why bother doing this when you’re just going to email your work anyway? Some people (myself included) find it easier to pull changes from another repository than to apply them from emails. It also gets your changes out into a public place so other developers can view and modify your modifications.

It’s also nice to have an easily accessible clone of your personal repository if you’re not at your own machine and want to do some more work.

Set Up Your Local Clone

By now the crew repository should be finished cloning to your local machine. The first thing we want to do is set up the paths so we can push and pull easily. Modify the .hg/hgrc file inside that repository (not your ~/.hgrc file!) to look like this:

[paths]
default = http://bitbucket.org/(username)/(repo name)/
crew = http://hg.intevation.org/mercurial/crew

Replace (username) with your BitBucket username and (repo name) with the name of the BitBucket repository, obviously.

Now you can run hg push to push any changes you commit locally to your public BitBucket repository, and hg pull crew to grab new changes from the crew repository.

Build Mercurial

Make sure you’re in your local crew repository, and run:

make local

This will build Mercurial from the source files. It should finish fairly quickly. Once that’s done, run the tests to make sure everything is working correctly:

make tests

Those will take a long time to run. While you wait, move on to the next step.

Create a Virtual Environment for Mercurial Development

While the test suite is running you’re going to make a new virtual environment to use for testing your changes to Mercurial. First you need to create it, so open up a new terminal window and run:

mkvirtualenv hg-dev

You can name it something else if you don’t like hg-dev. Notice how your shell prompt now has (hg-dev) prepended to it? That’s there to remind you that you’re working in that environment.

Creating the environment automatically puts you into it for that shell session. To get into it in the future you’ll use:

workon hg-dev

Now you need to link the Mercurial libraries you’re going to change to this environment, so you can use it to test your code. Run the following three commands:

ln -s (full path to crew)/mercurial $WORKON_HOME/hg-dev/lib/python2.6/site-packages/
ln -s (full path to crew)/hgext $WORKON_HOME/hg-dev/lib/python2.6/site-packages/
ln -s (full path to crew)/hg $WORKON_HOME/hg-dev/bin/

Replace (full path to crew) with the full path to your local crew repository, and hg-dev with the name of your virtual environment (if you named it differently than I did).

Why Bother With a Virtual Environment?

It’s probably the most work in this entire process, so why did I insist you set it up? Watch this:

sjl at ecgtheow in ~/src/hg-crew-sjl on default 
$ hg --version
Mercurial Distributed SCM (version 1.2.1)

Copyright (C) 2005-2009 Matt Mackall <mpm@selenic.com> and others
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

sjl at ecgtheow in ~/src/hg-crew-sjl on default 
$ workon hg-dev

(hg-dev)
sjl at ecgtheow in ~/src/hg-crew-sjl on default 
$ hg --version
Mercurial Distributed SCM (version 7956a823daa3)

Copyright (C) 2005-2009 Matt Mackall <mpm@selenic.com> and others
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

(hg-dev)
sjl at ecgtheow in ~/src/hg-crew-sjl on default 
$

Look at the version numbers. Notice how they changed?

You can now use workon hg-dev to make the hg command run Mercurial with your changes!

This makes it simple to test your work as you go. I keep two terminal windows open while working on the Mercurial source:

  • One is in my local crew repository to diff, commit, etc with the standard version of Mercurial.
  • The other is in a test repository. I’ve run workon hg-dev to switch to my modified version so I can try out my changes.

Get To Work!

That’s it for the initial setup. Now you can start adding all the wonderful features you have in mind. A simple workflow might go something like this:

  1. Open two terminal windows and use workon hg-dev in the second one to switch to the virtual environment.
  2. Make some changes to the Mercurial codebase.
  3. Test your changes in the second terminal window.
  4. Repeat 3 and 4, using the first terminal window to commit to the local repository as you go.

Once you’re finished with your shiny new feature, how do you get it into the main Mercurial code so we can all benefit from it?

Modify the Test Suite

Before you go off and submit a patch, you need to make sure it works with the current tests. You can run the tests by going to the repository and using:

cd tests
python run-tests.py

It’s going to take a while, just like before. If you changed something that modifies the output of a command it will probably break some of the existing tests. That’s alright, but you’ll need to fix them to reflect your changes. The simplest way to do this is to use the --interactive flag:

python run-tests.py --interactive

Using --interactive will prompt you each time a test fails and ask if you want to accept the new output as the “correct” output. This makes it simple to update the existing tests to take your changes into account.

Obviously you should only accept these updates if they make sense. Pay attention to the output and make sure you didn’t accidentally break something!

Add to the Test Suite

If you’ve implemented an entirely new feature (as opposed to just modifying an existing one) you should add some tests of your own. This will make sure no one else will inadvertently break whatever you’ve added.

There’s a nice guide to this on the Mercurial site. Follow it to add some tests for your changes.

From what I’ve heard it’s preferable to append to an existing set of tests (if it makes sense) instead of adding a brand new test file — the tests run pretty slowly and adding new files make them even slower.

Email Your Patches to the Mailing List

You’ve finally got your code working and updated the tests. It’s time to share your changes with the rest of the Mercurial developers.

If you want your patches to get accepted, you should take the time to read the guide to contributing changes and the guide to making successful patches on the Mercurial site. They’re short but contain most of what you need to know.

I prefer to use the patchbomb extension to email the patches because it takes care of the formatting for me. To enable and configure it you’ll need to make some additions to your ~/.hgrc file:

[extensions]
hgext.patchbomb =

[email]
method = smtp from = Your Name <you@yourdomain.com>

[smtp]
host = smtp.youremailhost.com username = yourusername tls = True

You’ll need to tweak those settings with your own email address, mail server, and so on. Once that’s done, you can have patchbomb package up your changes and email them out:

hg email --rev (your first revision):(your last revision)

The command will guide you through filling out some more information, and then send out the patches. hg help email has a lot more information on how to use it.

I usually email them to myself first just in case I mess up a revision number and email six thousand changesets. Flooding my own inbox is much better than flooding the list’s.

That’s It, Now Go Contribute!

That’s about the extent of my knowledge on contributing to Mercurial. As I said before, I’ve only done one patch so far, but I wanted to write this while everything about setting up the environment was fresh in my mind.

If any of the other Mercurial developers would like to chime in and suggest corrections or additions — or if anyone has any questions — I’ll be happy to update the post. I hope it’s helpful!