Next: , Previous: , Up: Hacker's guide   [Contents][Index]


3.16 Using GNU Arch

3.16.1 About GNU Arch

Since March, 4th 2010, Liquid War 6 uses GIT to handle source code, track changes, branches, and the rest. It replaces the GNU Arch repository. This old repository contains all sources up to version 0.0.7beta, following versions, including 0.0.8beta, must be retrieved from GIT.

So the following informations only concern those who are interested in previous versions of the game. Anybody else - probably you - should use GIT instead.

See Using GIT.

Still, this quick Arch survival guide is kept in the documentation.

Read the GNU Arch tutorial to learn how Arch works. Note that there are many other source control managers available, some of which provide functionnalities similar to GNU Arch / tla. GNU Arch has been chosen for Liquid War 6 because:

3.16.2 Getting the latest version from the repository

The repository for Liquid War 6 is accessible on http://arch.savannah.gnu.org/archives/liquidwar6. This is a read-only access, but with the distributed nature of GNU Arch, it still allows you to keep track of your own changes, and submit patches. Accessing it in read/write mode with sftp requires a Savannah account and special rights on the Liquid War 6 project.

Here are typicall commands one can use to get Liquid War 6 source from the GNU Arch repository:

tla register-archive http://arch.savannah.gnu.org/archives/liquidwar6
tla get -A liquidwar6@sv.gnu.org liquidwar6--beta

All the patches in the archive are signed with GnuPG, so you can check their authenticity with my public key.

You might need to edit your $HOME/.arch-params/signing/=default.check file and put the following text in it:

tla-gpg-check gpg_command="gpg --verify-files -"

3.16.3 Setting up your own arch repository

This section is for those who want to hack the game and set up their own repositories. This will enable you to keep track of your patches, package them, and help the core maintainer merging them in the main repository.

You can introduce yourself and create a repository by issuing commands like:

You can introduce yourself and create a repository by issuing commands like:

tla my-id me@home.net
tla register-archive me@home.net--2008 /home/me/tla-archives

Then, you can get create your own repository, with a command like:

tla tag -S liquidwar6@sv.gnu.org/liquidwar6--beta--0.1 me@home.net--2008/liquidwar6--beta--0.4

The idea is that you create, locally, a depot which has a name that matches the name on savannah (this is for convenience, you could technically give it any name...) and indicate that they represent, today, the same thing.

You can get a working copy of your depot with the command:

tla get me@home.net--2008/liquidwar6--beta--0.4

This will create a complete source tree, which you are free to modify, this is where you should hack.

3.16.4 Synchronizing your repository with upstream releases

To synchronize yourself with upstream developments, go into your copy (the directory created by tla get) and type:

tla star-merge liquidwar6@sv.gnu.org/liquidwar6--beta--0.1

This will apply locally all the changes that happened since the last synchronization. Of course this is one way to work, you can decide to cherry pick patches and such stuff, but for most dayly uses, a good’ol star-merge is fine.

Not that star-merge will only apply patches on your working copy, not on your repository. The only way to actually commit the modifications on the repository is to use the commit command.

3.16.5 Submitting patches

When using Arch, you can of course still send patches created with diff, or even send updated files directly, the way you would without revision control.

But it can be more convenient to either

Here’s an example of an mkpatch command, and which will compute the differences between a previous liquidwar6--beta--0.4--patch-2 snapshot and a not yet commited latest version:

tla mkpatch {arch}/++pristine-trees/unlocked/liquidwar6/liquidwar6--beta/liquidwar6--beta--0.4/me@home.net--2006/liquidwar6--beta--0.4--patch-2 . my-patch

This will create a my-patch directory, which can be gzipped and sent by mail.

3.16.6 Recover from broken lock

Sometimes, when signing a patch, you might enter the wrong passphrase several times, or you might press CTRL+D inadvertantly. In that case, tla will be in a half-broken state, telling you it can’t acquire revision lock... A quick workarround for this is to go to the depot, find the latest patch, and in this repository, create the following folders:

++revision-lock/+contents

Both are directories, note the two ++ and the single +. the +contents directory can be empty. Once you’ve done this, try to re-commit.


Next: , Previous: , Up: Hacker's guide   [Contents][Index]