Take a look at the wiki for user instructions.

Important note: building from a github-generated tarball does not work! You must either build from a git checkout or from a 'make dist' tarball.

The rest of this file contains information relevant only to:

1. Distributors.
2. Contributors.

TMWA has been maintained by o11c (Ben Longbons) since early 2011 or so. Before that, it never really had a proper maintainer, since everyone thought that ManaServ was going to be the thing. But it won't ever be, at least not for TMW.

TMWA has a bugtracker.

But it's probably worth getting on IRC first:

• Use an IRC client: irc://chat.freenode.net/tmwa
• Or just use the webchat.

Note that this channel is only for technical discussion of TMWA (and attoconf), not general chat or TMW content development.

I'm active in the Pacific timezone, but I might not have internet access all the time. I'm usually never AFK longer than 48 hours; when there is an exception, I always tell the content devs who also idle there.

1. Distributors.

Important notes:

• Go read version.make
• TMWA requires git to build by default, use 'make dist' to get a tarball.

Platform dependencies:

Architecture:

tested: x86, amd64, x32
likely to work: all architectures (patches welcome if they don't)

Operating system:

known bad: Linux 2.6.26 and earlier
maintained: Linux 3.2 and later
likely to break: Cygwin, BSD

Filesystem:

must support symlinks

Build dependencies:

Python:

required: 2.7.x only, installed in $PATH as 'python'

C library:

recommended: glibc 2.17 or higher
supported: glibc 2.13
known bad: glibc 2.8 or below
unsupported: anything that's not glibc

C++ compiler:

required: g++ 4.7.2 or higher
recommended: g++ 4.8.1 or higher
not recommended: clang++ 3.3 or higher (all versions have unfixed bugs)

C++ library:

recommended: libstdc++ to match g++; may need patch for clang++
may work: libc++

attoconf:

special: see below

Runtime dependencies:

glibc:

depends on what it was built against

libstdc++:

depends on what it was built against

Instructions:

Configuration:

./configure

Takes most of the options GNU Autoconf's configure does - I won't repeat the output of ./configure --help here.

--prefix=/usr, not --prefix usr, in order to prevent an ambiguity. "In the face of ambiguity, refuse the temptation to guess."

Out-of-tree builds work.

Note that there is no option to disable dependency tracking, as it is also used to generate link information. There is also no option to ignore unknown options - I refuse to lie.

Build:

make -jN

Build test:

make test

Nowhere near complete useful yet. Requires source of Google Test.

make format; git diff --exit-code

Install:

make install DESTDIR=/whatever

See what is installed below

Install test:

not implemented

Distribution tarballs:

make dist
make bindist

Note about attoconf:

TMWA's ./configure script is implemented using a python package 'attoconf', which I wrote over a weekend after reading GNU autoconf's documentation and realizing that it was 1. insane, and 2. still trying to solve the wrong sort of problem.

Currently, attoconf's API is still in the "experimental" stage, so the real rule is "does ./configure work?". When it gets to 1.0, it will start guaranteeing compatibility.

Attoconf is available at Github and is a well-behaving python package.

Attoconf requires Python 2.7; a port to Python 2.6 is doable with a bit of work, but it is not known if this would benefit anybody.

If you're Arch - you broke Python for us all, you clean up your own mess. Patches to call a nonexistent /usr/bin/python2 will NOT be accepted.

What is installed:

Overview:

Currently, make install installs 5 binaries, along with a handful of libraries, headers, data files, config files, and debug files, each of which has a make install-something target.

The 4 main programs below are typically running on the same machine, though in theory they may be configured to run on different machines in a fast LAN. Also, the internal protocol between the programs is subject to change without notice, so they must be upgraded simultaneously.

These programs currently read most of their files relative to the current working directory; this was the only thing that makes sense since the files are dependent on the server-data. However, a migration to installed files has begun.

tmwa-monitor:

<DEPRECATED> Formerly known as eathena-monitor.

An unmaintained tool whose job was to keep restarting the servers every time they crashed. It still builds in case anyone was using it, but it proved inflexible and has't really been kept up-to-date with our (TMW's) server-data, and besides, the server doesn't crash much now. There are also a number of other Open Source programs that monitor services already.

There is a run-all script in the server-data repo that starts the appropriate server for that config. On the main server, we instead start the servers (and bots) individually in a tmux.

tmwa-admin:

Formerly known as ladmin ("local").

This is an essential tool to maintain the server's flatfile "databases". It doesn't actually touch the files directly, just connects to tmwa-login.

Even when everything is rewritten to use SQL, it will be kept, if just to keep a consistent interface. In fact, if we use SQLite we can't edit the databases independently. This wouldn't be a problem with Postgres, but people seem to think it's hard to install (that's not my experience on Debian though. Did they ever try themselves, or are they just repeating what they've heard?)

tmwa-login:

Formerly known as login-server.

User-facing server to deal with account checks.

Also accepts internal connections from tmwa-admin and tmwa-char, subject to a plaintext password (and for tmwa-admin, also an IP check).

tmwa-char:

Formerly known as char-server.

User-facing server to deal with character persistence.

Connects to tmwa-login; also takes internal connections from tmwa-map.

Note that it is fully supported for more than one tmwa-char to connect to the same tmwa-login; the client will be presented with a list of "worlds" before leaving the login server.

tmwa-map:

Formerly known as map-server.

Connects to tmwa-char.

It is technically possible for more than one tmwa-map to connect to a single tmwa-char, but this is poorly supported by our current config and moderation tools, and there are likely undiscovered bugs.

About server data:

Just having the binaries is not enough: you also need a complete set of content: config files, game scripts, savefiles, and client updates.

A web server to serve the updates is also strongly recommended, as even developers get annoyed when wushin makes us work straight from his client-data repo.

Currently, there is only one set of server data that is known to be compatible with TMWA: https://github.com/themanaworld/tmwa-server-data

The only recommended way of using this is by following the instructions in the How to Develop article. These instructions are only designed for people contributing to TMW itself, not to people trying to start a fork - we know all forks are doomed to be unsuccessful anyway, so please don't split the development effort, and you can't split the player community.

In particular, the instructions do NOT provide information on how to secure a server by changing all the default passwords.

There are 3 other known sets of complete server data: regional ones for Germany and Brasil, and Evol. Evol requires their own fork of the tmwa server (for some reason they don't like me to call it evola), and nobody seems to know of the foreign servers are keeping active.

Note also that The Mana World has not investigated the copyright status of other sets of server data.

2. Contributors.

The most important thing if you want to help improve TMWA is talk to me. No, wait, that's the second most important thing.

The real most important thing if you want to help improve TMWA is that it's work. You can't just stop by and chat for a few hours and help at all. If you're going to work on TMWA, you have to be work months in the future.

TMWA was terrible when I got it, and I've only fixed enough to make it sane, not pretty. Even a minimal change is likely to touch the whole tree, so merge conflicts are a constant problem.

That said, there are several tasks that I could use help with. Several essential tasks have been left undone just because they don't conflict with the main body of my work.

But I do not want someone who will just work for a few hours, go to bed, then never return. I have wasted far too many hours answering their questions. If you're going to help, you have to actually help.

The following skills are good to know required for various tasks:

• ability to read
• ability to write
• ability to notice error messages
• ability to solve your own problems
• willingness to accept review of your changes. It's not personal if I say your work is wrong, I'm just seeing more than you do, and tiny details are often incredibly important.
• familiarity with gdb
• Python (A low entry barrier, but Python alone is not enough for the tasks. Particularly, reread the bit about review.)
• C++11 (Not a low entry barrier. I'm not really expecting help with this, and since this is conflict heavy, please do the other tasks first).