Open main menu

OpenTasks

Revision as of 16:12, 6 March 2007 by Fingolfin (talk | contribs) (added basic rules, removed TOC)


This page contains a list of open tasks. Completing any of them would benefit the ScummVM project a lot. At least in theory, all of them should be doable even by somebody relatively new to the project, in particular, as part of the Google Summer of Code. Besides the tasks listed here, there is of course also the TODO page listing more things that need to be done (but with far less details).

Some basic rules

Below follow some basic rules that anybody interested in one of these tasks should adhere to. Sometimes exceptions may be possible -- as always, commen sense applies, and if in doubt, ask.

The projects below are sketches and ideas. Plus, things evolve over time, so the descriptions might be slightly outdated by the time you read them (although we strive to keep them up-to-date). Hence, you should talk with somebody from the team, probably the person(s) listed as Tech Contact, before starting work on any of them.

All code, unless stated differently, must be written in clean and portable C++, in particular, GCC must be able to compile it (portability exceptions can be made for platform specific code, of course). Using the standard C++ lib for code used inside ScummVM is at this time not possible. Using it inside non-essential tool should be fine, though.

All of the code submitted must be contributed under the terms of the GPL.

We only accept clean and maintainable code. This is a somewhat vague requirement, but as a rule of thumb, if the code does what it is supposed to do, but is not extensible, a quick hack, or we need to rewrite it from scratch to properly integrate it, we will not accept it. In particular, we would rather have a maintainable, clean and incomplete piece of code that we could extend than a complete but unmaintainable piece of code.

Generic infrastructure tasks

Plugins

Technical Contact: Max Horn

Background:

ScummVM contains by now over a dozen different engines for different adventure game systems. Each of these engines forms a "plugin", and it is possible to compile these plugins as loadable modules separate from the core of ScummVM itself. This makes it possible to (un)load engine code at will, ideal for systems with tight memory constraints; if you want to play a game, you only need to load the engine for that specific game, and can unload the others.

We always compile the engines as "plugins", but by default, they are compiled as so-called "static" plugins; these are not real plugins, but rather the engine gets compiled into the application (i.e. "static linked", hence the name), but is still treated by the rest of the application as a plugin (only that it is not dynamically loaded). Support for true "dynamic" plugin is only available on a subset of all plugins -- essentially most POSIX systems (including Linux, *BSD, Mac OS X) and the Dreamcast.

The Task:

There are various ways in which this plugin code could be improved:

  • Add support for real "dynamic" plugins on Windows. This is currently not possible, because the ScummVM plugins rely on "backlinking" from the loaded modules to the core binary. This is not supported on Windows. There are various ways to solve this issue (use Google for more info). This is further complicated by the fact that while the interface of the *plugins* is well-defined (although not quite as well documented -- see base/plugins.h), the converse is not true -- plugins can call virtually arbitrary code within the core of ScummVM
  • Add support for "dynamic" plugins on more systems which might benefit from them.
  • Improve the way we search for plugins. Right now we only look in the current directory. Should this be made configurable? On Unix, maybe /use/share/scummvm should be searched, too? On Mac OS X, the CoreFoundation bundle API should be used to search for plugins inside the ScummVM.app bundle (which would also required the build system to be enhanced to properly bundle all plugins). Etc.

Small devices backend

Technical Contact: Eugene Sandulenko, Max Horn

Background:

ScummVM has been ported to many platforms, often by simply re-using the SDL backend (which is based on SDL, which by itself has been ported to many platforms, making it fairly easy to port ScummVM to any of these platforms, too).

But for some platforms, dedicated backends are required, either because SDL doesn't support them, or because we can't achieve all our needs by using the SDL port (e.g. because we need more speed, more control, etc.). These backends are typically made for what we call "small systems" -- systems like PDAs or smartphones (PalmOS, Symbian, WinCE), or game consoles (Dreamcast, GP2X/GP32, Nintendo DS, PlayStation 2, PlayStation Portable). These systems share many features. In particular they often have no (full) keyboard and quite limited resources: Little RAM, little permanent storage space, not that much CPU power, or a limited screen resolution.

This makes it often necessary to (re)implement certain functionality, like virtual keyboards, or graphic downscalers.

The Task:

Since the same needs occur again and again, it would be nice to implement such functionality only once in a sufficiently portable and flexible way, making it possible for backends to pick and use whatever they need. Details and further suggestions as to how to achieved this can be found on the Small Devices Backend page.

Besides this, optimizing code for speed and memory usage benefits all our targets, and in particular these "small devices". Hence doing this is a worthy goal on its own.

File system abstraction layer

Technical Contact: Max Horn

Background:

ScummVM runs on a great variety of platforms, ranging from full-blown personal computers running Windows, Linux, Mac OS X, etc., to game consoles, smart phones and PDAs (sorry, no support for toasters yet :).

To achieve that, we have to provide abstract APIs for many things, like audio and graphics, user input, and general I/O. One particular area that needs to be covered are filesystems -- great differences exist with regards to this between the platforms we support. In fact, some of them don't even have such a thing as a filesystem.

Still, we need to access files: we have to find and load the game data, GUI themes, save states, config files and more (and also have to store save states, obviously). To compensate for the differences between our target systems, we introduced the FilesystemNode API.

In its present state, it fulfills our absolute minimal requirements. However, the abstraction level it provides is flawed and likely should be overhauled / redesigned. Also, not all parts of ScummVM strictly use this API; there are still many parts of ScummVM which try to work with "paths" to access file data.

The Task:

To work on this, you should be familiar with at least Windows and Unix, ideally more, and be willing to learn about the differences between file systems on many different platforms (our team will be happy to help you with this). You will have to design a revised FilesystemNode API, and implement it, based on the needs of ScummVM engines and infrastructure code (requiring close communication with the rest of the team).

GUI

Technical Contact: Eugene Sandulenko, Johannes Schickel

Background:

ScummVM implements its own GUI code. We can't use any of the portable GUI kits out there, as they are mostly not portable enough, use too many resources, or simply are not flexible enough to work atop our backend system. Also, our GUI must look equally good with 8bit, 16bit or 32bit graphics, at 320x200, 320x240 or 640x480 resolutions -- and ideally also at 240x200 or at 1280x1024.

Currently we have a quite flexible system for defining the look and feel of our GUI. Almost every aspect of it can be specified via external theme file described here.

The Task:

There are many things missing:

  • Shadows are not configurable
  • Not all widget aspects such as font to use are configurable
  • We need to implement possibility for backends to extend GUI by adding their own tab in options dialog
  • It would be nice to have antialiased fonts (possibly optionally -- we still need to support relatively low end devices, after all)
  • Perhaps implement portable support for TrueType fonts
  • We need better About dialog
  • We need dialog for save game selection
  • There are a number of small things to do which will make GUI behavior more predictive
  • Modern Theme renderer should be reworked to make it faster, for example currently shadow drawing is rather slow (and hacked)

Audio related tasks

Mixer improvements

Technical Contact: Max Horn

Background:

Most games we supported need to play sampled audio data (PCM encoded). In addition, some code which emulates certain audio devices (like the MT-32 emulator, the FMOPL code or Amiga MOD players) need to do so. To facilitate this, we provide certain APIs: The so-called Mixer object takes a number of AudioStream objects and mixes the audio data it fetches from these together, taking into consideration various external parameters like volume and pan controls. The abstract AudioStream class in turn provides a convenient and flexible way for code which needs to generate and output sampled audio data.

Audiostreams can provide their data with arbitrary sample rates, and the mixer has to convert these to sample rate by the audio output device provided by the backend (which usually means converting to a sample rate of 22050 Hz or 44100 Hz). Currently, we only have a simple linear rate converter.

The Task:

There are much better ways to resample audio data. It should be band pass filtered and other things done to it. Code to do this exists out there. Integrate such code into ScummVM or roll your own. Make sure to keep resource constraints (CPU, memory) in mind. It should be possible to select which resample to use (the low quality but fast linear filter, or your high quality but not quite so fast fancy filter).

Furthermore, on some systems, ScummVM suffers from stutter even though there is plenty CPU power left. The problem here is latency (not throughput). That is, while we can generate data fast, we do so too late. The problem on these system likely could be fixed by using a double buffering approach, i.e. generate audio data slightly before it is requested to avoid latency issues.

Finally, our AudioStreams currently only can have a single fixed sample rate. But for some usages (in particular, Amiga MOD players), it would be convenient to allow streams to change their sample rate on the fly. Adding support for this would be nice, but should not slow down streams which have a constant sample rate.

Add support for TFMX, and more Amiga MOD formats

Technical Contact: Eugene Sandulenko, Sven Hesse

Background:

Since we're adding (and have added so far) support for different Amiga game versions, we need support for the sound files too. Currently we need support for following formats:

  • TFMX a (non-MOD) format used by Monkey Island 1. There is a player for it already, which is written in rather poor C, also it misses marco implementation. This documentation may be useful, it isn't complete though. If you don't have the Amiga version of Monkey Island 1 you can use the demo of Monkey Island 1.
  • MaxTrax support for Kyrandia 1. The Amiga version isn't supported yet, but LordHoto will be working on it when his copy arrives. A Motorola 68000 (short: 68k) assembler implementation by Joe Pearce can be found here. It would be necessary to re-implement it in proper C++, naturally a must for this task is 68k assembler knowledge. If you don't have the Amiga version of the Legend of Kyrandia, there are sample modules available.
  • SoundFX is used by the Future Wars and Operation Stealth games of the cinE engine. The player is already implemented, but it cannot load music stored in Amiga format with combined track info and instrument info. The existing code only supports loading them separately, which happens to be the way they are stored in the DOS version. If you don't have the Amiga versions of the required games, you can use the demo of Future Wars and the demo of Operation Stealth.

Rewrite FMOPL emulator

Technical Contact: Eugene Sandulenko

Background:

Many games supported by ScummVM contain sound data intended for the AdLib Music Synthesizer Card, or other AdLib-compatible hardware like the SoundBlaster. These sound cards used the Yamaha YM3812 sound chip (also known as OPL2) to produce sound via FM synthesis. For more information about how this is done, see Jeffrey S. Lee's web page on Programming the AdLib/Sound Blaster FM Music Chips.

In ScummVM we use the well-known FMOPL emulator from MAME V0.37a with some modifications (the name is the composition of "FM" and "OPL", see above). That was one of the few versions on MAME released under the GPL, making it legal for us to use this particular version of the FMOPL code. Jarek Burczynski later rewrote the FMOPL emulator to be bit-precise, but unfortunately for us his work has only been released under a GPL-incompatible license.

The Task:

What needs to be done is to understand Jarek's current code and write a new emulator. Efforts must be made to avoid verbatim copying of his code, to ensure no license violation takes place.

There are several approaches to do this, but most are not suitable for a single developer. A possible approach would be to fix our current FMOPL code by looking at how Jarek's code works. Another would be to start with a second, totally different emulator which we have and gradually improve that one (which was written by Ken Silverman and released under the GPL).

Fingolfin has some doubts whether this is a realistic project. In particular, the approaches suggested above would possibly still violate Jarek's license, as most license do cover "derived" work -- and looking at some code and then writing your own definitely falls into that category

MIDI enhancements

Many of the adventures supported by ScummVM make use of MIDI music. Which is why we already include several device drivers for various MIDI APIs and emulators (e.g. ALSA, Windows MIDI, Mac OS X CoreAudio/CoreMIDI, fluidsynth...).

MIDI device configuration

Technical Contact: Max Horn

Background:

Right now, the MIDI drivers are treated by ScummVM in a rather single minded fashion: Either a driver is linked in and hence "available", or not. It's not possible to configure anything about them (like ports to be used etc.), nor does it ever take into account that a single driver might correspond to multiple devices (after all, you can plug several sequencers into your MIDI port; or you could have configured several different sound font settings in your MIDI emulator).

The Task:

  • Add an API for querying the OSystem backend for a list of available MIDI devices (not drivers)
  • Information about the selected device must be serializable, so that it can be stored in the config file
  • Selection of devices via command line should be possible
  • It must deal with devices being added/removed (at least between runs of ScummVM, ideally also while ScummVM is running)
  • Devices should be configurable via the GUI; this needs to be done in a flexible (different devices/drivers offer different settings) and portable fashion.

XMIDI parser

Several of our games make use of the XMIDI format. We already have a parser for it (see sound/midiparser_xmidi.cpp), which was based on code from the Exult project, but it is incomplete. Specifically we require support for XMIDI_CONTROLLER_FOR_LOOP and XMIDI_CONTROLLER_NEXT_BREAK, the XMIDI code (see audio/midi_drivers/XMidiSequence.cpp) from the Exult project, could be used as reference again.

Revive ScummEX (likely from scratch)

Technical Contact: Max Horn

Background:

ScummEX (short for "Scumm EXplorer") is a tool which allows browsing the data files of games supported by ScummVM. That's the theory. Current reality is that the ScummEX version in our Subversion repository only supports a subset of SCUMM games (i.e. not even all SCUMM games, not even talking about *all* games supported by ScummVM). It also is based on an old version of wxWidgets, and written in a rather crude way. Hence, starting from a clean new code base likely is the best way to approach this.

The Task:

The goal would be to write a new graphical tool which allows browsing of data files of games supported by ScummVM. Some requirements:

  • It must run on at least Linux, Windows XP, Mac OS X -- more are preferable, of course. You could use wx.
  • Should likely be written in C++ (other languages, like Python or Java, might be acceptable, if the result is sufficiently portable.
  • Browse at least the data of SCUMM v5 till v8 games.
  • Support for older SCUMM versions, and HE SCUMM games, as well for non-SCUMM games, is not necessary, but the code should be written modular/abstract enough to make it possible to add support for more data formats w/o too much trouble
  • Only viewing data is necessary, not being able to edit it (although we won't complain if that's possible :)
  • The interface can be designed in many different ways, but at least SCUMM data tends to suggest a main view with some kind of "data tree", in which one can select specific resources, and view them.
  • Viewing the raw hex data of resources is required.
  • It should be possible to add custom "views" for data (e.g. viewing images as images, not just as hex data; or hooking up a script disassembler; etc.). It is *not* necessary to provide such viewers for all resources types initially, but it must be possible to add them.

Tools

Tools: Game script (bytecode) decompiler

Technical Contact: Max Horn

Background:

Most (all?) adventure engines ScummVM supports are driven by bytecode scripts, which control the game behavior. For various reasons (in particular, debugging), it's very useful to be able to take such a script and decompile it into something resembling regular code (with variables, loops, if statements etc.). We have a tool for doing this for the bytecode used by SCUMM games. This tool is called descumm.

Currently, descumm use a relatively crude heuristic to detect "if" statements, loops and so on. In particular, it only supports loops with the condition at the start (think "while(...) { ... }" ), not such with the condition at the end (as in "do { ... } while(...)" and also suffers from various other limitations (like not being able to correctly recognize "continue" and "break" statements.

The whole thing is somewhat complicated by the fact that there are two main versions of this bytecode: V5 and older (register based bytecode), and V6 and newer (stack based).

The Task:

Write an improved bytecode decompiler for SCUMM games, based on the existing "descumm" tool. Initially, support for only one version of the SCUMM bytecode (e.g. plain V5 or V6 or V8) is needed, but everything should be designed and implemented such that it is possible to support all SCUMM versions eventually. Even better would be if the core of the decompiler was so generic that it could be reused to decompile other bytecode dialects (and hence write improved versions of the "dekyra" and "desword2" tools).

A good starting point for this might be the [Jode] Java bytecode decompiler, the author of which used to be on the ScummVM team. The technical contact has several ideas on how to approach this project, too.

The tool should run on at least Windows, Mac OS X and Linux. Acceptable languages include C/C++, Python, Perl, Ruby -- other languages might be OK, but please consult with us first (after all, other developers also will have to use and maintain your code).

Tools: Overhaul the compression tools

Technical Contact: Max Horn

Background:

We offer a multitude of command line tools in a separate package (scummvm-tools). The majority of these tools are used to (re)compress audio data. This greatly helps users who want to play their games on devices with limited storage, like PDAs and smart phones.

The user can choose between MP3, Ogg Vorbis and (in those cases where it makes sense) FLAC compression. The tools take the original data files, extract sound data, compress them, and reassemble everything into new (smaller) data files.

The Task:

There are various ways in which these tools could be improved:

  • Write a portable GUI for them -- that's a big task and hence has its own section (see below)
  • Unify their command line interfaces
  • Unify the actual compression code -- some partial work on this exists, but with a crude ad-hoc API, which makes it very difficult to use them and also inefficient
  • Right now, compression works by invoking external encoder binaries (oggenc, lame, flac). To this end we write data into a temporary file, invoke the external encoder, and read back the file it created. It would be nice if one could instead link against suitable encoder libraries, and provide compression binaries which are fully self-contained. Also this would make it possible to perform in-memory compression (which could provide a noticeable speedup).

Tools: Write a portable GUI for the tools

Technical Contact: Max Horn

Background:

We offer a multitude of command line tools in a separate packages (scummvm-tools). Several of these tools are to be used by end users, mainly to (re)compress their data files (good for systems which have not that much storage, like cell phones). But asking end users to use command line tools to perform common tasks is somewhat problematic.

The Task:

We would like to make it easier for users to compress their data files, by offering them a nice graphical user interface (GUI). It should support at least Linux, Windows, Mac OS X, ideally more. This could for example be achieved by using wxWidgets or Qt.

Note that the command line versions of the tools should still be available, for use in shell scripts or on systems where the GUI is not available. This likely requires modularizing the code sufficiently.

Also note that this task meshes well with the previous task (of overhauling the compression tools).


Engine/game specific

Residual: Light-weight software renderer

Technical Contact: Pawel Kolodziejski

Background:

Residual currently offers two different renderers: OpenGL and TinyGL. Neither offers an ideal solution. Not all OpenGL drivers accelerate the operations needed by Grim Fandango, and the TinyGL renderer, apart from being somewhat glitchy, doesn't run well on low-end hardware.

The Task:

It would be nice to have a more light-weight cross-platform renderer. Since Grim Fandango uses mostly static backgrounds, we could probably gain a lot just from more intelligent screen redrawing.

This shouldn't require any deeper knowledge of the game engine. The renderer appears to be fairly well separated from the rest of the engine already.

Objectify CinE engine

Technical Contact: Eugene Sandulenko, Pawel Kolodziejski

Background:

The cinE engine started out as an external project started by Yaz0r. Originally it was written in plain C. ScummVM is a C++ project, so we need to objectify this engine without changing/breaking its behavior.

The engine itself is well-structured, hence many functions / variables which might be good candidates for being grouped together into a C++ class are already grouped by files.

No deep knowledge of the engine internals is required, but in any case the Engine is not that big, and thus it should be possible to learn enough about it to start working in a relatively short amount of time.

We have previously "objectified" several other engines, namely SAGA, Gob and AGI, so one can learn a lot about various approaches how to do this by tracing through our SVN repository.

Improve AGI engine

Technical Contact: Eugene Sandulenko, Stuart George

Background:

Several tasks in AGI engine do not require to own original Sierra games. Generally, Sierra games work better than fan made ones.

  • Implement AGIPAL, AGI256 and AGI256-2 engine extensions
  • Fix bugs triggered by fan made games
  • Finish implementation of Apple IIgs sound
  • Finish implementation of predictive input
  • Amiga style (colored) menus
  • Hercules mode

Some of above mentioned tasks will require engine knowledge, but the engine itself is tiny and well documented, so it should be doable. Additionally NAGI and DAGII could be used as a reference. Particularly, NAGI is known to be more compatible with fan made games

Add 16bit graphics support to SCUMM engine

Technical Contact: Travis Howell, Eugene Sandulenko

Background:

The SCUMM engine was originally developed for palette-based graphics. At version 6 it was forked by Humongous Entertainment, which extended it significantly. Their later games started to use 16bit graphics for backgrounds and actors. See here for more detailed information.

This task requires good knowledge of C++, as we need a solution which will not clobber our code, will have minimal impact on 8bit games, and can be optionally turned off at compilation stage.

If you don't have any of the required games, there are several demos available. A 16bit graphics demo example would be the Freddi5 demo.

Implement "return to launcher" feature

Technical Contact: Max Horn

Background:

Presently we have to exit the ScummVM application completely when users exit a game because most engines do not clean up memory properly on their exit.

The Task:

What we need is to analyze what's going on there and plug all memory leaks, properly shut down subsystems like sound, so it will be possible to play more than a single game within one session.

The task will require good use of a memory leak analyzer and C++.

Port specific projects

Porters, add your ideas here -- and soon, before we submit the project for review!