Thursday, February 22, 2024

“Alex the Allegator 2” for Linux

Alex the Allegator 2 also didn’t have a Linux port yet :-)

This time it’s a puzzle game.



Tuesday, February 20, 2024

“Alex the Allegator 1” for Linux

I recently stumbled over some old Linux games by Free Lunch Design (Alex the Allegator 4 to be exact) and found out even more had the source code available!

This is a quick DOS to Linux port of the original Alex the Allegator 1 from the Alex the Allegator series, which doesn’t seem to have been ported yet.

I admit the game is very minimal - my first playthrough lasted about 3 minutes - but now I (and hopefully also you) know what we missed at least ;-)



Have fun!

Sunday, February 18, 2024

Getting various MIDI applications to work

From time to time I stumble about older applications (usually games) which use MIDI for music playback. These are some notes on how to get things working with various libraries and will be extended from now on every time I stumble upon something.


MIDI was originally meant to be supported by hardware, but when we are talking about today’s soundcards you won’t find any consumer sound card with hardware MIDI support any more. Because of that you need software to do the job.

This is where software such as TiMidity++ or FluidSynth come into play, but this article will only cover the requirements for specific libraries. The focus of this article are the requirements of specific applications / libraries however; for a general overview see e.g. the ArchWiki entry on MIDI.

The most important thing first

If everything you try just doesn’t want to work, make sure that the applications and libraries you are using are compiled with MIDI support built-in! Nothing is more frustrating than debugging for hours just to find out that the packager didn’t care about MIDI support when creating the package. Yes, that happened to me several times, so if in doubt, do yourself a favor and compile the relevant parts yourself just to make sure.

Allegro 4

Allegro is a cross-platform library for creating multimedia applications, including games. Up to version 4 it also included MIDI support.

On Linux the following MIDI drivers are supported:

MIDI_OSS - Open Sound System
MIDI_DIGMID - sample-based software wavetable player

I never tested OSS MIDI setup.
Allegro’s ALSA driver unfortunately only supports raw MIDI devices (which could be emulated by TiMidity++ if one wanted).
If you don’t have one set up, the autodetection will use the DIGMID software emulation, and that’s the easiest to set up. It just requires one file with the MIDI instruments in a specific location - otherwise you simply won’t hear anything.

I’ll quote the readme.txt file:

The DIGMID wavetable driver uses standard GUS format .pat files, and you
will need a collection of such instruments before you can use it. This
can either be in the standard GUS format (a set of .pat files and a
default.cfg index), or a patches.dat file as produced by the pat2dat
utility. You can also use pat2dat to convert AWE32 SoundFont banks into
the patches.dat format […].


  • Get a sound bank in “SoundFont 2” or “GUS” format (see link to ArchWiki above for links).
  • If it’s in “SoundFont 2” format: Convert it using pat2dat -o patches.dat /usr/share/soundfonts/CT8MGM.SF2 (I’ve used the SoundFonts included with my ancient SoundBlaster 16 for this).
  • Copy the file(s) to /usr/share/allegro/ (which I’d recommend, then all Allegro based applications will use it) or directly into the application’s directory.

Tuesday, February 22, 2022

Playing German “Marathon 2: Durandal” with Aleph One

Or: How to extract “Resource Forks” on Linux

Or: What happens if you want to access Mac data from a game disc without a Mac? A self experiment.


You may know the Marathon series: Three first person shooters games, which may be seen as a predecessor of the later “Halo” series by the same company. There is one caveat though: Back then, developer Bungie Software was a very Mac centric company, and most of their games didn’t even have ports to Windows or other platforms.

The exception in the Marathon series is the second part, “Marathon 2: Durandal”, which is the only one that was also ported to Windows. But not just that, the game was even localized into German (and probably a few other languages as well).

Fast forward: In 2000, something wonderful happened. The source code of the game engine was released as Open Source, and in 2005 the games themselves were released as freeware. That is: The Mac version of the game engine was open-sourced, and the Mac version of the games were released as freeware. And just the English versions of them.

So now I’m looking at my beautiful German box of Marathon 2 and am wondering: Can I also play that German version on Linux?

Of course that’s possible, but it turned out that getting the data files is more complicated than expected…

Mounting the Mac part

So let’s have a look at the disc: It contains both an ISO9660 part for the Windows version and a HFS part for the Mac version.

The source code, meanwhile curated by the wonderful people of the Aleph One project, is - of course - only compatible with the Mac game data, so let’s try to mount the HFS part of the disc (by default the ISO9660 part would be mounted):

mount /dev/sr0 -t hfs /media/cdrom

If that command fails for you, your distribution probably didn’t include the hfs kernel module - but don’t worry, you won’t need it anyway. Apart from some strange characters (the kernel hfs module doesn’t seem to be very good with special characters - you can see the output in the “HFS” link above) all files seem to be there.

So: Let’s try to start the game:

$ alephone /media/cdrom/Marathon 2 �/
fatal alert (ID=-1): Please be sure the files 'Map', 'Shapes', 'Images' and 'Sounds' are correctly installed and try again. (csalerts_sdl.cpp:202)

OK, the manual mentions that you can start the Mac version directly from CD, so something seems to be off.

$ ls -l /media/cdrom/Marathon 2 �/
total 33648
-rw-r--r-- 1 root root        0 Nov  6  1995 Iconr
-rw-r--r-- 1 root root        0 Oct 10  1996 Images
-rw-r--r-- 1 root root        0 Nov  6  1995 Manual (english)
-rw-r--r-- 1 root root  5704325 Oct 14  1996 Map
-rw-r--r-- 1 root root   407802 Oct 21  1996 Marathon 2
-rw-r--r-- 1 root root  4136840 Jul 13  1995 Music
drwxr-xr-x 1 root root        4 Dec 10  1996 Physics Models/
-rw-r--r-- 1 root root 10016800 Nov  6  1995 Shapes
-rw-r--r-- 1 root root 14183536 Nov  6  1995 Sounds

What’s that? The Images file is empty? What happened here?

(One may wonder: How did I enter the unprintable character if tab completion stops at exactly that place because of another directory starting with the same name? cd /media/cdrom/Marathon 2 [^S], followed by pressing tab. I know zsh users will probably laugh now because they can simply cycle through the directories with tab by default…)

Exporting Resource Forks

It was clear now that either the kernel driver had a bug here, or that the information would be stored elsewhere. To exclude a kernel driver bug I decided to cross-check with another open source tool: HFSExplorer also promises to be able to view HFS file systems.

And surprise: HFSExplorer is able to decode the special file name characters, so we now know that the game is stored in folder called “Marathon 2 ƒ”, but the Images file is also shown with a size of 0 byte here.

The HFSExplorer download however does also include a command line utility called unhfs, and Heureka! The following command revealed the mystery:

unhfs -resforks APPLEDOUBLE /dev/sr0

Now there are suddenly two files: Images and ._Images - with a size of 4,8M that seems to be the file we are looking for. file ._Images reveals that this is a AppleDouble encoded Macintosh file. So: what the heck is this again?

Decoding AppleDouble encoded Macintosh files

For a short description of what the AppleDouble format is see e.g. the AppleDouble documentation of the Archive Team. In short: It’s a Metadata store - basically some parts of the data is stored as metadata, while other parts are stored in the actual file.

Unfortunately HFSExplorer can only export the metadata in AppleDouble format, while Aleph One can only read AppleSingle, MacBinary and RAW data. One option may be to convert the data to AppleSingle or MacBinary, but I haven’t found a tool which would create working files. So let’s create RAW files instead.

Analyzing the metadata is possible with several tools: apple_dump (from the Netatalk project / package will print some information, including an offset of 38 and a raw dump in hex and ASCII.

$ apple_dump Images | less
"Images" is not AppleSingle/AppleDouble format.
"/xxx/Marathon 2 ƒ/._Images" is found.
Dumping "/xxx/Marathon 2 ƒ/._Images"...
MagicNumber: 00051607                                        : AppleDouble
Version    : 00020000                                        : Version 2
Filler     : 4D 61 63 20 4F 53 20 58 20 20 20 20 20 20 20 20 : Mac OS X        
Num. of ent: 0001                                            : 1

Entry ID   : 00000002 : Resource Fork
Offset     : 00000026 : 38 
Length     : 004C21F8 : 4989432

-RAW DUMP--:  0  1  2  3  4  5  6  7  8  9  A  B  C  D  E  F : (ASCII)
00000000   : 00 00 01 00 00 4C 20 F2 00 4C 1F F2 00 00 01 06 : .....L ..L......
00000010   : 00 00 23 4C 0B 74 72 6F 6F 70 65 72 2E 72 61 77 : ..#L.trooper.raw
00000020   : 02 00 00 00 72 73 72 63 52 53 45 44 01 00 FF FF : ....rsrcRSED....
00000030   : 06 49 6D 61 67 65 73 49 02 00 00 00 69 6D 67 32 : .ImagesI....img2
00000040   : 35 32 2E 34 01 00 00 35 00 2B 00 00 00 00 1E A7 : 52.4...5.+......
00000050   : 00 00 69 6D 67 32 35 32 2E 34 01 00 00 35 00 2B : ..img252.4...5.+
00000060   : 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 : ................
00000070   : 00 02 AC 28 59 B0 00 00 00 00 00 4C 21 F8 00 00 : ...(Y......L!...
00000080   : 00 00 00 00 00 00 1B 00 00 00 23 4C 15 77 65 61 : ..........#L.wea
00000090   : 70 6F 6E 73 2D 69 6E 2D 68 61 6E 64 2E 70 6C 61 : pons-in-hand.pla
000000A0   : 74 65 02 00 00 00 50 49 43 54 38 42 49 4D 05 00 : te....PICT8BIM..

Despite of its name I ironically couldn’t find a way to just dump the data with apple_dump instead of printing the nicely formatted output.

Similarly Lsar from The Unarchiver (a collection of unpackers for various typical Macintosh formats) will also print the offset:

$ lsar -L ._Images
._Images: AppleSingle
  Name:                      ._Images
  Size:                      4.99 MB (4.989.432 bytes)
  Compressed size:           4.99 MB (4.989.432 bytes)
  Is a Mac OS resource fork: Yes
  Index in file:             0
  Start of data:             38
  Length of data:            4989432

But also unar, the extraction part, didn’t want to dump the actual data.

But if it’s really just an offset of 38: dd to the rescue ;-)

dd if=._Images of=Images.rsrc bs=38 skip=1

In case you are wondering about the extension: .rsrc is used by Aleph One to look for the RAW resource data.

Repeat that for the Map file:

dd if=._Map of=Map.rsrc bs=38 skip=1

And voilà: Alph One will now accept the data files, and most components, including the main menu and the ingame texts are localized now! Just the UI itself seems to be hardcoded.

Saturday, August 28, 2021

Recover boot sector from floppy disks with Parity Boot Virus

I recently found some old floppies, back from the times when I was young, was using Microsoft DOS and didn’t have any real clue about computers. As a result, almost all of my floppy disks were infected with the Parity Boot Virus.

The Parity Boot Virus had the habit of overwriting the complete boot sector, i.e. the first sector of the floppy disk (if it wasn’t write protected at least). BUT it was also fair enough to create a backup copy of the boot sector and - on 3,5′’ floppy disks - put it into sector 32. See the Parity Boot Virus page on the Malware Wiki for more information.

25 years later I wanted to mount one of my floppy disk images I had backed up back then on my Linux machine and was surprised to see a lot garbage files like |úI|ë?K|.╕ - certainly not a valid file name. It turned out that sector 32 is the last sector to contain the root directory¹. And it also turned out that Linux reads all clusters and scans them for directory entries, while Windows seems to stop earlier, probably as soon as it encounters an empty file entry.

So how can you restore the original boot sector and get a clean root directory again? Simple, just call

dd if=$file count=1 skip=31 of=$file conv=notrunc seek=0
dd if=/dev/zero count=1 of=$file conv=notrunc seek=31

myfloppy can either be a physical floppy drive (e.g. /dev/fd0) or the file name of an image.

Please just make sure that sector 32 really does contain a valid boot sector - you can view the disk’s contents with hexdump -C myfloppy, check the contents between 00003e00 - 00004000. It should at least contain the label and the file system (e.g. NO NAME FAT12 in lines 3 and 4.

¹ usually, on a 3,5′’ HD floppy disk at least - see this website for an excellent disassembly of the FAT12 file system (German) and this Wikipedia article about the actual default sector size for the directory on a given floppy type if you want to know more.