Sunday, December 23, 2018

Seasons greeting and off on holiday

Seasons greetings to all, and I hope everyone has a happy time as we finish 2018 and look forward to the beginning of 2019. As I sit in the Los Angeles airport writing this, I'm reminded yet again of how much fun I have doing reverse engineering, since even the smallest thread of work can lead you in unexpected directions as you trace the flow of what methods call a given method, and what methods it calls.

For something to do, I opened up my Companions of Xanth in progress disassembly to do a bit of pottering around with it. Just by chance, scrolling through the disassembly, I noticed a call to a method I was pretty sure took in a string, but in this case was passing in a value. I then quickly realised it was an offset n the data segment that hadn't been properly resolved, so I did that. The message in question turned out to be a generic "insufficient memory". So from it's usage, I was able to identify the entire method it was in.. it's called during startup, and ensures that there's enough free memory to play the game. Ironically, it looks like the code to actually display the message if there isn't enough was broken, as evidenced when I ran the game in DosBox debugger.

From there, identified a method just before the message call that took in a number and passed out a string.. it wasn't too hard to identify it as a number formatting routine. I confirmed it in the DosBox debugger by putting a breakpoint in the machine code, and seeing the result from the method. I then noticed that there was one other method calling it, so I was intrigued to find out what it was. It turned out it too was a block of code in a method that printed out the free memory, in a method that looked like a big switch based on a single value. It looked like a keypress. Playing the game, I was able to confirm that pressing 'M' on the keyboard did call the code, which printed out the amount of free memory in the status area.

This was a good woo-hoo method. I had identified the method for handling keypresses. Likely, some of the other keypresses do fairly identifiable things, like saving and loading, so knowing where each key's code is, I'll be able to trace into what they do, and what methods they call. In fact, the free memory printing for the 'M' key already prove dividends. From it, I was able to identify the method that cleared the status area, and a version of 'sprintf' for printing text in the status area. In the future I'll be able to look further into other places that call the method to add lines to the status area, and start identifying other parts of the game code. For example, looking at an item produces a description in the status area, so by setting a breakpoint in the status text writing code, I'll be able to identify what method calls it, and start identifying the structure of items that likely have a "description" as one of it's fields.

All this progress today came from a minor diversion to pass the time sitting in the airport. The notice of an incorrect parameter to a method call has unraveled more of the game's secrets, and giving me multiple further areas I can investigate. A nice way to start the holidays.

Saturday, December 15, 2018

It's time for a very verbose change of pace

One of the things I like about my work on ScummVM is that it's a hobby rather than a profession, which gives me the freedom to switch what I'm working on when I desire. I've previously had encouragement to resurrect my Gargoyle engine work from about a decade ago. It was an early attempt to add support for Frotz into ScummVM, and at the time it was rejected due to multiple reasons. It was specific only for Infocom/zcode games, lacked clipboard, truetype fonts, and Unicode support. Most of these things have since been solved by the ScummVM framework itself, which now supports all of them. So I thought, why the heck not, and decided to work on it again.

However, my original Gargoyle engine being limited to just Zcode was a problem. Not only wasn't it extensible to other IF systems, but it would also take work to properly reimplement it to support Unicode. In the end, it was just simpler to scrap it all, and start from scratch. Luckily, there's the very convenient Glk specification, that was specifically created for easing porting the various Interactive Fiction interpreters to other systems. I knew that by supporting it, it would make it easier to create a framework that could support multiple different interpreters as sub-engines.

So I created a new Gargoyle engine, which I've since renamed it to ScummGlk to avoid ambiguity with the stand-alone Gargoyle project which also implements the Glk specification. After over a month of work, I've created an engine that implements the bulk of the API, and implements two sub-engines under it so far.. Scott for Scott Adams games, and Frotz for playing Z-Code/Infocom games. Scott was useful as a test case, due to it's simplicity. With it as the first sub-engine, I was able to focus on getting the ScummGlk core working. Then Frotz was a fairly obvious second choice, given it's popularity and large number of available fan games.

As of last weekend, the review period was completed for the engine, and it was merged into master. Daily builds now include support for both sub-engines. I've spend the last week fleshing out detection entries for all the ZCode games on the if-archive, so all the games will be automatically detected by the launcher. I'm going to wait on announcing an official testing period until I finish support in Frotz for V6 games. But if you can't wait, you're certainly welcome to jump the gun and try playing some games. Maybe one of the early Zork from GOG, or the immortal classic Freefall :)

Having the engine included into master will be a real nostalgia trip for me. Even prior to my original Gargoyle module, and before I even started working on implementing adventures in ScummVM, I spent several years as part of the Interactive Fiction community. Indeed, the Hitchhiker's Guide to the Galaxy was one of the first computer games I ever had. I spent way too many hours getting frustrated with the game's puzzles.