Leaked Original SCI Source

[lance.ewing]

There are also references to other SCI manuals.

Ah, yes. Just spotted the reference to "Object Oriented Programming in Script". Is that the only one?

[lance.ewing]

Did you notice that break, breakif, continue and continueif take an optional parameter to indicate the number of loop levels to break/continue from?

That might actually explain some of the weird compiled code we see from time to time. We were thinking there might be a continue, but there seems to be a number of variations. Good luck to the decompiler.

[Kawa]

Except for the part where Phil doesn't want to touch the decompiler

[OmerMor]

lance: You can go ahead and use this material.

I'll probably share some more docs.

[troflip]

Hmmm. You know Omer does imply he has a lot of this code but can't share most of it. But maybe an extraction of all of the original class names and procedure names, particularly the ones from the common framework, would be very useful for the decompiler.

The problem with this is that it would be game-specific. I mean, the "long class names" seem fairly consistent over different games (but the short names are already pretty easy to understand anyway, so not much added value). Having public procedure names would be great, but that will only work if the procedure is at the same export index in the same script number in every game. There would undoubtedly be a lot of mismatches if this were done automatically by the decompiler (nothing stopping you from renaming them as you decompile though, based on some suggested list).

[OmerMor]

Here's the Kernel document.

BTW - the SCI sources I have are from version 1.001.099.

[troflip]

That might actually explain some of the weird compiled code we see from time to time. We were thinking there might be a continue, but there seems to be a number of variations. Good luck to the decompiler.

That's undoubtedly true. Multi-level break/continue, and just regular continue especially, are not patterns the decompiler expects to see, nor is it feasible to reconstruct them in SCI Studio syntax.

[lance.ewing]

BTW - the SCI sources I have are from version 1.001.099.

So version 1.001.099 I can't seem to find any references to at all on the Internet. Ignoring the PQ4 and GK demos, the closest for a full game is Freddy Pharkas. I'm guessing this is where it is from then.

[OmerMor]

From going over the sources/docs it seems that the SCI infrastructure consisted of:

• Interpreter
• System Classes
• Tools (compiler, etc.)

This set evolved with time, and games just took the version that was available at that time and worked with it. Perhaps they might sync to a newer version later. But the team that wrote the infrastructure advanced it regardless of specific games. It was an ongoing project, developed in parallel to the games.

[OmerMor]

Ah, yes. Just spotted the reference to "Object Oriented Programming in Script". Is that the only one?

This is the one.

[lance.ewing]

The SCRIPT, KERNEL, and OBJECT documents all refer, in multiple places, to yet another document called "Script Classes for Adventure Games".   

[lance.ewing]

Spotted another one. "Vocabulary documentation" is referred to twice in the SCRIPT document.

[Collector]

Omer, would you mind putting some of this on the Wiki, at least what you are comfortable doing?

[troflip]

Any guesses as to why a comma was used to separate selector send params? Seems redundant, since the boundary could easily be detected by encountering the next selector name with : at the end of it.

[lance.ewing]

Yeah, that's a good question. It isn't consistent. The OBJECT pdf has an example without using the commas, from page 11, which suggests it wasn't required, at least not at first:

Code: [Select]

Any number of messages can be sent in one fell swoop:

(ego
    x:50
    y:50
    setMotion: MoveTo 100 100
    setCycle: Reverse self
)

Using a comma feels impure to me. It distracts from the smalltalk/Objective-C keyword message syntax. Maybe the comma was optional. Maybe some developers preferred using it. It doesn't seem to make sense as a requirement, unless there were a possibility of encountering a colon in a place where it doesn't follow a selector name.