SCI Companion V3 - alpha build notes/bugs/feature requests

[troflip]

Yeah, Sierra used that a lot to check ranges.

Code: [Select]

(if (< 5 (ego x?) 20)
    (Prints "Ego x is between 5 and 20")
)

It's one of the few places that uses the pprev register, and the decompiler doesn't handle that properly (and it's not trivial to fix).


http://scicompanion.com/Documentation/Compiler/relationaloperators.html

[lskovlun]

wait what

I can understand (+ 1 2 3) and such but <?

Code: [Select]

if 2 < a and a < 5 then ....
becpmes

Code: [Select]

(if (< 2 a 5) ...)

[OmerMor]

Reminds me of Python's chained comparisons: if (a < b < c) ...

[lskovlun]

Reminds of Python's chained comparisons: if (a < b < c) ...

Python has that? *goes to test* ... hm, sure does. I learned something new there.

[Kawa]

I am intrigued.

[OmerMor]

Reminds of Python's chained comparisons: if (a < b < c) ...

Python has that? *goes to test* ... hm, sure does. I learned something new there.

Yeah, I was shocked the first time I learned that. Now I wish every other language had that as well... 

[OmerMor]

Minor bug:
When decompiling both "old" and "new" versions of KQ4-SCI (1.000.111 / 1.006.004), script #120, I noticed that the old version sometimes confuse properties with methods, and call class methods with getter notation '?'.
For example, in introSc::changeState()

Old:

Code: [Select]

      (addToPics dispose?)

New:

Code: [Select]

      (addToPics dispose:)

Their disassembly is the same:

Code: [Select]

      pushi    #dispose
      pushi    0
      lag      addToPics
      send     4

the dispose method is declared in the Collect class, and is decompiled correctly in both versions.
This behavior is not consistent: sometimes method are called correctly with the ':' notation:

Code: [Select]

(hatActor hide: stopUpd:)
I am using the Sierra Script flavor in both games.

[troflip]

Well it doesn't really matter from the compiler's view point, as you know. But it's still weird that this is happening.

I'm confused by what you're saying is inconsistent... are you saying that the same method (dispose) is being output with both ? and : notation within the same version of a game?

The logic should be that a ':' is used when there are more than zero parameters passed, or when a method of that selector name is implemented on any object in the game and never used as a property on any object in the game. Given that dispose() is almost certainly a method on some objects in any version of KQ4, it should always be output with ':' notation. So it looks like I have a bug somewhere.

[OmerMor]

are you saying that the same method (dispose) is being output with both ? and : notation within the same version of a game?

Yes. Even for the same objects:

Room29.sc:

Code: [Select]

(if (timers contains: local4) (local4 dispose?))
Room3.sc:

Code: [Select]

(if (timers contains: local5) (local5 dispose: delete:))
Both scripts from version 1.000.111.

[troflip]

Ok, so the early KQ4 uses a slightly different script resource format. I'm surprised it even decompiles :-). At any rate, my code that attempts to discern props vs methods doesn't read the selectors properly.

I actually doubt anything will compile successfully unless I do more work to make that happen. I could at least put up a warning or something saying compiling to this version of SCI isn't supported.

Although it might not be too much work to support it...

fyi, SV.exe doesn't even show KQ4 scripts...

[OmerMor]

fyi, SV.exe doesn't even show KQ4 scripts...

Yes, I know - I assumed you had special treatment for early KQ4, and it was related to the bug.

Just wanted to repeat that this is a very minor bug - so don't bother too much with it. I just wanted to let you know.

[Collector]

That is only with the earlier versions of the game. SV does see the scripts in later versions. I do not have all versions, but I suspect that it does from 1.003.006 on. SV does not see the scripts in the earlier versions 1.000.106 and 1.000.111. They also do not support patch files. 1.000.111 seems to be one of the more commonly available, so I assume that is the version that Phil has.

[troflip]

Correct, 1.000.111 is the KQ4 I have.

[lskovlun]

There is one very old version of LSL2 that uses the old script format as well.

[Kawa]

I can see the scripts in SV's list, but I can't actually see the disassemblies. So I guess I have those KQ4 and LSL2 versions.