[Scummvm-devel] SCI versions...

[Max Horn]

Am 14.05.2009 um 16:30 schrieb Filippos Karapetis:

[...]

> Well, the way that SCI versions are implemented, there are certain  
> games
> which don't work with the detected SCI version, which kind of  
> defeats the
> whole purpose of relying on the interpreter version in the first  
> place.
>
> An example is Conquests of Camelot (Amiga), where the detected  
> version is
> 1.002.030, but it's actually version 0.000.685. Same goes for Longbow:
> detected version is 1.005.001, with the actual version being  
> 1.000.510.
> Then, there are the weird version numbers for KQ1 (S.old.010), Eco  
> Quest
> (1.ECO.013), SQ4 (1.SQ4.057), QFG2 (L.rry.083), Pharkas (l.cfs.081),  
> where
> we actually guess what the SCI version should be (and we don't know  
> for
> some cases, e.g. Pharkas).

Then we shouldn't use these versions at all (and only keep them in  
comments along with the detection entries; I still consider that  
useful for purposes of understanding the history of these games).



> As you can see, the current engine version scheme is problematic,  
> and it
> will get more complicated while support for new games is added in.  
> Plus, there
> are features which existed only in certain games, e.g. the way that  
> windows
> are handled in PQ3.

Sounds like a perfect argument for using a combination of feature  
flags, gameids ("pq3") and *major* VM revisions (0, 1, 1.1, 32) only,  
not the interpreter versions.
And this data should be encoded in the detection entries.

Of course it would be nice if we were able to auto-detect all of this,  
but if we aren't, well, so be it :).


Bye,
Max