[freesci-develop] SCI 1.1 support in Glutton

[Lars Skovlund]

Hi all!

I have postponed announcing this for a while, but support for SCI 1.1
is now nominally included in the Glutton branch. This means that games
such as Space Quest 5, King's Quest 6 and others have a chance of
running. You'll need to use the -r7 switch; I have left out
autodetection support because the games are still very flaky. There is
a number of known bugs and missing features, and calling the games
playable would be an overstatement: 

- The X driver crashes when playing SCI 1.1 games; this is presumably
  due to a bug in the implementation of colored mouse pointers, which
  we haven't had a way to test for real until now. 

  Solution: Use the SDL driver instead.

- No message file support. This means that the games can't show any
text (they show the current save dir instead; I realize I should
probably have written a proper dummy message). A prerequisite to
implementing this would be an extension to the resource manager so it
can load from other sources than RESOURCE.000/RESOURCE.MAP
(RESOURCE.MSG/MESSAGE.MAP in particular).

- Many CD games can be expected to lock up immediately if the speech
is not disabled; the responsible kernel call has not been implemented,
but I'll try to put together a stub. Otherwise same prerequisites as
above (for this, we need support for map resources contained within
the standard resource volume; in addition, a different numbering
scheme is used for this kind of audio resources - this is described in
a document I wrote which for some reason hasn't been added to the
official docs yet).

- Speed problems. This likely points to a flawed design change made by
Sierra in their later games, which might also account for the
general rise in the number of speed-related bugs in the original games
from SCI1.1 on.

- Ye Olde SCI1 User Interface Bug(tm) also affects SCI1.1 games. I
have been testing two different approaches to fixing this bug. At
this point it seems there is no "best" way to fix it given our gfx
subsystem; there might be a "least bad" way which varies with each
game (yuck...).

- Still no palette cycling/cross-fading/brightness (but this is mostly
cosmetic; I can't think of any puzzles that depend on it), let alone
SCI32-specific palette features.

- Song playing with adlibemu still acts funny in places (but General
MIDI should work with SCI 1.1; please test).

- Garbage collection is sometimes performed at inconvenient times,
leading to VM crashes (observed in CotL where it results in a null
pointer being dereferenced).

- And probably many others...

On the bright side, I have managed to put together a cross-compilation
framework. This should enable us to provide timely releases for our
Windows-based users.

Lars