The DOS versions of Simon the Sorcerer 2 now use the Miles drivers. AdLib now
sounds like the original does. ScummVM would use the second set of XMIDI data,
written for the MT-32, for GM with instrument remapping. Now the first set is
used, which uses GM instruments and has extra instruments on several pieces.
For all versions music fade-outs have been added during screen fades.
MIDI music is now paused by pausing the parser(s) instead of relying on the
mixer streams to stop providing callbacks. This gives more consistent behavior
as external MIDI devices or softsynths now also pause playback.
Muting sound now sets the volume to 0 instead of pausing playback.
This updates the Simon 1 DOS AdLib driver and adds an AdLib driver for Windows.
The DOS driver now has OPL3 support, which eliminates the note cut off that
frequently occurs on an OPL2. It also allows for some limited use of stereo.
The driver now plays the floppy SFX accurately compared to the original
interpreter. It also implements multisource functionality, allowing for
separate volume control for music and SFX for the floppy version.
The Windows driver just maps the MT-32 instruments of the Simon 1 MIDI data to
the equivalent GM instruments.
This commit adds MIDI parser subclasses for the Simon 1 GMF format and for the
SMF variant used by the Windows versions of Simon 1 and 2.
The GMF format was handled by the SMF parser, which can now be removed from
that class. It also fixes the tempos not matching the DOS interpreter.
The Simon Windows SMF variant was handled in the AGOS midi class. Moving it to
a separate parser allows for some cleaner code in that class. It now also
corrects the tempos to match those of the DOS interpreter.
At this point, the driver still tries to play notes at the maximum
allowed volume or louder, but ScummVM's music volume setting can bring
them back down to the expected range. That may have to be good enough
for now.
Also, since the master volume is only used internally there's no need to
keep converting back and forth. Just use the interval 0-255 throughout.
After feedback from NMIError, and some more guesswork. The master volume
isn't applied until noteOnSetVolume(), so that the correct adjusted
volume can be stored in the channel data structure.
Now the volume seems too low to me, but of course that could be fixed by
having some other default value than 15 for _masterVolume.
Because only the unused channelVolueAdjust value was clipped, the actual
velocity of the note would often overflow, causing the note to be almost
muted instead of played at full volume.
I assume it was an oversight that chanelVolumeAdjust wasn't used, so
I've fixed that and moved the clipping to after all adjustments have
been made. Also eliminated some unneeded floating-point math.
Is it correct now? I honestly don't know. It seems that, at least in
Waxworks, almost all notes end up being played at max volume. But maybe
that's how it should be?
The autosave refactoring that was done in
7adad5aaf5 used g_engine for identifying the
autosave slot. This worked for in-game save/load, but doesn't fit when
called from the launcher.
Fix by passing MetaEngine to SaveStateDescriptor ctor and using it for this
query.
Amends 7adad5aaf5.
This employs a "lazy" approach: the "format" for the credits stays
exactly as it was, i.e., perl code. Of course one may want to change
this to another format (e.g. YAML, JSON, XML; or also shell script or
AWK, like `configure.engine` uses). But I deliberately kept it simple,
to get a minimal change that is easy to verify. Any further changes to
e.g. the format can be layered atop this.
I have also made the necessary changes to the PC-98 audio drivers which were written to work around the current situation, but now also need a different order in the destructors.
The whole restoreBlock() function looked strange. It said it took x, y,
w and h, but it seems pretty clear that what it expects is left, top,
right and bottom.
Except in one case, where the order of the parameters had been swapped.
The most visible result of all this was that The Feeble Files crashed
when you pressed the "Off" button in the Oracle interface.
I've changed the names of the parameters, and that one strange call, and
changed how the function calculates the dirty rect. Hopefully that
should be correct.
Added support for InstallShield v6 cabinets, which can be made up of
multiple files. The interface for creating an Archive instance now takes
the base filename (e.g. the "data" in "data1.cab") for all cabinets,
including single-file ones.
Co-Authored-By: clone2727 <236052+clone2727@users.noreply.github.com>
Co-Authored-By: Walter van Niftrik <615114+waltervn@users.noreply.github.com>
- Fix messed up inventory (bug no. 11998) caused by the asynchronous (outsize of the normal engine) handling of the mouse wheel. This often lead to wrong inventory item icons.
- Also prevent crash due to nullptr deref. This can happen when using the mouse wheel while the inventory items are not visible (e. g. during the copy protection, when the symbol and words for the code wheel are displayed in the inventory window).
I assume that this happens accidently, since the mouse wheel was never supported by the original. Currently, you can still scroll the inventory using the mouse wheel when a save/load/pause dialog is open. It causes a text glitch in the PC-98 version. Instead of just fixing that glitch I've decided to fix the underlying bug (limited to Elvira 1 though; no idea what the behavior is in the other games).
Just walking around the castle fixing bugs as I find them, e. g.:
- make decompression routine safer (this could write a couple of invalid byte in at least one case)
- also support compressed data header without actual data (required in the kitchen scene when killing the evil cook)
- fix some graphics/palette related glitches
- Add dual layer graphics display to display the Japanese characters in their proper (double) resolution on top of the graphics. The original actually uses text mode.
- Adapt mouse handling to dual layer code.
- For consistency I got rid of all direct calls to _system->lockScreen() and _system->unlockScreen(), even in the sub engines that aren't affected. I find it easier to debug and to maintain this way.
