AGOS: implement Accolade AdLib + MT32 music drivers

- both known variants are supported (INSTR.DAT + MUSIC.DRV)
- INSTR.DAT/MUSIC.DRV holds channel mapping, instrument mapping, etc.
- fixed bug inside S1D MidiParser, that ruined some instrument changes
0xFC header was seen as 2 byte header, but it's 4 bytes in Elvira 2
and 5 bytes in Waxworks / Simon 1 demo
- dynamic channel allocation for the MUSIC.DRV adlib driver is not
implemented atm, simply because at least the demos of Waxworks and
Simon 1 do not use this feature
- sound effects of Waxworks are not implemented atm
- note: the game "Altered Destiny" uses Accolade INSTR.DAT variant too

1 files changed, 16 insertions, 4 deletions

diff --git a/engines/agos/midiparser_s1d.cpp b/engines/agos/midiparser_s1d.cpp
index c2c08bf451..f07ef5be41 100644
--- a/engines/agos/midiparser_s1d.cpp
+++ b/engines/agos/midiparser_s1d.cpp
@@ -181,10 +181,22 @@ bool MidiParser_S1D::loadMusic(byte *data, uint32 size) {
 
 	// The original actually just ignores the first two bytes.
 	byte *pos = data;
-	if (*(pos++) != 0xFC)
-		debug(1, "Expected 0xFC header but found 0x%02X instead", (int) *pos);
-
-	pos += 1;
+	if (*pos == 0xFC) {
+		// SysEx found right at the start
+		// this seems to happen since Elvira 2, we currently ignore it
+		// the original Accolade code does see 0xFC as end of track, which means there must have been a change
+		if ((pos[1] == 0x29) && (pos[2] == 0x07) && (pos[3] == 0x01)) {
+			// Security check
+			// Last byte is either 0x00 or 0x01. Maybe some looping indicator?
+			pos += 5; // Waxworks / Simon 1 demo
+		} else {
+			if ((pos[1] == 0x04) && (pos[2] == 0x06) && (pos[3] == 06)) {
+				pos += 4; // Elvira 2
+			} else {
+				warning("0xFC startup without proper signature");
+			}
+		}
+	}
 
 	// And now we're at the actual data. Only one track.
 	_numTracks = 1;